public-mirrors/linux
1
0
Fork 0
mirror of git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git synced 2025-08-05 16:54:27 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
linux/tools/power/x86/turbostat/turbostat.c

10143 lines
268 KiB
C
Raw Permalink Normal View History

treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 335 Based on 1 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms and conditions of the gnu general public license version 2 as published by the free software foundation this program is distributed in the hope it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details you should have received a copy of the gnu general public license along with this program if not write to the free software foundation inc 51 franklin st fifth floor boston ma 02110 1301 usa extracted by the scancode license scanner the SPDX license identifier GPL-2.0-only has been chosen to replace the boilerplate/reference in 111 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Alexios Zavras <alexios.zavras@intel.com> Reviewed-by: Allison Randal <allison@lohutok.net> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190530000436.567572064@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2019-05-29 16:57:49 -07:00
// SPDX-License-Identifier: GPL-2.0-only
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
/*
* turbostat -- show CPU frequency and C-state residency
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
* on modern Intel and AMD processors.
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
*
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
* Copyright (c) 2024 Intel Corporation.
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
* Len Brown <len.brown@intel.com>
*/
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 21:44:40 -04:00
#define _GNU_SOURCE
turbostat: Don't put unprocessed uapi headers in the include path turbostat's Makefile puts arch/x86/include/uapi/ in the include path, so that it can include <asm/msr.h> from it. It isn't in general safe to include even uapi headers directly from the kernel tree without processing them through scripts/headers_install.sh, but asm/msr.h happens to work. However, that include path can break with some versions of system headers, by overriding some system headers with the unprocessed versions directly from the kernel source. For instance: In file included from /build/x86-generic/usr/include/bits/sigcontext.h:28:0, from /build/x86-generic/usr/include/signal.h:339, from /build/x86-generic/usr/include/sys/wait.h:31, from turbostat.c:27: ../../../../arch/x86/include/uapi/asm/sigcontext.h:4:28: fatal error: linux/compiler.h: No such file or directory This occurs because the system bits/sigcontext.h on that build system includes <asm/sigcontext.h>, and asm/sigcontext.h in the kernel source includes <linux/compiler.h>, which scripts/headers_install.sh would have filtered out. Since turbostat really only wants a single header, just include that one header rather than putting an entire directory of kernel headers on the include path. In the process, switch from msr.h to msr-index.h, since turbostat just wants the MSR numbers. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Cc: stable@vger.kernel.org Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:12 -07:00
#include MSRHEADER
tools/power/turbostat: Switch to new Intel CPU model defines New CPU #defines encode vendor and family as well as model. N.B. Copied VFM_*() defines here from <asm/cpu_device_id.h> to avoid an application picking a second internal kernel header file. Signed-off-by: Tony Luck <tony.luck@intel.com> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-06-11 10:19:50 -07:00
// copied from arch/x86/include/asm/cpu_device_id.h
#define VFM_MODEL_BIT 0
#define VFM_FAMILY_BIT 8
#define VFM_VENDOR_BIT 16
#define VFM_RSVD_BIT 24
#define VFM_MODEL_MASK GENMASK(VFM_FAMILY_BIT - 1, VFM_MODEL_BIT)
#define VFM_FAMILY_MASK GENMASK(VFM_VENDOR_BIT - 1, VFM_FAMILY_BIT)
#define VFM_VENDOR_MASK GENMASK(VFM_RSVD_BIT - 1, VFM_VENDOR_BIT)
#define VFM_MODEL(vfm) (((vfm) & VFM_MODEL_MASK) >> VFM_MODEL_BIT)
#define VFM_FAMILY(vfm) (((vfm) & VFM_FAMILY_MASK) >> VFM_FAMILY_BIT)
#define VFM_VENDOR(vfm) (((vfm) & VFM_VENDOR_MASK) >> VFM_VENDOR_BIT)
#define VFM_MAKE(_vendor, _family, _model) ( \
((_model) << VFM_MODEL_BIT) | \
((_family) << VFM_FAMILY_BIT) | \
((_vendor) << VFM_VENDOR_BIT) \
)
// end copied section
tools/power turbostat: Allow using cpu device in perf counters on hybrid platforms Intel hybrid platforms expose different perf devices for P and E cores. Instead of one, "/sys/bus/event_source/devices/cpu" device, there are "/sys/bus/event_source/devices/{cpu_core,cpu_atom}". This, however makes it more complicated for the user, because most of the counters are available on both and had to be handled manually. This patch allows users to use "virtual" cpu device that is seemingly translated to cpu_core and cpu_atom perf devices, depending on the type of a CPU we are opening the counter for. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-20 18:47:59 +02:00
#define CPUID_LEAF_MODEL_ID 0x1A
#define CPUID_LEAF_MODEL_ID_CORE_TYPE_SHIFT 24
tools/power/turbostat: Switch to new Intel CPU model defines New CPU #defines encode vendor and family as well as model. N.B. Copied VFM_*() defines here from <asm/cpu_device_id.h> to avoid an application picking a second internal kernel header file. Signed-off-by: Tony Luck <tony.luck@intel.com> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-06-11 10:19:50 -07:00
#define X86_VENDOR_INTEL 0
tools/power turbostat: use intel-family.h model strings All except for model 1F, a Nehalem, which is currently incorrectly indentified as a Westmere in that new header. Signed-off-by: Len Brown <len.brown@intel.com>
2016-06-16 23:22:37 -04:00
#include INTEL_FAMILY_HEADER
tools/power turbostat: Add local build_bug.h header for snapshot target Fixes compilation errors for Makefile snapshot target described in: commit 231ce08b662a ("tools/power turbostat: Add "snapshot:" Makefile target") Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-30 09:16:39 +02:00
#include BUILD_BUG_HEADER
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:16 -07:00
#include <stdarg.h>
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
#include <stdio.h>
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:18 -07:00
#include <err.h>
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
#include <unistd.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <sys/stat.h>
tools/power turbostat: end current interval upon newline input In turbostat interval mode, a newline typed on standard input will now conclude the current interval. Data will immediately be collected and printed for that interval, and the next interval will be started. This is similar to the recently added SIGUSR1 feature. But that is for use by programs, while this is for interactive use. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-19 19:28:37 -04:00
#include <sys/select.h>
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
#include <sys/resource.h>
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
#include <sys/mman.h>
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
#include <fcntl.h>
#include <signal.h>
#include <sys/time.h>
#include <stdlib.h>
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
#include <getopt.h>
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
#include <dirent.h>
#include <string.h>
#include <ctype.h>
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 21:44:40 -04:00
#include <sched.h>
tools/power turbostat: allow sub-sec intervals turbostat -i interval_sec will sample and display statistics every interval_sec. interval_sec used to be a whole number of seconds, but now we accept a decimal, as small as 0.001 sec (1 ms). Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-12 22:44:48 -05:00
#include <time.h>
turbostat: Use GCC's CPUID functions to support PIC turbostat uses inline assembly to call cpuid. On 32-bit x86, on systems that have certain security features enabled by default that make -fPIC the default, this causes a build error: turbostat.c: In function ‘check_cpuid’: turbostat.c:1906:2: error: PIC register clobbered by ‘ebx’ in ‘asm’ asm("cpuid" : "=a" (fms), "=c" (ecx), "=d" (edx) : "a" (1) : "ebx"); ^ GCC provides a header cpuid.h, containing a __get_cpuid function that works with both PIC and non-PIC. (On PIC, it saves and restores ebx around the cpuid instruction.) Use that instead. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Cc: stable@vger.kernel.org Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:14 -07:00
#include <cpuid.h>
tools/power turbostat: Fix 32-bit capabilities warning warning: `turbostat' uses 32-bit capabilities (legacy support in use) Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 23:24:17 -04:00
#include <sys/capability.h>
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 00:36:50 -04:00
#include <errno.h>
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
#include <math.h>
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
#include <linux/perf_event.h>
#include <asm/unistd.h>
tools/power turbostat: Fix Core C6 residency on Atom CPUs For Atom CPUs that have core cstate deeper than C6, MSR_CORE_C6_RESIDENCY actually returns the residency for both CC6 and deeper Core cstates. Thus, the real Core C6 residency should be the subtraction of MSR_CORE_C6_RESIDENCY return value and MSR_CORE_C6_RESIDENCY return value. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-21 22:22:47 +08:00
#include <stdbool.h>
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
#include <assert.h>
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
#include <linux/kernel.h>
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: No build warnings with -Wextra Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 23:45:18 -04:00
#define UNUSED(x) (void)(x)
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
/*
* This list matches the column headers, except
* 1. built-in only, the sysfs counters are not here -- we learn of those at run-time
* 2. Core and CPU are moved to the end, we can't have strings that contain them
* matching on them for --show and --hide.
*/
/*
* buffer size used by sscanf() for added column names
* Usually truncated to 7 characters, but also handles 18 columns for raw 64-bit counters
*/
#define NAME_BYTES 20
#define PATH_BYTES 128
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
#define PERF_NAME_BYTES 128
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
tools/power turbostat: Increase the limit for fd opened When running turbostat, a system with 512 cpus reaches the limit for maximum number of file descriptors that can be opened. To solve this problem, the limit is raised to 2^15, which is a large enough number. Below data is collected from AMD server systems while running turbostat: |-----------+-------------------------------| | # of cpus | # of opened fds for turbostat | |-----------+-------------------------------| | 128 | 260 | |-----------+-------------------------------| | 192 | 388 | |-----------+-------------------------------| | 512 | 1028 | |-----------+-------------------------------| So, the new max limit would be sufficient up to 2^14 cpus (but this also depends on how many counters are enabled). Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Wyes Karny <wyes.karny@amd.com> Signed-off-by: Len Brown <len.brown@intel.com>
2023-10-03 05:07:51 +00:00
#define MAX_NOFILE 0x8000
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
#define COUNTER_KIND_PERF_PREFIX "perf/"
#define COUNTER_KIND_PERF_PREFIX_LEN strlen(COUNTER_KIND_PERF_PREFIX)
#define PERF_DEV_NAME_BYTES 32
#define PERF_EVT_NAME_BYTES 32
tools/power turbostat: Allow using cpu device in perf counters on hybrid platforms Intel hybrid platforms expose different perf devices for P and E cores. Instead of one, "/sys/bus/event_source/devices/cpu" device, there are "/sys/bus/event_source/devices/{cpu_core,cpu_atom}". This, however makes it more complicated for the user, because most of the counters are available on both and had to be handled manually. This patch allows users to use "virtual" cpu device that is seemingly translated to cpu_core and cpu_atom perf devices, depending on the type of a CPU we are opening the counter for. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-20 18:47:59 +02:00
#define INTEL_ECORE_TYPE 0x20
#define INTEL_PCORE_TYPE 0x40
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
enum counter_scope { SCOPE_CPU, SCOPE_CORE, SCOPE_PACKAGE };
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
enum counter_type { COUNTER_ITEMS, COUNTER_CYCLES, COUNTER_SECONDS, COUNTER_USEC, COUNTER_K2M };
enum counter_format { FORMAT_RAW, FORMAT_DELTA, FORMAT_PERCENT, FORMAT_AVERAGE };
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
enum counter_source { COUNTER_SOURCE_NONE, COUNTER_SOURCE_PERF, COUNTER_SOURCE_MSR };
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
struct perf_counter_info {
struct perf_counter_info *next;
/* How to open the counter / What counter it is. */
char device[PERF_DEV_NAME_BYTES];
char event[PERF_EVT_NAME_BYTES];
/* How to show/format the counter. */
char name[PERF_NAME_BYTES];
unsigned int width;
enum counter_scope scope;
enum counter_type type;
enum counter_format format;
double scale;
/* For reading the counter. */
int *fd_perf_per_domain;
size_t num_domains;
};
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
struct sysfs_path {
char path[PATH_BYTES];
int id;
struct sysfs_path *next;
};
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
struct msr_counter {
unsigned int msr_num;
char name[NAME_BYTES];
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
struct sysfs_path *sp;
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
unsigned int width;
enum counter_type type;
enum counter_format format;
struct msr_counter *next;
unsigned int flags;
#define FLAGS_HIDE (1 << 0)
#define FLAGS_SHOW (1 << 1)
#define SYSFS_PERCPU (1 << 1)
};
struct msr_counter bic[] = {
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
{ 0x0, "usec", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Time_Of_Day_Seconds", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Package", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Node", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Avg_MHz", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Busy%", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Bzy_MHz", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "TSC_MHz", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "IRQ", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "SMI", NULL, 32, 0, FORMAT_DELTA, NULL, 0 },
{ 0x0, "sysfs", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "CPU%c1", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "CPU%c3", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "CPU%c6", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "CPU%c7", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "ThreadC", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "CoreTmp", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "CoreCnt", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "PkgTmp", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "GFX%rc6", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "GFXMHz", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Pkg%pc2", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Pkg%pc3", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Pkg%pc6", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Pkg%pc7", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Pkg%pc8", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Pkg%pc9", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Pk%pc10", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "CPU%LPI", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "SYS%LPI", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "PkgWatt", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "CorWatt", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "GFXWatt", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "PkgCnt", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "RAMWatt", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "PKG_%", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "RAM_%", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Pkg_J", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Cor_J", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "GFX_J", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "RAM_J", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Mod%c6", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Totl%C0", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Any%C0", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "GFX%C0", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "CPUGFX%", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Core", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "CPU", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "APIC", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "X2APIC", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Die", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "GFXAMHz", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "IPC", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "CoreThr", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "UncMHz", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "SAM%mc6", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "SAMMHz", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "SAMAMHz", NULL, 0, 0, 0, NULL, 0 },
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
{ 0x0, "Die%c6", NULL, 0, 0, 0, NULL, 0 },
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
{ 0x0, "SysWatt", NULL, 0, 0, 0, NULL, 0 },
{ 0x0, "Sys_J", NULL, 0, 0, 0, NULL, 0 },
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
};
#define MAX_BIC (sizeof(bic) / sizeof(struct msr_counter))
#define BIC_USEC (1ULL << 0)
#define BIC_TOD (1ULL << 1)
#define BIC_Package (1ULL << 2)
#define BIC_Node (1ULL << 3)
#define BIC_Avg_MHz (1ULL << 4)
#define BIC_Busy (1ULL << 5)
#define BIC_Bzy_MHz (1ULL << 6)
#define BIC_TSC_MHz (1ULL << 7)
#define BIC_IRQ (1ULL << 8)
#define BIC_SMI (1ULL << 9)
#define BIC_sysfs (1ULL << 10)
#define BIC_CPU_c1 (1ULL << 11)
#define BIC_CPU_c3 (1ULL << 12)
#define BIC_CPU_c6 (1ULL << 13)
#define BIC_CPU_c7 (1ULL << 14)
#define BIC_ThreadC (1ULL << 15)
#define BIC_CoreTmp (1ULL << 16)
#define BIC_CoreCnt (1ULL << 17)
#define BIC_PkgTmp (1ULL << 18)
#define BIC_GFX_rc6 (1ULL << 19)
#define BIC_GFXMHz (1ULL << 20)
#define BIC_Pkgpc2 (1ULL << 21)
#define BIC_Pkgpc3 (1ULL << 22)
#define BIC_Pkgpc6 (1ULL << 23)
#define BIC_Pkgpc7 (1ULL << 24)
#define BIC_Pkgpc8 (1ULL << 25)
#define BIC_Pkgpc9 (1ULL << 26)
#define BIC_Pkgpc10 (1ULL << 27)
#define BIC_CPU_LPI (1ULL << 28)
#define BIC_SYS_LPI (1ULL << 29)
#define BIC_PkgWatt (1ULL << 30)
#define BIC_CorWatt (1ULL << 31)
#define BIC_GFXWatt (1ULL << 32)
#define BIC_PkgCnt (1ULL << 33)
#define BIC_RAMWatt (1ULL << 34)
#define BIC_PKG__ (1ULL << 35)
#define BIC_RAM__ (1ULL << 36)
#define BIC_Pkg_J (1ULL << 37)
#define BIC_Cor_J (1ULL << 38)
#define BIC_GFX_J (1ULL << 39)
#define BIC_RAM_J (1ULL << 40)
#define BIC_Mod_c6 (1ULL << 41)
#define BIC_Totl_c0 (1ULL << 42)
#define BIC_Any_c0 (1ULL << 43)
#define BIC_GFX_c0 (1ULL << 44)
#define BIC_CPUGFX (1ULL << 45)
#define BIC_Core (1ULL << 46)
#define BIC_CPU (1ULL << 47)
#define BIC_APIC (1ULL << 48)
#define BIC_X2APIC (1ULL << 49)
#define BIC_Die (1ULL << 50)
#define BIC_GFXACTMHz (1ULL << 51)
#define BIC_IPC (1ULL << 52)
#define BIC_CORE_THROT_CNT (1ULL << 53)
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
#define BIC_UNCORE_MHZ (1ULL << 54)
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
#define BIC_SAM_mc6 (1ULL << 55)
#define BIC_SAMMHz (1ULL << 56)
#define BIC_SAMACTMHz (1ULL << 57)
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
#define BIC_Diec6 (1ULL << 58)
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
#define BIC_SysWatt (1ULL << 59)
#define BIC_Sys_J (1ULL << 60)
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
#define BIC_TOPOLOGY (BIC_Package | BIC_Node | BIC_CoreCnt | BIC_PkgCnt | BIC_Core | BIC_CPU | BIC_Die )
#define BIC_THERMAL_PWR ( BIC_CoreTmp | BIC_PkgTmp | BIC_PkgWatt | BIC_CorWatt | BIC_GFXWatt | BIC_RAMWatt | BIC_PKG__ | BIC_RAM__)
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
#define BIC_FREQUENCY (BIC_Avg_MHz | BIC_Busy | BIC_Bzy_MHz | BIC_TSC_MHz | BIC_GFXMHz | BIC_GFXACTMHz | BIC_SAMMHz | BIC_SAMACTMHz | BIC_UNCORE_MHZ)
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
#define BIC_IDLE (BIC_sysfs | BIC_CPU_c1 | BIC_CPU_c3 | BIC_CPU_c6 | BIC_CPU_c7 | BIC_GFX_rc6 | BIC_Pkgpc2 | BIC_Pkgpc3 | BIC_Pkgpc6 | BIC_Pkgpc7 | BIC_Pkgpc8 | BIC_Pkgpc9 | BIC_Pkgpc10 | BIC_CPU_LPI | BIC_SYS_LPI | BIC_Mod_c6 | BIC_Totl_c0 | BIC_Any_c0 | BIC_GFX_c0 | BIC_CPUGFX | BIC_SAM_mc6 | BIC_Diec6)
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
#define BIC_OTHER ( BIC_IRQ | BIC_SMI | BIC_ThreadC | BIC_CoreTmp | BIC_IPC)
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
#define BIC_DISABLED_BY_DEFAULT (BIC_USEC | BIC_TOD | BIC_APIC | BIC_X2APIC | BIC_SysWatt | BIC_Sys_J)
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
unsigned long long bic_enabled = (0xFFFFFFFFFFFFFFFFULL & ~BIC_DISABLED_BY_DEFAULT);
unsigned long long bic_present = BIC_USEC | BIC_TOD | BIC_sysfs | BIC_APIC | BIC_X2APIC;
#define DO_BIC(COUNTER_NAME) (bic_enabled & bic_present & COUNTER_NAME)
#define DO_BIC_READ(COUNTER_NAME) (bic_present & COUNTER_NAME)
#define ENABLE_BIC(COUNTER_NAME) (bic_enabled |= COUNTER_NAME)
#define BIC_PRESENT(COUNTER_BIT) (bic_present |= COUNTER_BIT)
#define BIC_NOT_PRESENT(COUNTER_BIT) (bic_present &= ~COUNTER_BIT)
#define BIC_IS_ENABLED(COUNTER_BIT) (bic_enabled & COUNTER_BIT)
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
/*
* MSR_PKG_CST_CONFIG_CONTROL decoding for pkg_cstate_limit:
* If you change the values, note they are used both in comparisons
* (>= PCL__7) and to index pkg_cstate_limit_strings[].
*/
#define PCLUKN 0 /* Unknown */
#define PCLRSV 1 /* Reserved */
#define PCL__0 2 /* PC0 */
#define PCL__1 3 /* PC1 */
#define PCL__2 4 /* PC2 */
#define PCL__3 5 /* PC3 */
#define PCL__4 6 /* PC4 */
#define PCL__6 7 /* PC6 */
#define PCL_6N 8 /* PC6 No Retention */
#define PCL_6R 9 /* PC6 Retention */
#define PCL__7 10 /* PC7 */
#define PCL_7S 11 /* PC7 Shrink */
#define PCL__8 12 /* PC8 */
#define PCL__9 13 /* PC9 */
#define PCL_10 14 /* PC10 */
#define PCLUNL 15 /* Unlimited */
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
struct amperf_group_fd;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
char *proc_stat = "/proc/stat";
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
FILE *outf;
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
int *fd_percpu;
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
int *fd_instr_count_percpu;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
struct timeval interval_tv = { 5, 0 };
struct timespec interval_ts = { 5, 0 };
tools/power turbostat: save original CPU model CPU model may get changed in intel_model_duplicates() for code reuse. But there are still some cases we need the original CPU model to handle minor differences between generations. Thus save the original CPU model. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-26 10:05:27 +08:00
tools/power turbostat: if --num_iterations, print for specific number of iterations There's a use case during test to only print specific round of iterations if --num_iterations is specified, for example, with this patch applied: turbostat -i 5 -n 4 will capture 4 samples with 5 seconds interval. [lenb: renamed to --num_iterations from --iterations] Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Reviewed-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-04-26 08:41:03 +08:00
unsigned int num_iterations;
tools/power turbostat: Allow printing header every N iterations This gives the ability to reprint the header every N iterations, so you can ensure that a scrolling display always has the header visible somewhere on the screen. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:39 -07:00
unsigned int header_iterations;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
unsigned int debug;
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
unsigned int quiet;
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
unsigned int shown;
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
unsigned int sums_need_wide_columns;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
unsigned int rapl_joules;
unsigned int summary_only;
tools/power turbostat: Add --list option to show available header names It is handy to know the list of column header names, so that they can be used with --add and --skip The new --list option shows them: sudo ./turbostat --list --hide sysfs ,Core,CPU,Avg_MHz,Busy%,Bzy_MHz,TSC_MHz,IRQ,SMI,CPU%c1,CPU%c3,CPU%c6,CPU%c7,CoreTmp,PkgTmp,GFX%rc6,GFXMHz,PkgWatt,CorWatt,GFXWatt Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 17:15:11 -05:00
unsigned int list_header_only;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
unsigned int dump_only;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
unsigned int has_aperf;
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
unsigned int has_aperf_access;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
unsigned int has_epb;
tools/power turbostat: Do not dump TRL if turbo is not supported Do not dump turbo ratio limits if platform does not support turbo, because it is confusing and the TRL MSRs may even include misleading information. And they are not supposed to be relied on if turbo is not supported. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-09-16 14:59:50 +03:00
unsigned int has_turbo;
tools/power turbostat: dump CPUID.7.EDX.Hybrid CPUID leaf 7 EDX now tells us if the processor has hybrid CPUs Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-31 17:08:25 -10:00
unsigned int is_hybrid;
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 00:55:19 -05:00
unsigned int units = 1000000; /* MHz etc */
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
unsigned int genuine_intel;
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
unsigned int authentic_amd;
tools/power turbostat: Add support for Hygon Fam 18h (Dhyana) RAPL Commit 9392bd98bba760be96ee ("tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL") and the commit 3316f99a9f1b68c578c5 ("tools/power turbostat: Also read package power on AMD F17h (Zen)") add AMD Fam 17h RAPL support. Hygon Family 18h(Dhyana) support RAPL in bit 14 of CPUID 0x80000007 EDX, and has MSRs RAPL_PWR_UNIT/CORE_ENERGY_STAT/PKG_ENERGY_STAT. So add Hygon Dhyana Family 18h support for RAPL. Already tested on Hygon multi-node systems and it shows correct per-core energy usage and the total package power. Signed-off-by: Pu Wen <puwen@hygon.cn> Reviewed-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2019-08-31 10:20:31 +08:00
unsigned int hygon_genuine;
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
unsigned int max_level, max_extended_level;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
unsigned int has_invariant_tsc;
tools/power turbostat: KNL workaround for %Busy and Avg_MHz KNL increments APERF and MPERF every 1024 clocks. This is compliant with the architecture specification, which requires that only the ratio of APERF/MPERF need be valid. However, turbostat takes advantage of the fact that these two MSRs increment every un-halted clock at the actual and base frequency: AVG_MHz = APERF_delta/measurement_interval %Busy = MPERF_delta/TSC_delta This quirk is needed for these calculations to also work on KNL, which would otherwise show a value 1024x smaller than expected. Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-14 13:31:00 +02:00
unsigned int aperf_mperf_multiplier = 1;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
double bclk;
tools/power turbostat: SKL: Adjust for TSC difference from base frequency On a Skylake with 1500MHz base frequency, the TSC runs at 1512MHz. This is because the TSC is no longer in the n*100 MHz BCLK domain, but is now in the m*24MHz crystal clock domain. (24 MHz * 63 = 1512 MHz) This adds error to several calculations in turbostat, unless the TSC sample sizes are adjusted for this difference. Note that calculations in the time domain are immune from this issue, as the timing sub-system has already calibrated the TSC against a known wall clock. AVG_MHz = APERF_delta/measurement_interval need no adjustment. APERF_delta is in the BCLK domain, and measurement_interval is in the time domain. TSC_MHz = TSC_delta/measurement_interval needs no adjustment -- as we really do want to report the actual measured TSC delta here, and measurement_interval is in the accurate time domain. %Busy = MPERF_delta/TSC_delta needs adjustment to use TSC_BCLK_DOMAIN_delta. TSC_BCLK_DOMAIN_delta = TSC_delta * base_hz / tsc_hz Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/measurement_interval need adjustment as above. No other metrics in turbostat need to be adjusted. Before: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 24.84 2216 1512 0 2191 98.73 2219 1514 2 0 0.01 2130 1512 1 9 0.43 2016 1512 3 2 0.08 2016 1512 After: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 25.05 2198 1512 0 2190 99.62 2199 1512 2 0 0.01 2152 1512 1 9 0.46 2000 1512 3 2 0.10 2000 1512 Note that in this example, the "Before" Bzy_MHz was reported as exceeding the 2200 max turbo rate. Also, even a pinned spin loop would not be reported as over 99% busy. Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-26 00:12:38 -04:00
double base_hz;
tools/power turbostat: simplify Bzy_MHz calculation Bzy_MHz = TSC_delta*tsc_tweak/APERF_delta/MPERF_delta/measurement_interval becomes Bzy_MHz = base_mhz/APERF_delta/MPERF_delta on systems which support MSR_NHM_PLATFORM_INFO. base_mhz is calculated directly from the base_ratio reported in MSR_NHM_PLATFORM_INFO * bclk, and bclk is discovered via MSR or cpuid. This reduces the dependency of Bzy_MHz calculation on the TSC. Previously, there were 4 TSC readings required in each caculation, the raw TSC delta combined with the measurement_interval. This also removes the "tsc_tweak" correction factor used when TSC runs on a different base clock from the CPU's bclk. After this change, tsc_tweak is used only for %Busy. The end-result should be a Bzy_MHz result slightly less prone to jitter. Signed-off-by: Len Brown <len.brown@intel.com>
2015-10-19 22:37:40 -04:00
unsigned int has_base_hz;
tools/power turbostat: SKL: Adjust for TSC difference from base frequency On a Skylake with 1500MHz base frequency, the TSC runs at 1512MHz. This is because the TSC is no longer in the n*100 MHz BCLK domain, but is now in the m*24MHz crystal clock domain. (24 MHz * 63 = 1512 MHz) This adds error to several calculations in turbostat, unless the TSC sample sizes are adjusted for this difference. Note that calculations in the time domain are immune from this issue, as the timing sub-system has already calibrated the TSC against a known wall clock. AVG_MHz = APERF_delta/measurement_interval need no adjustment. APERF_delta is in the BCLK domain, and measurement_interval is in the time domain. TSC_MHz = TSC_delta/measurement_interval needs no adjustment -- as we really do want to report the actual measured TSC delta here, and measurement_interval is in the accurate time domain. %Busy = MPERF_delta/TSC_delta needs adjustment to use TSC_BCLK_DOMAIN_delta. TSC_BCLK_DOMAIN_delta = TSC_delta * base_hz / tsc_hz Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/measurement_interval need adjustment as above. No other metrics in turbostat need to be adjusted. Before: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 24.84 2216 1512 0 2191 98.73 2219 1514 2 0 0.01 2130 1512 1 9 0.43 2016 1512 3 2 0.08 2016 1512 After: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 25.05 2198 1512 0 2190 99.62 2199 1512 2 0 0.01 2152 1512 1 9 0.46 2000 1512 3 2 0.10 2000 1512 Note that in this example, the "Before" Bzy_MHz was reported as exceeding the 2200 max turbo rate. Also, even a pinned spin loop would not be reported as over 99% busy. Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-26 00:12:38 -04:00
double tsc_tweak = 1.0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
unsigned int show_pkg_only;
unsigned int show_core_only;
char *output_buffer, *outp;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
unsigned int do_dts;
unsigned int do_ptm;
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
unsigned int do_ipc;
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
unsigned long long cpuidle_cur_cpu_lpi_us;
unsigned long long cpuidle_cur_sys_lpi_us;
tools/power turbostat: rename tcc variables There are two TCC activation temeprature. One is the default TCC activation temperature, also known as TJ_MAX. Another one is the effective TCC activation temperature, which is the subtraction of default TCC activation temperature and TCC offset. The name of variable tcc_activation_temp might be misleading here. Thus rename tcc_activation_temp to tj_max, and use tcc_default and tcc_offset to calculate the effective TCC activation temperature. No functional change in this patch. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-26 18:49:26 +08:00
unsigned int tj_max;
unsigned int tj_max_override;
tools/power turbostat: correct DRAM RAPL units on recent Xeon processors While not yet documented in the Software Developer's Manual, the data-sheet for modern Xeon states that DRAM RAPL ENERGY units are fixed at 15.3 uJ, rather than being discovered via MSR. Before this patch, DRAM energy on these products is over-stated by turbostat because the RAPL units are 4x larger. ref: "Xeon E5-2600 v3/E5-1600 v3 Datasheet Volume 2" http://www.intel.com/content/dam/www/public/us/en/documents/datasheets/xeon-e5-v3-datasheet-vol-2.pdf Signed-off-by: Andrey Semin <andrey.semin@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-12-05 00:07:00 -05:00
double rapl_power_units, rapl_time_units;
double rapl_dram_energy_units, rapl_energy_units;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
double rapl_joule_counter_range;
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-01 21:02:57 -04:00
unsigned int crystal_hz;
unsigned long long tsc_hz;
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 08:34:28 -04:00
int base_cpu;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
unsigned int has_hwp; /* IA32_PM_ENABLE, IA32_HWP_CAPABILITIES */
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
/* IA32_HWP_REQUEST, IA32_HWP_STATUS */
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
unsigned int has_hwp_notify; /* IA32_HWP_INTERRUPT */
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
unsigned int has_hwp_activity_window; /* IA32_HWP_REQUEST[bits 41:32] */
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
unsigned int has_hwp_epp; /* IA32_HWP_REQUEST[bits 31:24] */
unsigned int has_hwp_pkg; /* IA32_HWP_REQUEST_PKG */
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
unsigned int first_counter_read = 1;
tools/power turbostat: read from pipes too Commit '47936f944e78 tools/power turbostat: fix printing on input' make a valid fix, but it completely disabled piped stdin support, which is a valuable use-case. Indeed, if stdin is a pipe, turbostat won't read anything from it, so it becomes impossible to get turbostat output at user-defined moments, instead of the regular intervals. There is no reason why this should works for terminals, but not for pipes. This patch improves the situation. Instead of ignoring pipes, we read data from them but gracefully handle the EOF case. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-08-14 20:12:55 +03:00
int ignore_stdin;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
bool no_msr;
tools/power turbostat: Add --no-perf option Add the --no-perf option to allow users to run turbostat without accessing perf. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:58:02 +01:00
bool no_perf;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power/turbostat: Cache graphics sysfs path Graphics drivers (i915/Xe) have different sysfs knobs on different platforms, and it is possible that different sysfs knobs fit into the same turbostat columns. Instead of specifying different sysfs knobs every time, detect them once and cache the path for future use. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-12 14:23:37 +08:00
enum gfx_sysfs_idx {
GFX_rc6,
GFX_MHz,
GFX_ACTMHz,
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
SAM_mc6,
SAM_MHz,
SAM_ACTMHz,
tools/power/turbostat: Cache graphics sysfs path Graphics drivers (i915/Xe) have different sysfs knobs on different platforms, and it is possible that different sysfs knobs fit into the same turbostat columns. Instead of specifying different sysfs knobs every time, detect them once and cache the path for future use. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-12 14:23:37 +08:00
GFX_MAX
};
struct gfx_sysfs_info {
tools/power/turbostat: Unify graphics sysfs snapshots Graphics sysfs snapshots share similar logic. Combine them into one function to avoid code duplication. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-13 10:12:19 +08:00
FILE *fp;
unsigned int val;
unsigned long long val_ull;
tools/power/turbostat: Cache graphics sysfs path Graphics drivers (i915/Xe) have different sysfs knobs on different platforms, and it is possible that different sysfs knobs fit into the same turbostat columns. Instead of specifying different sysfs knobs every time, detect them once and cache the path for future use. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-12 14:23:37 +08:00
};
static struct gfx_sysfs_info gfx_info[GFX_MAX];
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
int get_msr(int cpu, off_t offset, unsigned long long *msr);
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
int add_counter(unsigned int msr_num, char *path, char *name,
unsigned int width, enum counter_scope scope,
enum counter_type type, enum counter_format format, int flags, int package_num);
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
/* Model specific support Start */
/* List of features that may diverge among different platforms */
struct platform_features {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
bool has_msr_misc_feature_control; /* MSR_MISC_FEATURE_CONTROL */
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
bool has_msr_misc_pwr_mgmt; /* MSR_MISC_PWR_MGMT */
tools/power/turbostat: Relocate cstate probing code Move all cstate probing related code into probe_cstates(). Note that dump_platform_info() actually dumps both MSR_PLATFORM_INFO and MSR_IA32_POWER_CTL. MSR_PLATFORM_INFO is for pstate and MSR_IA32_POWER_CTL is for cstate. So split dump_platform_info() and dump MSR_IA32_POWER_CTL in probe_cstates(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-29 13:47:58 +08:00
bool has_nhm_msrs; /* MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, MSR_IA32_POWER_CTL, TRL MSRs */
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
bool has_config_tdp; /* MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL, MSR_TURBO_ACTIVATION_RATIO */
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
int bclk_freq; /* CPU base clock */
tools/power/turbostat: Abstract hardcoded Crystal Clock frequency Abstract the support for hardcoded Crystal Clock frequency, which is used when crystal clock is not available from CPUID.15. Delete CPU model checks in process_cpuid(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 11:39:49 +08:00
int crystal_freq; /* Crystal clock to use when not available from CPUID.15 */
tools/power/turbostat: Add skeleton support for cstate enumeration Add skeleton support for cstate enumeration. Note that the previous logic may override the cstate setting for multiple times for different reasons. The conversion to new cstate enumeration must be done step by step following the previous code order strictly. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:37 +08:00
int supported_cstates; /* Core cstates and Package cstates supported */
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
int cst_limit; /* MSR_PKG_CST_CONFIG_CONTROL */
tools/power/turbostat: Abstract Automatic Cstate Conversion support Abstract the support for AUTOMATIC_CSTATE_CONVERSION bit in MSR_PKG_CST_CONFIG_CONTROL. Delete automatic_cstate_conversion_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 11:29:18 +08:00
bool has_cst_auto_convension; /* AUTOMATIC_CSTATE_CONVERSION bit in MSR_PKG_CST_CONFIG_CONTROL */
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
bool has_irtl_msrs; /* MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL */
tools/power/turbostat: Abstract MSR_CORE_C1_RES support Abstract the support for MSR_CORE_C1_RES. Delete is_dnv() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:41:30 +08:00
bool has_msr_core_c1_res; /* MSR_CORE_C1_RES */
tools/power/turbostat: Abstract MSR_MODULE_C6_RES_MS support Abstract MSR_MODULE_C6_RES_MS support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:47:30 +08:00
bool has_msr_module_c6_res_ms; /* MSR_MODULE_C6_RES_MS */
tools/power/turbostat: Abstract MSR_CC6/MC6_DEMOTION_POLICY_CONFIG support Abstract the support for MSR_CC6/MC6_DEMOTION_POLICY_CONFIG. Delete has_slv_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:52:24 +08:00
bool has_msr_c6_demotion_policy_config; /* MSR_CC6_DEMOTION_POLICY_CONFIG/MSR_MC6_DEMOTION_POLICY_CONFIG */
tools/power/turbostat: Abstract MSR_ATOM_PKG_C6_RESIDENCY support Abstract the support for MSR_ATOM_PKG_C6_RESIDENCY. Delete is_slm() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 15:26:14 +08:00
bool has_msr_atom_pkg_c6_residency; /* MSR_ATOM_PKG_C6_RESIDENCY */
tools/power/turbostat: Abstract MSR_KNL_CORE_C6_RESIDENCY support Abstract the support for MSR_KNL_CORE_C6_RESIDENCY. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 15:00:58 +08:00
bool has_msr_knl_core_c6_residency; /* MSR_KNL_CORE_C6_RESIDENCY */
tools/power/turbostat: Abstract extended cstate MSRs support Abstract the support for MSR_PKG_WEIGHTED_CORE_C0_RES, MSR_PKG_ANY_CORE_C0_RES, MSR_PKG_ANY_GFXE_C0_RES and MSR_PKG_BOTH_CORE_GFXE_C0_RES. Delete has_skl_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 15:07:43 +08:00
bool has_ext_cst_msrs; /* MSR_PKG_WEIGHTED_CORE_C0_RES/MSR_PKG_ANY_CORE_C0_RES/MSR_PKG_ANY_GFXE_C0_RES/MSR_PKG_BOTH_CORE_GFXE_C0_RES */
tools/power/turbostat: Abstract cstate prewake bit support Abstract cstate prewake bit support. Delete is_icx()/is_spr() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 15:57:44 +08:00
bool has_cst_prewake_bit; /* Cstate prewake bit in MSR_IA32_POWER_CTL */
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
int trl_msrs; /* MSR_TURBO_RATIO_LIMIT/LIMIT1/LIMIT2/SECONDARY, Atom TRL MSRs */
tools/power/turbostat: Abstract Perf Limit Reasons MSRs support Abstract the support for MSR_CORE/GFX/RING_PERF_LIMIT_REASONS MSRs. Delete perf_limit_reasons_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-07-30 14:25:17 +08:00
int plr_msrs; /* MSR_CORE/GFX/RING_PERF_LIMIT_REASONS */
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
int rapl_msrs; /* RAPL PKG/DRAM/CORE/GFX MSRs, AMD RAPL MSRs */
tools/power/turbostat: Abstract Per Core RAPL support Abstract the support for Per Core RAPL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:37:46 +08:00
bool has_per_core_rapl; /* Indicates cores energy collection is per-core, not per-package. AMD specific for now */
tools/power/turbostat: Abstract RAPL divisor support INTEL_FAM6_ATOM_SILVERMONT model needs a divisor to convert the raw Energy Units value from MSR_RAPL_POWER_UNIT. Abstract the support for RAPL divisor. Delete CPU model check in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 11:59:04 +08:00
bool has_rapl_divisor; /* Divisor for Energy unit raw value from MSR_RAPL_POWER_UNIT */
tools/power/turbostat: Abstract fixed DRAM Energy unit support Abstract the support for fixed Dram domain energy unit. Delete rapl_dram_energy_units_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 12:21:58 +08:00
bool has_fixed_rapl_unit; /* Fixed Energy Unit used for DRAM RAPL Domain */
tools/power/turbostat: Abstract hardcoded TDP value Different hardcoded TDP values are used when TDP can not be retrieved from the hardware. Abstract hardcoded TDP value. Delete CPU model checks in get_tdp_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:21:10 +08:00
int rapl_quirk_tdp; /* Hardcoded TDP value when cannot be retrieved from hardware */
tools/power/turbostat: Abstract TCC Offset bits support Abstract the support for different TCC Offset bits in MSR_IA32_TEMPERATURE_TARGET. Delete check_tcc_offset() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 23:04:44 +08:00
int tcc_offset_bits; /* TCC Offset bits in MSR_IA32_TEMPERATURE_TARGET */
tools/power/turbostat: Abstract TSC tweak support On some models, the CPU base frequency is different from the TSC frequency, and the aperf/mperf counters are running at CPU base frequency instead of TSC frequency. Abstract support for TSC tweak. Given that tsc_tweak depends on base_hz, move the code to probe_bclk() after base_hz is available. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:37:23 +08:00
bool enable_tsc_tweak; /* Use CPU Base freq instead of TSC freq for aperf/mperf counter */
tools/power/turbostat: Abstract aperf/mperf multiplier support Abstract aperf/mperf multiplier support. Delete is_knl() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 15:49:10 +08:00
bool need_perf_multiplier; /* mperf/aperf multiplier */
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
struct platform_data {
tools/power/turbostat: Switch to new Intel CPU model defines New CPU #defines encode vendor and family as well as model. N.B. Copied VFM_*() defines here from <asm/cpu_device_id.h> to avoid an application picking a second internal kernel header file. Signed-off-by: Tony Luck <tony.luck@intel.com> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-06-11 10:19:50 -07:00
unsigned int vfm;
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
const struct platform_features *features;
};
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
/* For BCLK */
enum bclk_freq {
BCLK_100MHZ = 1,
BCLK_133MHZ,
BCLK_SLV,
};
#define SLM_BCLK_FREQS 5
double slm_freq_table[SLM_BCLK_FREQS] = { 83.3, 100.0, 133.3, 116.7, 80.0 };
double slm_bclk(void)
{
unsigned long long msr = 3;
unsigned int i;
double freq;
if (get_msr(base_cpu, MSR_FSB_FREQ, &msr))
fprintf(outf, "SLM BCLK: unknown\n");
i = msr & 0xf;
if (i >= SLM_BCLK_FREQS) {
fprintf(outf, "SLM BCLK[%d] invalid\n", i);
i = 3;
}
freq = slm_freq_table[i];
if (!quiet)
fprintf(outf, "SLM BCLK: %.1f Mhz\n", freq);
return freq;
}
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
/* For Package cstate limit */
enum package_cstate_limit {
CST_LIMIT_NHM = 1,
CST_LIMIT_SNB,
CST_LIMIT_HSW,
CST_LIMIT_SKX,
CST_LIMIT_ICX,
CST_LIMIT_SLV,
CST_LIMIT_AMT,
CST_LIMIT_KNL,
CST_LIMIT_GMT,
};
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
/* For Turbo Ratio Limit MSRs */
enum turbo_ratio_limit_msrs {
TRL_BASE = BIT(0),
TRL_LIMIT1 = BIT(1),
TRL_LIMIT2 = BIT(2),
TRL_ATOM = BIT(3),
TRL_KNL = BIT(4),
TRL_CORECOUNT = BIT(5),
};
tools/power/turbostat: Abstract Perf Limit Reasons MSRs support Abstract the support for MSR_CORE/GFX/RING_PERF_LIMIT_REASONS MSRs. Delete perf_limit_reasons_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-07-30 14:25:17 +08:00
/* For Perf Limit Reason MSRs */
enum perf_limit_reason_msrs {
PLR_CORE = BIT(0),
PLR_GFX = BIT(1),
PLR_RING = BIT(2),
};
tools/power/turbostat: Redefine RAPL macros Redefine RAPL macros to make the code more readable. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-26 14:38:38 +08:00
/* For RAPL MSRs */
enum rapl_msrs {
RAPL_PKG_POWER_LIMIT = BIT(0), /* 0x610 MSR_PKG_POWER_LIMIT */
RAPL_PKG_ENERGY_STATUS = BIT(1), /* 0x611 MSR_PKG_ENERGY_STATUS */
RAPL_PKG_PERF_STATUS = BIT(2), /* 0x613 MSR_PKG_PERF_STATUS */
RAPL_PKG_POWER_INFO = BIT(3), /* 0x614 MSR_PKG_POWER_INFO */
RAPL_DRAM_POWER_LIMIT = BIT(4), /* 0x618 MSR_DRAM_POWER_LIMIT */
RAPL_DRAM_ENERGY_STATUS = BIT(5), /* 0x619 MSR_DRAM_ENERGY_STATUS */
RAPL_DRAM_PERF_STATUS = BIT(6), /* 0x61b MSR_DRAM_PERF_STATUS */
RAPL_DRAM_POWER_INFO = BIT(7), /* 0x61c MSR_DRAM_POWER_INFO */
RAPL_CORE_POWER_LIMIT = BIT(8), /* 0x638 MSR_PP0_POWER_LIMIT */
RAPL_CORE_ENERGY_STATUS = BIT(9), /* 0x639 MSR_PP0_ENERGY_STATUS */
RAPL_CORE_POLICY = BIT(10), /* 0x63a MSR_PP0_POLICY */
RAPL_GFX_POWER_LIMIT = BIT(11), /* 0x640 MSR_PP1_POWER_LIMIT */
RAPL_GFX_ENERGY_STATUS = BIT(12), /* 0x641 MSR_PP1_ENERGY_STATUS */
RAPL_GFX_POLICY = BIT(13), /* 0x642 MSR_PP1_POLICY */
RAPL_AMD_PWR_UNIT = BIT(14), /* 0xc0010299 MSR_AMD_RAPL_POWER_UNIT */
RAPL_AMD_CORE_ENERGY_STAT = BIT(15), /* 0xc001029a MSR_AMD_CORE_ENERGY_STATUS */
RAPL_AMD_PKG_ENERGY_STAT = BIT(16), /* 0xc001029b MSR_AMD_PKG_ENERGY_STATUS */
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
RAPL_PLATFORM_ENERGY_LIMIT = BIT(17), /* 0x64c MSR_PLATFORM_ENERGY_LIMIT */
RAPL_PLATFORM_ENERGY_STATUS = BIT(18), /* 0x64d MSR_PLATFORM_ENERGY_STATUS */
tools/power/turbostat: Redefine RAPL macros Redefine RAPL macros to make the code more readable. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-26 14:38:38 +08:00
};
#define RAPL_PKG (RAPL_PKG_ENERGY_STATUS | RAPL_PKG_POWER_LIMIT)
#define RAPL_DRAM (RAPL_DRAM_ENERGY_STATUS | RAPL_DRAM_POWER_LIMIT)
#define RAPL_CORE (RAPL_CORE_ENERGY_STATUS | RAPL_CORE_POWER_LIMIT)
#define RAPL_GFX (RAPL_GFX_POWER_LIMIT | RAPL_GFX_ENERGY_STATUS)
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
#define RAPL_PSYS (RAPL_PLATFORM_ENERGY_STATUS | RAPL_PLATFORM_ENERGY_LIMIT)
tools/power/turbostat: Redefine RAPL macros Redefine RAPL macros to make the code more readable. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-26 14:38:38 +08:00
#define RAPL_PKG_ALL (RAPL_PKG | RAPL_PKG_PERF_STATUS | RAPL_PKG_POWER_INFO)
#define RAPL_DRAM_ALL (RAPL_DRAM | RAPL_DRAM_PERF_STATUS | RAPL_DRAM_POWER_INFO)
#define RAPL_CORE_ALL (RAPL_CORE | RAPL_CORE_POLICY)
#define RAPL_GFX_ALL (RAPL_GFX | RAPL_GFX_POLIGY)
#define RAPL_AMD_F17H (RAPL_AMD_PWR_UNIT | RAPL_AMD_CORE_ENERGY_STAT | RAPL_AMD_PKG_ENERGY_STAT)
tools/power/turbostat: Add skeleton support for cstate enumeration Add skeleton support for cstate enumeration. Note that the previous logic may override the cstate setting for multiple times for different reasons. The conversion to new cstate enumeration must be done step by step following the previous code order strictly. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:37 +08:00
/* For Cstates */
enum cstates {
CC1 = BIT(0),
CC3 = BIT(1),
CC6 = BIT(2),
CC7 = BIT(3),
PC2 = BIT(4),
PC3 = BIT(5),
PC6 = BIT(6),
PC7 = BIT(7),
PC8 = BIT(8),
PC9 = BIT(9),
PC10 = BIT(10),
};
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
static const struct platform_features nhm_features = {
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_133MHZ,
tools/power/turbostat: Adjust cstate for models with .cst_limit set Enable PC3/PC6 for platforms with .cst_limit set because package cstates are guarded by pkg_cstate_limit. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:17:23 +08:00
.supported_cstates = CC1 | CC3 | CC6 | PC3 | PC6,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_NHM,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features nhx_features = {
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_133MHZ,
tools/power/turbostat: Adjust cstate for models with .cst_limit set Enable PC3/PC6 for platforms with .cst_limit set because package cstates are guarded by pkg_cstate_limit. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:17:23 +08:00
.supported_cstates = CC1 | CC3 | CC6 | PC3 | PC6,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_NHM,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features snb_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for has_snb_msrs() models Enable PC7 for has_snb_msrs() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:17:58 +08:00
.supported_cstates = CC1 | CC3 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_SNB,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG | RAPL_CORE_ALL | RAPL_GFX | RAPL_PKG_POWER_INFO,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features snx_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for has_snb_msrs() models Enable PC7 for has_snb_msrs() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:17:58 +08:00
.supported_cstates = CC1 | CC3 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_SNB,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_CORE_ALL | RAPL_DRAM_ALL,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features ivb_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for has_snb_msrs() models Enable PC7 for has_snb_msrs() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:17:58 +08:00
.supported_cstates = CC1 | CC3 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_SNB,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG | RAPL_CORE_ALL | RAPL_GFX | RAPL_PKG_POWER_INFO,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features ivx_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for has_snb_msrs() models Enable PC7 for has_snb_msrs() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:17:58 +08:00
.supported_cstates = CC1 | CC3 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_SNB,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE | TRL_LIMIT1,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_CORE_ALL | RAPL_DRAM_ALL,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features hsw_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for has_snb_msrs() models Enable PC7 for has_snb_msrs() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:17:58 +08:00
.supported_cstates = CC1 | CC3 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_HSW,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract Perf Limit Reasons MSRs support Abstract the support for MSR_CORE/GFX/RING_PERF_LIMIT_REASONS MSRs. Delete perf_limit_reasons_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-07-30 14:25:17 +08:00
.plr_msrs = PLR_CORE | PLR_GFX | PLR_RING,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG | RAPL_CORE_ALL | RAPL_GFX | RAPL_PKG_POWER_INFO,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features hsx_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for has_snb_msrs() models Enable PC7 for has_snb_msrs() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:17:58 +08:00
.supported_cstates = CC1 | CC3 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_HSW,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE | TRL_LIMIT1 | TRL_LIMIT2,
tools/power/turbostat: Abstract Perf Limit Reasons MSRs support Abstract the support for MSR_CORE/GFX/RING_PERF_LIMIT_REASONS MSRs. Delete perf_limit_reasons_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-07-30 14:25:17 +08:00
.plr_msrs = PLR_CORE | PLR_RING,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL,
tools/power/turbostat: Abstract fixed DRAM Energy unit support Abstract the support for fixed Dram domain energy unit. Delete rapl_dram_energy_units_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 12:21:58 +08:00
.has_fixed_rapl_unit = 1,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features hswl_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for has_c8910_msrs() models Enable PC8/PC9/PC10 for has_c8910_msrs() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:20:02 +08:00
.supported_cstates = CC1 | CC3 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7 | PC8 | PC9 | PC10,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_HSW,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract Perf Limit Reasons MSRs support Abstract the support for MSR_CORE/GFX/RING_PERF_LIMIT_REASONS MSRs. Delete perf_limit_reasons_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-07-30 14:25:17 +08:00
.plr_msrs = PLR_CORE | PLR_GFX | PLR_RING,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG | RAPL_CORE_ALL | RAPL_GFX | RAPL_PKG_POWER_INFO,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features hswg_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for has_snb_msrs() models Enable PC7 for has_snb_msrs() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:17:58 +08:00
.supported_cstates = CC1 | CC3 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_HSW,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract Perf Limit Reasons MSRs support Abstract the support for MSR_CORE/GFX/RING_PERF_LIMIT_REASONS MSRs. Delete perf_limit_reasons_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-07-30 14:25:17 +08:00
.plr_msrs = PLR_CORE | PLR_GFX | PLR_RING,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG | RAPL_CORE_ALL | RAPL_GFX | RAPL_PKG_POWER_INFO,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features bdw_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for has_c8910_msrs() models Enable PC8/PC9/PC10 for has_c8910_msrs() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:20:02 +08:00
.supported_cstates = CC1 | CC3 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7 | PC8 | PC9 | PC10,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_HSW,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG | RAPL_CORE_ALL | RAPL_GFX | RAPL_PKG_POWER_INFO,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features bdwg_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for has_snb_msrs() models Enable PC7 for has_snb_msrs() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:17:58 +08:00
.supported_cstates = CC1 | CC3 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_HSW,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG | RAPL_CORE_ALL | RAPL_GFX | RAPL_PKG_POWER_INFO,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features bdx_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for is_bdx() models Disable CC7/PC7 for is_bdx() models. Delete is_bdx() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:16:52 +08:00
.supported_cstates = CC1 | CC3 | CC6 | PC2 | PC3 | PC6,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_HSW,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Automatic Cstate Conversion support Abstract the support for AUTOMATIC_CSTATE_CONVERSION bit in MSR_PKG_CST_CONFIG_CONTROL. Delete automatic_cstate_conversion_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 11:29:18 +08:00
.has_cst_auto_convension = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL,
tools/power/turbostat: Abstract fixed DRAM Energy unit support Abstract the support for fixed Dram domain energy unit. Delete rapl_dram_energy_units_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 12:21:58 +08:00
.has_fixed_rapl_unit = 1,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features skl_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Abstract hardcoded Crystal Clock frequency Abstract the support for hardcoded Crystal Clock frequency, which is used when crystal clock is not available from CPUID.15. Delete CPU model checks in process_cpuid(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 11:39:49 +08:00
.crystal_freq = 24000000,
tools/power/turbostat: Adjust cstate for has_c8910_msrs() models Enable PC8/PC9/PC10 for has_c8910_msrs() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:20:02 +08:00
.supported_cstates = CC1 | CC3 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7 | PC8 | PC9 | PC10,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_HSW,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract extended cstate MSRs support Abstract the support for MSR_PKG_WEIGHTED_CORE_C0_RES, MSR_PKG_ANY_CORE_C0_RES, MSR_PKG_ANY_GFXE_C0_RES and MSR_PKG_BOTH_CORE_GFXE_C0_RES. Delete has_skl_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 15:07:43 +08:00
.has_ext_cst_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract TCC Offset bits support Abstract the support for different TCC Offset bits in MSR_IA32_TEMPERATURE_TARGET. Delete check_tcc_offset() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 23:04:44 +08:00
.tcc_offset_bits = 6,
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_CORE_ALL | RAPL_DRAM | RAPL_DRAM_PERF_STATUS | RAPL_GFX | RAPL_PSYS,
tools/power/turbostat: Abstract TSC tweak support On some models, the CPU base frequency is different from the TSC frequency, and the aperf/mperf counters are running at CPU base frequency instead of TSC frequency. Abstract support for TSC tweak. Given that tsc_tweak depends on base_hz, move the code to probe_bclk() after base_hz is available. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:37:23 +08:00
.enable_tsc_tweak = 1,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features cnl_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for is_slm()/is_knl()/is_cnl()/is_ehl() models Disable CC3 for is_slm()/is_knl()/is_cnl()/is_ehl() models. Delete is_cnl()/is_ehl() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:24:19 +08:00
.supported_cstates = CC1 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7 | PC8 | PC9 | PC10,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_HSW,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Enable MSR_CORE_C1_RES on recent Intel client platforms All recent Intel client platforms have MSR_CORE_C1_RES. Enable the support on these platforms, including CNL/ICL/LKF/RKL/TGL/ADL/RPL/MTL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 15:33:27 +08:00
.has_msr_core_c1_res = 1,
tools/power/turbostat: Abstract extended cstate MSRs support Abstract the support for MSR_PKG_WEIGHTED_CORE_C0_RES, MSR_PKG_ANY_CORE_C0_RES, MSR_PKG_ANY_GFXE_C0_RES and MSR_PKG_BOTH_CORE_GFXE_C0_RES. Delete has_skl_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 15:07:43 +08:00
.has_ext_cst_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract TCC Offset bits support Abstract the support for different TCC Offset bits in MSR_IA32_TEMPERATURE_TARGET. Delete check_tcc_offset() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 23:04:44 +08:00
.tcc_offset_bits = 6,
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_CORE_ALL | RAPL_DRAM | RAPL_DRAM_PERF_STATUS | RAPL_GFX | RAPL_PSYS,
tools/power/turbostat: Abstract TSC tweak support On some models, the CPU base frequency is different from the TSC frequency, and the aperf/mperf counters are running at CPU base frequency instead of TSC frequency. Abstract support for TSC tweak. Given that tsc_tweak depends on base_hz, move the code to probe_bclk() after base_hz is available. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:37:23 +08:00
.enable_tsc_tweak = 1,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
tools/power turbostat: Enhance platform divergence description In various generations, platforms often share a majority of features, diverging only in a few specific aspects. The current approach of using hardcoded values in 'platform_features' structure fails to effectively represent these divergences. To improve the description of platform divergence: 1. Each newly introduced 'platform_features' structure must have a base, typically derived from the previous generation. 2. Platform feature values should be inherited from the base structure rather than being hardcoded. This approach ensures a more accurate and maintainable representation of platform-specific features across different generations. Converts `adl_features` and `lnl_features` to follow this new scheme. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:42 +08:00
/* Copied from cnl_features, with PC7/PC9 removed */
tools/power/turbostat: Remove PC7/PC9 support on ADL/RPL Compared with other platforms that share cnl_features, ADL/RPL don't have PC7/PC9. Clone a new platform feature set from cnl_features for ADL/RPL, with PC7/PC9 removed. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-09 13:26:51 +08:00
static const struct platform_features adl_features = {
tools/power turbostat: Enhance platform divergence description In various generations, platforms often share a majority of features, diverging only in a few specific aspects. The current approach of using hardcoded values in 'platform_features' structure fails to effectively represent these divergences. To improve the description of platform divergence: 1. Each newly introduced 'platform_features' structure must have a base, typically derived from the previous generation. 2. Platform feature values should be inherited from the base structure rather than being hardcoded. This approach ensures a more accurate and maintainable representation of platform-specific features across different generations. Converts `adl_features` and `lnl_features` to follow this new scheme. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:42 +08:00
.has_msr_misc_feature_control = cnl_features.has_msr_misc_feature_control,
.has_msr_misc_pwr_mgmt = cnl_features.has_msr_misc_pwr_mgmt,
.has_nhm_msrs = cnl_features.has_nhm_msrs,
.has_config_tdp = cnl_features.has_config_tdp,
.bclk_freq = cnl_features.bclk_freq,
.supported_cstates = CC1 | CC6 | CC7 | PC2 | PC3 | PC6 | PC8 | PC10,
.cst_limit = cnl_features.cst_limit,
.has_irtl_msrs = cnl_features.has_irtl_msrs,
.has_msr_core_c1_res = cnl_features.has_msr_core_c1_res,
.has_ext_cst_msrs = cnl_features.has_ext_cst_msrs,
.trl_msrs = cnl_features.trl_msrs,
.tcc_offset_bits = cnl_features.tcc_offset_bits,
.rapl_msrs = cnl_features.rapl_msrs,
.enable_tsc_tweak = cnl_features.enable_tsc_tweak,
};
/* Copied from adl_features, with PC3/PC8 removed */
tools/power turbostat: Rename arl_features to lnl_features As ARL shares the same features with ADL/RPL/MTL, now 'arl_features' is used by Lunarlake platform only. Rename 'arl_features' to 'lnl_features'. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:39 +08:00
static const struct platform_features lnl_features = {
tools/power turbostat: Enhance platform divergence description In various generations, platforms often share a majority of features, diverging only in a few specific aspects. The current approach of using hardcoded values in 'platform_features' structure fails to effectively represent these divergences. To improve the description of platform divergence: 1. Each newly introduced 'platform_features' structure must have a base, typically derived from the previous generation. 2. Platform feature values should be inherited from the base structure rather than being hardcoded. This approach ensures a more accurate and maintainable representation of platform-specific features across different generations. Converts `adl_features` and `lnl_features` to follow this new scheme. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:42 +08:00
.has_msr_misc_feature_control = adl_features.has_msr_misc_feature_control,
.has_msr_misc_pwr_mgmt = adl_features.has_msr_misc_pwr_mgmt,
.has_nhm_msrs = adl_features.has_nhm_msrs,
.has_config_tdp = adl_features.has_config_tdp,
.bclk_freq = adl_features.bclk_freq,
.supported_cstates = CC1 | CC6 | CC7 | PC2 | PC6 | PC10,
.cst_limit = adl_features.cst_limit,
.has_irtl_msrs = adl_features.has_irtl_msrs,
.has_msr_core_c1_res = adl_features.has_msr_core_c1_res,
.has_ext_cst_msrs = adl_features.has_ext_cst_msrs,
.trl_msrs = adl_features.trl_msrs,
.tcc_offset_bits = adl_features.tcc_offset_bits,
.rapl_msrs = adl_features.rapl_msrs,
.enable_tsc_tweak = adl_features.enable_tsc_tweak,
tools/power turbostat: Enhance ARL/LNL support ARL/LNL don't have PC8, other than that, it behaves the same as CNL. Copy cnl_features for ARL/LNL, except that PC8 support is removed. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-27 22:35:03 +08:00
};
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
static const struct platform_features skx_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for is_skx()/is_icx()/is_spr() models Disable CC3/CC7/PC3/PC7 for is_skx()/is_icx()/is_spr() models. Delete is_skx() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:27 +08:00
.supported_cstates = CC1 | CC6 | PC2 | PC6,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_SKX,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Automatic Cstate Conversion support Abstract the support for AUTOMATIC_CSTATE_CONVERSION bit in MSR_PKG_CST_CONFIG_CONTROL. Delete automatic_cstate_conversion_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 11:29:18 +08:00
.has_cst_auto_convension = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE | TRL_CORECOUNT,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL,
tools/power/turbostat: Abstract fixed DRAM Energy unit support Abstract the support for fixed Dram domain energy unit. Delete rapl_dram_energy_units_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 12:21:58 +08:00
.has_fixed_rapl_unit = 1,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features icx_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for is_skx()/is_icx()/is_spr() models Disable CC3/CC7/PC3/PC7 for is_skx()/is_icx()/is_spr() models. Delete is_skx() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:27 +08:00
.supported_cstates = CC1 | CC6 | PC2 | PC6,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_ICX,
tools/power/turbostat: Enable MSR_CORE_C1_RES support for ICX Enable Core C1 hardware residency counter (MSR_CORE_C1_RES) on ICX. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-12 11:19:15 +08:00
.has_msr_core_c1_res = 1,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract cstate prewake bit support Abstract cstate prewake bit support. Delete is_icx()/is_spr() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 15:57:44 +08:00
.has_cst_prewake_bit = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE | TRL_CORECOUNT,
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL | RAPL_PSYS,
tools/power/turbostat: Abstract fixed DRAM Energy unit support Abstract the support for fixed Dram domain energy unit. Delete rapl_dram_energy_units_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 12:21:58 +08:00
.has_fixed_rapl_unit = 1,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features spr_features = {
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
.has_msr_misc_feature_control = 1,
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for is_skx()/is_icx()/is_spr() models Disable CC3/CC7/PC3/PC7 for is_skx()/is_icx()/is_spr() models. Delete is_skx() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:27 +08:00
.supported_cstates = CC1 | CC6 | PC2 | PC6,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_SKX,
tools/power/turbostat: Add MSR_CORE_C1_RES support for spr_features Add MSR_CORE_C1_RES support for spr_features because both Sapphirerapids and Emeraldrapids support this MSR. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-19 11:04:33 +08:00
.has_msr_core_c1_res = 1,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract cstate prewake bit support Abstract cstate prewake bit support. Delete is_icx()/is_spr() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 15:57:44 +08:00
.has_cst_prewake_bit = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE | TRL_CORECOUNT,
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL | RAPL_PSYS,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
tools/power/turbostat: Add initial support for SierraForest Add initial support for SierraForest. It shares the same features with SapphireRapids, except that it has MSR_MODULE_C6_RES_MS support. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-09-28 13:09:02 +08:00
static const struct platform_features srf_features = {
.has_msr_misc_feature_control = 1,
.has_msr_misc_pwr_mgmt = 1,
.has_nhm_msrs = 1,
.has_config_tdp = 1,
.bclk_freq = BCLK_100MHZ,
.supported_cstates = CC1 | CC6 | PC2 | PC6,
.cst_limit = CST_LIMIT_SKX,
.has_msr_core_c1_res = 1,
.has_msr_module_c6_res_ms = 1,
.has_irtl_msrs = 1,
.has_cst_prewake_bit = 1,
.trl_msrs = TRL_BASE | TRL_CORECOUNT,
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL | RAPL_PSYS,
tools/power/turbostat: Add initial support for SierraForest Add initial support for SierraForest. It shares the same features with SapphireRapids, except that it has MSR_MODULE_C6_RES_MS support. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-09-28 13:09:02 +08:00
};
tools/power/turbostat: Add initial support for GrandRidge Add initial support for GrandRidge. It shares the same features as SierraForest, except that it does not support PC2/PC6. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-19 11:00:07 +08:00
static const struct platform_features grr_features = {
.has_msr_misc_feature_control = 1,
.has_msr_misc_pwr_mgmt = 1,
.has_nhm_msrs = 1,
.has_config_tdp = 1,
.bclk_freq = BCLK_100MHZ,
.supported_cstates = CC1 | CC6,
.cst_limit = CST_LIMIT_SKX,
.has_msr_core_c1_res = 1,
.has_msr_module_c6_res_ms = 1,
.has_irtl_msrs = 1,
.has_cst_prewake_bit = 1,
.trl_msrs = TRL_BASE | TRL_CORECOUNT,
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL | RAPL_PSYS,
tools/power/turbostat: Add initial support for GrandRidge Add initial support for GrandRidge. It shares the same features as SierraForest, except that it does not support PC2/PC6. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-19 11:00:07 +08:00
};
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
static const struct platform_features slv_features = {
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_SLV,
tools/power/turbostat: Adjust cstate for is_slm()/is_knl()/is_cnl()/is_ehl() models Disable CC3 for is_slm()/is_knl()/is_cnl()/is_ehl() models. Delete is_cnl()/is_ehl() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:24:19 +08:00
.supported_cstates = CC1 | CC6 | PC6,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_SLV,
tools/power/turbostat: Abstract MSR_CORE_C1_RES support Abstract the support for MSR_CORE_C1_RES. Delete is_dnv() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:41:30 +08:00
.has_msr_core_c1_res = 1,
tools/power/turbostat: Abstract MSR_MODULE_C6_RES_MS support Abstract MSR_MODULE_C6_RES_MS support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:47:30 +08:00
.has_msr_module_c6_res_ms = 1,
tools/power/turbostat: Abstract MSR_CC6/MC6_DEMOTION_POLICY_CONFIG support Abstract the support for MSR_CC6/MC6_DEMOTION_POLICY_CONFIG. Delete has_slv_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:52:24 +08:00
.has_msr_c6_demotion_policy_config = 1,
tools/power/turbostat: Abstract MSR_ATOM_PKG_C6_RESIDENCY support Abstract the support for MSR_ATOM_PKG_C6_RESIDENCY. Delete is_slm() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 15:26:14 +08:00
.has_msr_atom_pkg_c6_residency = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_ATOM,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG | RAPL_CORE,
tools/power/turbostat: Abstract RAPL divisor support INTEL_FAM6_ATOM_SILVERMONT model needs a divisor to convert the raw Energy Units value from MSR_RAPL_POWER_UNIT. Abstract the support for RAPL divisor. Delete CPU model check in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 11:59:04 +08:00
.has_rapl_divisor = 1,
tools/power/turbostat: Abstract hardcoded TDP value Different hardcoded TDP values are used when TDP can not be retrieved from the hardware. Abstract hardcoded TDP value. Delete CPU model checks in get_tdp_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:21:10 +08:00
.rapl_quirk_tdp = 30,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features slvd_features = {
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_SLV,
tools/power/turbostat: Adjust cstate for is_slm()/is_knl()/is_cnl()/is_ehl() models Disable CC3 for is_slm()/is_knl()/is_cnl()/is_ehl() models. Delete is_cnl()/is_ehl() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:24:19 +08:00
.supported_cstates = CC1 | CC6 | PC3 | PC6,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_SLV,
tools/power/turbostat: Abstract MSR_ATOM_PKG_C6_RESIDENCY support Abstract the support for MSR_ATOM_PKG_C6_RESIDENCY. Delete is_slm() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 15:26:14 +08:00
.has_msr_atom_pkg_c6_residency = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG | RAPL_CORE,
tools/power/turbostat: Abstract hardcoded TDP value Different hardcoded TDP values are used when TDP can not be retrieved from the hardware. Abstract hardcoded TDP value. Delete CPU model checks in get_tdp_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:21:10 +08:00
.rapl_quirk_tdp = 30,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features amt_features = {
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_133MHZ,
tools/power/turbostat: Adjust cstate for models with .cst_limit set Enable PC3/PC6 for platforms with .cst_limit set because package cstates are guarded by pkg_cstate_limit. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:17:23 +08:00
.supported_cstates = CC1 | CC3 | CC6 | PC3 | PC6,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_AMT,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features gmt_features = {
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Abstract hardcoded Crystal Clock frequency Abstract the support for hardcoded Crystal Clock frequency, which is used when crystal clock is not available from CPUID.15. Delete CPU model checks in process_cpuid(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 11:39:49 +08:00
.crystal_freq = 19200000,
tools/power/turbostat: Adjust cstate for has_c8910_msrs() models Enable PC8/PC9/PC10 for has_c8910_msrs() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:20:02 +08:00
.supported_cstates = CC1 | CC3 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7 | PC8 | PC9 | PC10,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_GMT,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE | TRL_CORECOUNT,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG | RAPL_PKG_POWER_INFO,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features gmtd_features = {
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Abstract hardcoded Crystal Clock frequency Abstract the support for hardcoded Crystal Clock frequency, which is used when crystal clock is not available from CPUID.15. Delete CPU model checks in process_cpuid(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 11:39:49 +08:00
.crystal_freq = 25000000,
tools/power/turbostat: Adjust cstate for is_dnv() models Enable CC1 and disable CC3/CC7/PC3/PC7 for is_dnv() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:12:29 +08:00
.supported_cstates = CC1 | CC6 | PC2 | PC6,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_GMT,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract MSR_CORE_C1_RES support Abstract the support for MSR_CORE_C1_RES. Delete is_dnv() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:41:30 +08:00
.has_msr_core_c1_res = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE | TRL_CORECOUNT,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL | RAPL_CORE_ENERGY_STATUS,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features gmtp_features = {
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Abstract hardcoded Crystal Clock frequency Abstract the support for hardcoded Crystal Clock frequency, which is used when crystal clock is not available from CPUID.15. Delete CPU model checks in process_cpuid(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 11:39:49 +08:00
.crystal_freq = 19200000,
tools/power/turbostat: Adjust cstate for has_c8910_msrs() models Enable PC8/PC9/PC10 for has_c8910_msrs() models. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:20:02 +08:00
.supported_cstates = CC1 | CC3 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7 | PC8 | PC9 | PC10,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_GMT,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG | RAPL_PKG_POWER_INFO,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features tmt_features = {
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for is_slm()/is_knl()/is_cnl()/is_ehl() models Disable CC3 for is_slm()/is_knl()/is_cnl()/is_ehl() models. Delete is_cnl()/is_ehl() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:24:19 +08:00
.supported_cstates = CC1 | CC6 | CC7 | PC2 | PC3 | PC6 | PC7 | PC8 | PC9 | PC10,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_GMT,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_CORE_ALL | RAPL_DRAM | RAPL_DRAM_PERF_STATUS | RAPL_GFX,
tools/power/turbostat: Abstract TSC tweak support On some models, the CPU base frequency is different from the TSC frequency, and the aperf/mperf counters are running at CPU base frequency instead of TSC frequency. Abstract support for TSC tweak. Given that tsc_tweak depends on base_hz, move the code to probe_bclk() after base_hz is available. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:37:23 +08:00
.enable_tsc_tweak = 1,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features tmtd_features = {
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for is_jvl() models Disable CC3/CC7/PC2/PC3/PC6/PC7 for is_jvl() models. Delete is_jvl() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:09:20 +08:00
.supported_cstates = CC1 | CC6,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_GMT,
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
.has_irtl_msrs = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_BASE | TRL_CORECOUNT,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG_ALL,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features knl_features = {
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
.has_msr_misc_pwr_mgmt = 1,
tools/power/turbostat: Abstract Nehalem MSRs support MSR_PLATFORM_INFO, MSR_IA32_TEMPERATURE_TARGET, MSR_SMI_COUNT, MSR_PKG_CST_CONFIG_CONTROL, and the TRL MSRs are always available for platforms since Nehalem. Support for these msrs can be described altogether. Abstract the support for these MSRs. Delete probe_nhm_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 16:52:23 +08:00
.has_nhm_msrs = 1,
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
.has_config_tdp = 1,
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
.bclk_freq = BCLK_100MHZ,
tools/power/turbostat: Adjust cstate for is_slm()/is_knl()/is_cnl()/is_ehl() models Disable CC3 for is_slm()/is_knl()/is_cnl()/is_ehl() models. Delete is_cnl()/is_ehl() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:24:19 +08:00
.supported_cstates = CC1 | CC6 | PC3 | PC6,
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
.cst_limit = CST_LIMIT_KNL,
tools/power/turbostat: Abstract MSR_KNL_CORE_C6_RESIDENCY support Abstract the support for MSR_KNL_CORE_C6_RESIDENCY. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 15:00:58 +08:00
.has_msr_knl_core_c6_residency = 1,
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
.trl_msrs = TRL_KNL,
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
.rapl_msrs = RAPL_PKG_ALL | RAPL_DRAM_ALL,
tools/power/turbostat: Abstract fixed DRAM Energy unit support Abstract the support for fixed Dram domain energy unit. Delete rapl_dram_energy_units_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 12:21:58 +08:00
.has_fixed_rapl_unit = 1,
tools/power/turbostat: Abstract aperf/mperf multiplier support Abstract aperf/mperf multiplier support. Delete is_knl() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 15:49:10 +08:00
.need_perf_multiplier = 1,
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
};
static const struct platform_features default_features = {
};
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
static const struct platform_features amd_features_with_rapl = {
.rapl_msrs = RAPL_AMD_F17H,
tools/power/turbostat: Abstract Per Core RAPL support Abstract the support for Per Core RAPL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:37:46 +08:00
.has_per_core_rapl = 1,
tools/power/turbostat: Abstract hardcoded TDP value Different hardcoded TDP values are used when TDP can not be retrieved from the hardware. Abstract hardcoded TDP value. Delete CPU model checks in get_tdp_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:21:10 +08:00
.rapl_quirk_tdp = 280, /* This is the max stock TDP of HEDT/Server Fam17h+ chips */
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
};
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
static const struct platform_data turbostat_pdata[] = {
tools/power/turbostat: Switch to new Intel CPU model defines New CPU #defines encode vendor and family as well as model. N.B. Copied VFM_*() defines here from <asm/cpu_device_id.h> to avoid an application picking a second internal kernel header file. Signed-off-by: Tony Luck <tony.luck@intel.com> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-06-11 10:19:50 -07:00
{ INTEL_NEHALEM, &nhm_features },
{ INTEL_NEHALEM_G, &nhm_features },
{ INTEL_NEHALEM_EP, &nhm_features },
{ INTEL_NEHALEM_EX, &nhx_features },
{ INTEL_WESTMERE, &nhm_features },
{ INTEL_WESTMERE_EP, &nhm_features },
{ INTEL_WESTMERE_EX, &nhx_features },
{ INTEL_SANDYBRIDGE, &snb_features },
{ INTEL_SANDYBRIDGE_X, &snx_features },
{ INTEL_IVYBRIDGE, &ivb_features },
{ INTEL_IVYBRIDGE_X, &ivx_features },
{ INTEL_HASWELL, &hsw_features },
{ INTEL_HASWELL_X, &hsx_features },
{ INTEL_HASWELL_L, &hswl_features },
{ INTEL_HASWELL_G, &hswg_features },
{ INTEL_BROADWELL, &bdw_features },
{ INTEL_BROADWELL_G, &bdwg_features },
{ INTEL_BROADWELL_X, &bdx_features },
{ INTEL_BROADWELL_D, &bdx_features },
{ INTEL_SKYLAKE_L, &skl_features },
{ INTEL_SKYLAKE, &skl_features },
{ INTEL_SKYLAKE_X, &skx_features },
{ INTEL_KABYLAKE_L, &skl_features },
{ INTEL_KABYLAKE, &skl_features },
{ INTEL_COMETLAKE, &skl_features },
{ INTEL_COMETLAKE_L, &skl_features },
{ INTEL_CANNONLAKE_L, &cnl_features },
{ INTEL_ICELAKE_X, &icx_features },
{ INTEL_ICELAKE_D, &icx_features },
{ INTEL_ICELAKE_L, &cnl_features },
{ INTEL_ICELAKE_NNPI, &cnl_features },
{ INTEL_ROCKETLAKE, &cnl_features },
{ INTEL_TIGERLAKE_L, &cnl_features },
{ INTEL_TIGERLAKE, &cnl_features },
{ INTEL_SAPPHIRERAPIDS_X, &spr_features },
{ INTEL_EMERALDRAPIDS_X, &spr_features },
{ INTEL_GRANITERAPIDS_X, &spr_features },
tools/power turbostat: Add initial support for GraniteRapids-D Add initial support for GraniteRapids-D. It shares the same features with SapphireRapids. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:41 +08:00
{ INTEL_GRANITERAPIDS_D, &spr_features },
tools/power/turbostat: Switch to new Intel CPU model defines New CPU #defines encode vendor and family as well as model. N.B. Copied VFM_*() defines here from <asm/cpu_device_id.h> to avoid an application picking a second internal kernel header file. Signed-off-by: Tony Luck <tony.luck@intel.com> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-06-11 10:19:50 -07:00
{ INTEL_LAKEFIELD, &cnl_features },
{ INTEL_ALDERLAKE, &adl_features },
{ INTEL_ALDERLAKE_L, &adl_features },
{ INTEL_RAPTORLAKE, &adl_features },
{ INTEL_RAPTORLAKE_P, &adl_features },
{ INTEL_RAPTORLAKE_S, &adl_features },
tools/power turbostat: Remove PC7/PC9 support on MTL Similar to ADL/RPL, MTL support CC1/CC6/CC7/PC2/PC3/PC6/PC8/CP10. Remove PC7/PC9 support on MTL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:37 +08:00
{ INTEL_METEORLAKE, &adl_features },
{ INTEL_METEORLAKE_L, &adl_features },
tools/power turbostat: Add back PC8 support on Arrowlake Similar to ADL/RPL/MTL, ARL supports CC1/CC6/CC7/PC2/PC3/PC6/PC8/PC10. Add back PC8 support on Arrowlake. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:38 +08:00
{ INTEL_ARROWLAKE_H, &adl_features },
{ INTEL_ARROWLAKE_U, &adl_features },
{ INTEL_ARROWLAKE, &adl_features },
tools/power turbostat: Rename arl_features to lnl_features As ARL shares the same features with ADL/RPL/MTL, now 'arl_features' is used by Lunarlake platform only. Rename 'arl_features' to 'lnl_features'. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:39 +08:00
{ INTEL_LUNARLAKE_M, &lnl_features },
tools/power/turbostat: Switch to new Intel CPU model defines New CPU #defines encode vendor and family as well as model. N.B. Copied VFM_*() defines here from <asm/cpu_device_id.h> to avoid an application picking a second internal kernel header file. Signed-off-by: Tony Luck <tony.luck@intel.com> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-06-11 10:19:50 -07:00
{ INTEL_ATOM_SILVERMONT, &slv_features },
{ INTEL_ATOM_SILVERMONT_D, &slvd_features },
{ INTEL_ATOM_AIRMONT, &amt_features },
{ INTEL_ATOM_GOLDMONT, &gmt_features },
{ INTEL_ATOM_GOLDMONT_D, &gmtd_features },
{ INTEL_ATOM_GOLDMONT_PLUS, &gmtp_features },
{ INTEL_ATOM_TREMONT_D, &tmtd_features },
{ INTEL_ATOM_TREMONT, &tmt_features },
{ INTEL_ATOM_TREMONT_L, &tmt_features },
{ INTEL_ATOM_GRACEMONT, &adl_features },
{ INTEL_ATOM_CRESTMONT_X, &srf_features },
{ INTEL_ATOM_CRESTMONT, &grr_features },
{ INTEL_XEON_PHI_KNL, &knl_features },
{ INTEL_XEON_PHI_KNM, &knl_features },
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
/*
* Missing support for
tools/power/turbostat: Switch to new Intel CPU model defines New CPU #defines encode vendor and family as well as model. N.B. Copied VFM_*() defines here from <asm/cpu_device_id.h> to avoid an application picking a second internal kernel header file. Signed-off-by: Tony Luck <tony.luck@intel.com> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-06-11 10:19:50 -07:00
* INTEL_ICELAKE
* INTEL_ATOM_SILVERMONT_MID
* INTEL_ATOM_AIRMONT_MID
* INTEL_ATOM_AIRMONT_NP
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
*/
{ 0, NULL },
};
static const struct platform_features *platform;
void probe_platform_features(unsigned int family, unsigned int model)
{
int i;
tools/power/turbostat: Improve probe_platform_features() logic AMD/Hygon platforms that don't have RAPL use 'amd_features' to describe the platform features. Unknown Intel platforms use 'default_features' to describe the platform features. As none of the platform feature is set for 'amd_features' or 'default_features', there is no need to maintain both of them. Remove 'amd_features' structure and improve the logic in probe_platform_features(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:44:06 +08:00
platform = &default_features;
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
tools/power/turbostat: Improve probe_platform_features() logic AMD/Hygon platforms that don't have RAPL use 'amd_features' to describe the platform features. Unknown Intel platforms use 'default_features' to describe the platform features. As none of the platform feature is set for 'amd_features' or 'default_features', there is no need to maintain both of them. Remove 'amd_features' structure and improve the logic in probe_platform_features(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:44:06 +08:00
if (authentic_amd || hygon_genuine) {
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (max_extended_level >= 0x80000007) {
unsigned int eax, ebx, ecx, edx;
__cpuid(0x80000007, eax, ebx, ecx, edx);
/* RAPL (Fam 17h+) */
tools/power/turbostat: Abstract Per Core RAPL support Abstract the support for Per Core RAPL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:37:46 +08:00
if ((edx & (1 << 14)) && family >= 0x17)
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
platform = &amd_features_with_rapl;
}
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
return;
}
tools/power/turbostat: Switch to new Intel CPU model defines New CPU #defines encode vendor and family as well as model. N.B. Copied VFM_*() defines here from <asm/cpu_device_id.h> to avoid an application picking a second internal kernel header file. Signed-off-by: Tony Luck <tony.luck@intel.com> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-06-11 10:19:50 -07:00
if (!genuine_intel)
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
return;
for (i = 0; turbostat_pdata[i].features; i++) {
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
if (VFM_FAMILY(turbostat_pdata[i].vfm) == family && VFM_MODEL(turbostat_pdata[i].vfm) == model) {
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
platform = turbostat_pdata[i].features;
return;
}
}
}
/* Model specific support End */
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
#define TJMAX_DEFAULT 100
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
/* MSRs that are not yet in the kernel-provided header. */
#define MSR_RAPL_PWR_UNIT 0xc0010299
#define MSR_CORE_ENERGY_STAT 0xc001029a
#define MSR_PKG_ENERGY_STAT 0xc001029b
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
#define MAX(a, b) ((a) > (b) ? (a) : (b))
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
int backwards_count;
char *progname;
tools/power turbostat: add --cpu parameter With the --cpu parameter, turbostat prints only lines for the specified set of CPUs: sudo ./turbostat --quiet --show Core,CPU --cpu 0,1,3..5,6-7 Core CPU - - 0 0 0 4 1 1 1 5 2 6 3 3 3 7 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 23:54:15 -05:00
#define CPU_SUBSET_MAXCPUS 1024 /* need to use before probe... */
tools/power/turbostat: Handle cgroup v2 cpu limitation CPUs can be isolated via cgroup settings and turbostat should avoid migrating to these CPUs, just like it does for the '-c' cpus. Introduce cpu_effective_set to save the cgroup cpu limitation info from /sys/fs/cgroup/cpuset.cpus.effective. And use cpu_allowed_set as the intersection of cpu_present_set, cpu_effective_set and cpu_subset. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:39:22 +08:00
cpu_set_t *cpu_present_set, *cpu_effective_set, *cpu_allowed_set, *cpu_affinity_set, *cpu_subset;
size_t cpu_present_setsize, cpu_effective_setsize, cpu_allowed_setsize, cpu_affinity_setsize, cpu_subset_size;
tools/power turbostat: add POLL and POLL% column Like the "C1" and "C1%" column, the new POLL and POLL% columns show invocations and residency% during the measurement interval. While it didn't seem important to track in the past, we've recently found some Linux cpuidle bugs related to POLL%. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:38:29 -04:00
#define MAX_ADDED_THREAD_COUNTERS 24
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
#define MAX_ADDED_CORE_COUNTERS 8
#define MAX_ADDED_PACKAGE_COUNTERS 16
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
#define PMT_MAX_ADDED_THREAD_COUNTERS 24
#define PMT_MAX_ADDED_CORE_COUNTERS 8
#define PMT_MAX_ADDED_PACKAGE_COUNTERS 16
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
#define BITMASK_SIZE 32
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: Add ZERO_ARRAY for zero initializing builtin array It makes it harder to shoot yourself in the foot, by using additional __must_be_array() check. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 13:04:33 +02:00
#define ZERO_ARRAY(arr) (memset(arr, 0, sizeof(arr)) + __must_be_array(arr))
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
/* Indexes used to map data read from perf and MSRs into global variables */
enum rapl_rci_index {
RAPL_RCI_INDEX_ENERGY_PKG = 0,
RAPL_RCI_INDEX_ENERGY_CORES = 1,
RAPL_RCI_INDEX_DRAM = 2,
RAPL_RCI_INDEX_GFX = 3,
RAPL_RCI_INDEX_PKG_PERF_STATUS = 4,
RAPL_RCI_INDEX_DRAM_PERF_STATUS = 5,
RAPL_RCI_INDEX_CORE_ENERGY = 6,
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
RAPL_RCI_INDEX_ENERGY_PLATFORM = 7,
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
NUM_RAPL_COUNTERS,
};
enum rapl_unit {
RAPL_UNIT_INVALID,
RAPL_UNIT_JOULES,
RAPL_UNIT_WATTS,
};
struct rapl_counter_info_t {
unsigned long long data[NUM_RAPL_COUNTERS];
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
enum counter_source source[NUM_RAPL_COUNTERS];
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
unsigned long long flags[NUM_RAPL_COUNTERS];
double scale[NUM_RAPL_COUNTERS];
enum rapl_unit unit[NUM_RAPL_COUNTERS];
tools/power turbostat: Remove anonymous union from rapl_counter_info_t fd_perf field used to be part of the union, but later moved out of it, because we test it with fd_perf != -1 to determine if any perf counter is opened, making the union unused. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:06:39 +02:00
unsigned long long msr[NUM_RAPL_COUNTERS];
unsigned long long msr_mask[NUM_RAPL_COUNTERS];
int msr_shift[NUM_RAPL_COUNTERS];
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
int fd_perf;
};
/* struct rapl_counter_info_t for each RAPL domain */
struct rapl_counter_info_t *rapl_counter_info_perdomain;
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs Save the highest core and package id when parsing topology to allocate enough memory when get_rapl_counters() is called with a core or a package id as a domain. Note that RAPL domains are per-package on Intel, but per-core on AMD. Thus, the RAPL code effectively runs in different modes on those two product lines. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-06 15:39:08 +02:00
unsigned int rapl_counter_info_perdomain_size;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
#define RAPL_COUNTER_FLAG_PLATFORM_COUNTER (1u << 0)
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
#define RAPL_COUNTER_FLAG_USE_MSR_SUM (1u << 1)
struct rapl_counter_arch_info {
int feature_mask; /* Mask for testing if the counter is supported on host */
const char *perf_subsys;
const char *perf_name;
unsigned long long msr;
unsigned long long msr_mask;
int msr_shift; /* Positive mean shift right, negative mean shift left */
double *platform_rapl_msr_scale; /* Scale applied to values read by MSR (platform dependent, filled at runtime) */
unsigned int rci_index; /* Maps data from perf counters to global variables */
unsigned long long bic;
double compat_scale; /* Some counters require constant scaling to be in the same range as other, similar ones */
unsigned long long flags;
};
static const struct rapl_counter_arch_info rapl_counter_arch_infos[] = {
{
.feature_mask = RAPL_PKG,
.perf_subsys = "power",
.perf_name = "energy-pkg",
.msr = MSR_PKG_ENERGY_STATUS,
.msr_mask = 0xFFFFFFFFFFFFFFFF,
.msr_shift = 0,
.platform_rapl_msr_scale = &rapl_energy_units,
.rci_index = RAPL_RCI_INDEX_ENERGY_PKG,
.bic = BIC_PkgWatt | BIC_Pkg_J,
.compat_scale = 1.0,
.flags = RAPL_COUNTER_FLAG_USE_MSR_SUM,
},
{
.feature_mask = RAPL_AMD_F17H,
.perf_subsys = "power",
.perf_name = "energy-pkg",
.msr = MSR_PKG_ENERGY_STAT,
.msr_mask = 0xFFFFFFFFFFFFFFFF,
.msr_shift = 0,
.platform_rapl_msr_scale = &rapl_energy_units,
.rci_index = RAPL_RCI_INDEX_ENERGY_PKG,
.bic = BIC_PkgWatt | BIC_Pkg_J,
.compat_scale = 1.0,
.flags = RAPL_COUNTER_FLAG_USE_MSR_SUM,
},
{
.feature_mask = RAPL_CORE_ENERGY_STATUS,
.perf_subsys = "power",
.perf_name = "energy-cores",
.msr = MSR_PP0_ENERGY_STATUS,
.msr_mask = 0xFFFFFFFFFFFFFFFF,
.msr_shift = 0,
.platform_rapl_msr_scale = &rapl_energy_units,
.rci_index = RAPL_RCI_INDEX_ENERGY_CORES,
.bic = BIC_CorWatt | BIC_Cor_J,
.compat_scale = 1.0,
.flags = RAPL_COUNTER_FLAG_USE_MSR_SUM,
},
{
.feature_mask = RAPL_DRAM,
.perf_subsys = "power",
.perf_name = "energy-ram",
.msr = MSR_DRAM_ENERGY_STATUS,
.msr_mask = 0xFFFFFFFFFFFFFFFF,
.msr_shift = 0,
.platform_rapl_msr_scale = &rapl_dram_energy_units,
.rci_index = RAPL_RCI_INDEX_DRAM,
.bic = BIC_RAMWatt | BIC_RAM_J,
.compat_scale = 1.0,
.flags = RAPL_COUNTER_FLAG_USE_MSR_SUM,
},
{
.feature_mask = RAPL_GFX,
.perf_subsys = "power",
.perf_name = "energy-gpu",
.msr = MSR_PP1_ENERGY_STATUS,
.msr_mask = 0xFFFFFFFFFFFFFFFF,
.msr_shift = 0,
.platform_rapl_msr_scale = &rapl_energy_units,
.rci_index = RAPL_RCI_INDEX_GFX,
.bic = BIC_GFXWatt | BIC_GFX_J,
.compat_scale = 1.0,
.flags = RAPL_COUNTER_FLAG_USE_MSR_SUM,
},
{
.feature_mask = RAPL_PKG_PERF_STATUS,
.perf_subsys = NULL,
.perf_name = NULL,
.msr = MSR_PKG_PERF_STATUS,
.msr_mask = 0xFFFFFFFFFFFFFFFF,
.msr_shift = 0,
.platform_rapl_msr_scale = &rapl_time_units,
.rci_index = RAPL_RCI_INDEX_PKG_PERF_STATUS,
.bic = BIC_PKG__,
.compat_scale = 100.0,
.flags = RAPL_COUNTER_FLAG_USE_MSR_SUM,
},
{
.feature_mask = RAPL_DRAM_PERF_STATUS,
.perf_subsys = NULL,
.perf_name = NULL,
.msr = MSR_DRAM_PERF_STATUS,
.msr_mask = 0xFFFFFFFFFFFFFFFF,
.msr_shift = 0,
.platform_rapl_msr_scale = &rapl_time_units,
.rci_index = RAPL_RCI_INDEX_DRAM_PERF_STATUS,
.bic = BIC_RAM__,
.compat_scale = 100.0,
.flags = RAPL_COUNTER_FLAG_USE_MSR_SUM,
},
{
.feature_mask = RAPL_AMD_F17H,
.perf_subsys = NULL,
.perf_name = NULL,
.msr = MSR_CORE_ENERGY_STAT,
.msr_mask = 0xFFFFFFFF,
.msr_shift = 0,
.platform_rapl_msr_scale = &rapl_energy_units,
.rci_index = RAPL_RCI_INDEX_CORE_ENERGY,
.bic = BIC_CorWatt | BIC_Cor_J,
.compat_scale = 1.0,
.flags = 0,
},
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
{
.feature_mask = RAPL_PSYS,
.perf_subsys = "power",
.perf_name = "energy-psys",
.msr = MSR_PLATFORM_ENERGY_STATUS,
.msr_mask = 0x00000000FFFFFFFF,
.msr_shift = 0,
.platform_rapl_msr_scale = &rapl_energy_units,
.rci_index = RAPL_RCI_INDEX_ENERGY_PLATFORM,
.bic = BIC_SysWatt | BIC_Sys_J,
.compat_scale = 1.0,
.flags = RAPL_COUNTER_FLAG_PLATFORM_COUNTER | RAPL_COUNTER_FLAG_USE_MSR_SUM,
},
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
};
struct rapl_counter {
unsigned long long raw_value;
enum rapl_unit unit;
double scale;
};
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
/* Indexes used to map data read from perf and MSRs into global variables */
enum ccstate_rci_index {
CCSTATE_RCI_INDEX_C1_RESIDENCY = 0,
CCSTATE_RCI_INDEX_C3_RESIDENCY = 1,
CCSTATE_RCI_INDEX_C6_RESIDENCY = 2,
CCSTATE_RCI_INDEX_C7_RESIDENCY = 3,
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
PCSTATE_RCI_INDEX_C2_RESIDENCY = 4,
PCSTATE_RCI_INDEX_C3_RESIDENCY = 5,
PCSTATE_RCI_INDEX_C6_RESIDENCY = 6,
PCSTATE_RCI_INDEX_C7_RESIDENCY = 7,
PCSTATE_RCI_INDEX_C8_RESIDENCY = 8,
PCSTATE_RCI_INDEX_C9_RESIDENCY = 9,
PCSTATE_RCI_INDEX_C10_RESIDENCY = 10,
NUM_CSTATE_COUNTERS,
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
};
struct cstate_counter_info_t {
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
unsigned long long data[NUM_CSTATE_COUNTERS];
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
enum counter_source source[NUM_CSTATE_COUNTERS];
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
unsigned long long msr[NUM_CSTATE_COUNTERS];
int fd_perf_core;
int fd_perf_pkg;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
};
struct cstate_counter_info_t *ccstate_counter_info;
unsigned int ccstate_counter_info_size;
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
#define CSTATE_COUNTER_FLAG_COLLECT_PER_CORE (1u << 0)
#define CSTATE_COUNTER_FLAG_COLLECT_PER_THREAD ((1u << 1) | CSTATE_COUNTER_FLAG_COLLECT_PER_CORE)
#define CSTATE_COUNTER_FLAG_SOFT_C1_DEPENDENCY (1u << 2)
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
struct cstate_counter_arch_info {
int feature_mask; /* Mask for testing if the counter is supported on host */
const char *perf_subsys;
const char *perf_name;
unsigned long long msr;
unsigned int rci_index; /* Maps data from perf counters to global variables */
unsigned long long bic;
unsigned long long flags;
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
int pkg_cstate_limit;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
};
static struct cstate_counter_arch_info ccstate_counter_arch_infos[] = {
{
.feature_mask = CC1,
.perf_subsys = "cstate_core",
.perf_name = "c1-residency",
.msr = MSR_CORE_C1_RES,
.rci_index = CCSTATE_RCI_INDEX_C1_RESIDENCY,
.bic = BIC_CPU_c1,
.flags = CSTATE_COUNTER_FLAG_COLLECT_PER_THREAD,
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
.pkg_cstate_limit = 0,
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
},
{
.feature_mask = CC3,
.perf_subsys = "cstate_core",
.perf_name = "c3-residency",
.msr = MSR_CORE_C3_RESIDENCY,
.rci_index = CCSTATE_RCI_INDEX_C3_RESIDENCY,
.bic = BIC_CPU_c3,
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
.flags = CSTATE_COUNTER_FLAG_COLLECT_PER_CORE | CSTATE_COUNTER_FLAG_SOFT_C1_DEPENDENCY,
.pkg_cstate_limit = 0,
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
},
{
.feature_mask = CC6,
.perf_subsys = "cstate_core",
.perf_name = "c6-residency",
.msr = MSR_CORE_C6_RESIDENCY,
.rci_index = CCSTATE_RCI_INDEX_C6_RESIDENCY,
.bic = BIC_CPU_c6,
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
.flags = CSTATE_COUNTER_FLAG_COLLECT_PER_CORE | CSTATE_COUNTER_FLAG_SOFT_C1_DEPENDENCY,
.pkg_cstate_limit = 0,
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
},
{
.feature_mask = CC7,
.perf_subsys = "cstate_core",
.perf_name = "c7-residency",
.msr = MSR_CORE_C7_RESIDENCY,
.rci_index = CCSTATE_RCI_INDEX_C7_RESIDENCY,
.bic = BIC_CPU_c7,
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
.flags = CSTATE_COUNTER_FLAG_COLLECT_PER_CORE | CSTATE_COUNTER_FLAG_SOFT_C1_DEPENDENCY,
.pkg_cstate_limit = 0,
},
{
.feature_mask = PC2,
.perf_subsys = "cstate_pkg",
.perf_name = "c2-residency",
.msr = MSR_PKG_C2_RESIDENCY,
.rci_index = PCSTATE_RCI_INDEX_C2_RESIDENCY,
.bic = BIC_Pkgpc2,
.flags = 0,
.pkg_cstate_limit = PCL__2,
},
{
.feature_mask = PC3,
.perf_subsys = "cstate_pkg",
.perf_name = "c3-residency",
.msr = MSR_PKG_C3_RESIDENCY,
.rci_index = PCSTATE_RCI_INDEX_C3_RESIDENCY,
.bic = BIC_Pkgpc3,
.flags = 0,
.pkg_cstate_limit = PCL__3,
},
{
.feature_mask = PC6,
.perf_subsys = "cstate_pkg",
.perf_name = "c6-residency",
.msr = MSR_PKG_C6_RESIDENCY,
.rci_index = PCSTATE_RCI_INDEX_C6_RESIDENCY,
.bic = BIC_Pkgpc6,
.flags = 0,
.pkg_cstate_limit = PCL__6,
},
{
.feature_mask = PC7,
.perf_subsys = "cstate_pkg",
.perf_name = "c7-residency",
.msr = MSR_PKG_C7_RESIDENCY,
.rci_index = PCSTATE_RCI_INDEX_C7_RESIDENCY,
.bic = BIC_Pkgpc7,
.flags = 0,
.pkg_cstate_limit = PCL__7,
},
{
.feature_mask = PC8,
.perf_subsys = "cstate_pkg",
.perf_name = "c8-residency",
.msr = MSR_PKG_C8_RESIDENCY,
.rci_index = PCSTATE_RCI_INDEX_C8_RESIDENCY,
.bic = BIC_Pkgpc8,
.flags = 0,
.pkg_cstate_limit = PCL__8,
},
{
.feature_mask = PC9,
.perf_subsys = "cstate_pkg",
.perf_name = "c9-residency",
.msr = MSR_PKG_C9_RESIDENCY,
.rci_index = PCSTATE_RCI_INDEX_C9_RESIDENCY,
.bic = BIC_Pkgpc9,
.flags = 0,
.pkg_cstate_limit = PCL__9,
},
{
.feature_mask = PC10,
.perf_subsys = "cstate_pkg",
.perf_name = "c10-residency",
.msr = MSR_PKG_C10_RESIDENCY,
.rci_index = PCSTATE_RCI_INDEX_C10_RESIDENCY,
.bic = BIC_Pkgpc10,
.flags = 0,
.pkg_cstate_limit = PCL_10,
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
},
};
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
/* Indexes used to map data read from perf and MSRs into global variables */
enum msr_rci_index {
MSR_RCI_INDEX_APERF = 0,
MSR_RCI_INDEX_MPERF = 1,
MSR_RCI_INDEX_SMI = 2,
NUM_MSR_COUNTERS,
};
struct msr_counter_info_t {
unsigned long long data[NUM_MSR_COUNTERS];
enum counter_source source[NUM_MSR_COUNTERS];
unsigned long long msr[NUM_MSR_COUNTERS];
unsigned long long msr_mask[NUM_MSR_COUNTERS];
int fd_perf;
};
struct msr_counter_info_t *msr_counter_info;
unsigned int msr_counter_info_size;
struct msr_counter_arch_info {
const char *perf_subsys;
const char *perf_name;
unsigned long long msr;
unsigned long long msr_mask;
unsigned int rci_index; /* Maps data from perf counters to global variables */
bool needed;
bool present;
};
enum msr_arch_info_index {
MSR_ARCH_INFO_APERF_INDEX = 0,
MSR_ARCH_INFO_MPERF_INDEX = 1,
MSR_ARCH_INFO_SMI_INDEX = 2,
};
static struct msr_counter_arch_info msr_counter_arch_infos[] = {
[MSR_ARCH_INFO_APERF_INDEX] = {
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
.perf_subsys = "msr",
.perf_name = "aperf",
.msr = MSR_IA32_APERF,
.msr_mask = 0xFFFFFFFFFFFFFFFF,
.rci_index = MSR_RCI_INDEX_APERF,
},
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
[MSR_ARCH_INFO_MPERF_INDEX] = {
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
.perf_subsys = "msr",
.perf_name = "mperf",
.msr = MSR_IA32_MPERF,
.msr_mask = 0xFFFFFFFFFFFFFFFF,
.rci_index = MSR_RCI_INDEX_MPERF,
},
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
[MSR_ARCH_INFO_SMI_INDEX] = {
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
.perf_subsys = "msr",
.perf_name = "smi",
.msr = MSR_SMI_COUNT,
.msr_mask = 0xFFFFFFFF,
.rci_index = MSR_RCI_INDEX_SMI,
},
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
};
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
/* Can be redefined when compiling, useful for testing. */
#ifndef SYSFS_TELEM_PATH
#define SYSFS_TELEM_PATH "/sys/class/intel_pmt"
#endif
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
#define PMT_COUNTER_MTL_DC6_OFFSET 120
#define PMT_COUNTER_MTL_DC6_LSB 0
#define PMT_COUNTER_MTL_DC6_MSB 63
#define PMT_MTL_DC6_GUID 0x1a067102
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
#define PMT_COUNTER_NAME_SIZE_BYTES 16
#define PMT_COUNTER_TYPE_NAME_SIZE_BYTES 32
struct pmt_mmio {
struct pmt_mmio *next;
unsigned int guid;
unsigned int size;
/* Base pointer to the mmaped memory. */
void *mmio_base;
/*
* Offset to be applied to the mmio_base
* to get the beginning of the PMT counters for given GUID.
*/
unsigned long pmt_offset;
} *pmt_mmios;
enum pmt_datatype {
PMT_TYPE_RAW,
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
PMT_TYPE_XTAL_TIME,
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
};
struct pmt_domain_info {
/*
* Pointer to the MMIO obtained by applying a counter offset
* to the mmio_base of the mmaped region for the given GUID.
*
* This is where to read the raw value of the counter from.
*/
unsigned long *pcounter;
};
struct pmt_counter {
struct pmt_counter *next;
/* PMT metadata */
char name[PMT_COUNTER_NAME_SIZE_BYTES];
enum pmt_datatype type;
enum counter_scope scope;
unsigned int lsb;
unsigned int msb;
/* BIC-like metadata */
enum counter_format format;
unsigned int num_domains;
struct pmt_domain_info *domains;
};
unsigned int pmt_counter_get_width(const struct pmt_counter *p)
{
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
return (p->msb - p->lsb) + 1;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
}
void pmt_counter_resize_(struct pmt_counter *pcounter, unsigned int new_size)
{
struct pmt_domain_info *new_mem;
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
new_mem = (struct pmt_domain_info *)reallocarray(pcounter->domains, new_size, sizeof(*pcounter->domains));
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
if (!new_mem) {
fprintf(stderr, "%s: failed to allocate memory for PMT counters\n", __func__);
exit(1);
}
/* Zero initialize just allocated memory. */
const size_t num_new_domains = new_size - pcounter->num_domains;
memset(&new_mem[pcounter->num_domains], 0, num_new_domains * sizeof(*pcounter->domains));
pcounter->num_domains = new_size;
pcounter->domains = new_mem;
}
void pmt_counter_resize(struct pmt_counter *pcounter, unsigned int new_size)
{
/*
* Allocate more memory ahead of time.
*
* Always allocate space for at least 8 elements
* and double the size when growing.
*/
if (new_size < 8)
new_size = 8;
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
new_size = MAX(new_size, pcounter->num_domains * 2);
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
pmt_counter_resize_(pcounter, new_size);
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
struct thread_data {
tools/power turbostat: if --debug, print sampling overhead The --debug option now pre-pends each row with the number of micro-seconds [usec] to collect the finishing snapshot for that row. Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-27 21:06:55 -07:00
struct timeval tv_begin;
struct timeval tv_end;
tools/power turbostat: Make interval calculation per thread to reduce jitter Turbostat currently normalizes TSC and other values by dividing by an interval. This interval is the delta between the start of one global (all counters on all CPUs) sampling and the start of another. However, this introduces a lot of jitter into the data. In order to reduce jitter, the interval calculation should be based on timestamps taken per thread and close to the start of the thread's sampling. Define a per thread time value to hold the delta between samples taken on the thread. Use the timestamp taken at the beginning of sampling to calculate the delta. Move the thread's beginning timestamp to after the CPU migration to avoid jitter due to the migration. Use the global time delta for the average time delta. Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-25 17:32:42 +00:00
struct timeval tv_delta;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
unsigned long long tsc;
unsigned long long aperf;
unsigned long long mperf;
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
unsigned long long c1;
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
unsigned long long instr_count;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
unsigned long long irq_count;
tools/power turbostat: display SMI count by default The SMI counter is popular -- so display it by default rather than requiring an option. What the heck, we've blown the 80 column budget on many systems already... Note that the value displayed is the delta during the measurement interval. The absolute value of the counter can still be seen with the generic 32-bit MSR option, ie. -m 0x34 Signed-off-by: Len Brown <len.brown@intel.com>
2013-02-10 17:19:24 -05:00
unsigned int smi_count;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
unsigned int cpu_id;
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
unsigned int apic_id;
unsigned int x2apic_id;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
unsigned int flags;
tools/power turbostat: Fix Core C6 residency on Atom CPUs For Atom CPUs that have core cstate deeper than C6, MSR_CORE_C6_RESIDENCY actually returns the residency for both CC6 and deeper Core cstates. Thus, the real Core C6 residency should be the subtraction of MSR_CORE_C6_RESIDENCY return value and MSR_CORE_C6_RESIDENCY return value. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-21 22:22:47 +08:00
bool is_atom;
tools/power turbostat: add POLL and POLL% column Like the "C1" and "C1%" column, the new POLL and POLL% columns show invocations and residency% during the measurement interval. While it didn't seem important to track in the past, we've recently found some Linux cpuidle bugs related to POLL%. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:38:29 -04:00
unsigned long long counter[MAX_ADDED_THREAD_COUNTERS];
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
unsigned long long perf_counter[MAX_ADDED_THREAD_COUNTERS];
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
unsigned long long pmt_counter[PMT_MAX_ADDED_THREAD_COUNTERS];
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
} *thread_even, *thread_odd;
struct core_data {
tools/power/turbostat: Obey allowed CPUs for primary thread/core detection Thread_id doesn't tell if a CPU is allowed or not. Detect allowed CPUs only and use the first detected thread/core as the primary thread/core of a core/package. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 09:59:14 +08:00
int base_cpu;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
unsigned long long c3;
unsigned long long c6;
unsigned long long c7;
tools/power turbostat: Baytrail c-state support The Baytrail SOC, with its Silvermont core, has some unique properties: 1. a hardware CC1 residency counter 2. a module-c6 residency counter 3. a package-c6 counter at traditional package-c7 counter address. The SOC does not support c3, pc3, c7 or pc7 counters. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:27:20 -05:00
unsigned long long mc6_us; /* duplicate as per-core for now, even though per module */
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
unsigned int core_temp_c;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
struct rapl_counter core_energy; /* MSR_CORE_ENERGY_STAT */
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
unsigned int core_id;
tools/power turbostat: Support thermal throttle count print The turbostat data is collected by end user for power evaluationit. However it looks like we are missing enough thermal context there. Already a couple of time we found that power management developer asking something like this: grep -r . /sys/devices/system/cpu/cpu*/thermal_throttle/* Print the per core thermal throttle count so as to get suffificent thermal context. turbostat -i 5 -s Core,CPU,CoreThr Core CPU CoreThr - - 104 0 0 61 0 4 1 1 0 1 5 2 2 104 2 6 3 3 7 3 7 Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-11-12 19:51:59 +08:00
unsigned long long core_throt_cnt;
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
unsigned long long counter[MAX_ADDED_CORE_COUNTERS];
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
unsigned long long perf_counter[MAX_ADDED_CORE_COUNTERS];
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
unsigned long long pmt_counter[PMT_MAX_ADDED_CORE_COUNTERS];
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
} *core_even, *core_odd;
struct pkg_data {
tools/power/turbostat: Obey allowed CPUs for primary thread/core detection Thread_id doesn't tell if a CPU is allowed or not. Detect allowed CPUs only and use the first detected thread/core as the primary thread/core of a core/package. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 09:59:14 +08:00
int base_cpu;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
unsigned long long pc2;
unsigned long long pc3;
unsigned long long pc6;
unsigned long long pc7;
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 05:22:43 -08:00
unsigned long long pc8;
unsigned long long pc9;
unsigned long long pc10;
tools/power turbostat: Do not print negative LPI residency turbostat prints the abnormal SYS%LPI across suspend-to-idle: SYS%LPI = 114479815993277.50 This is reproduced by: Run a freeze cycle, e.g. "sleepgraph -m freeze -rtcwake 15". Then do a reboot. After boot up, launch the suspend-idle-idle and check the SYS%LPI field. The slp_so residence counter is in LPIT table, and BIOS does not clears this register across reset. The PMC expects the OS to calculate the LPI residency based on the delta. However, there is an firmware issue that the LPIT gets cleared to 0 during the second suspend to idle after the reboot, which brings negative delta value. [lenb: updated to print "neg" upon this BIOS failure] Reported-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2023-10-22 13:52:21 +08:00
long long cpu_lpi;
long long sys_lpi;
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
unsigned long long pkg_wtd_core_c0;
unsigned long long pkg_any_core_c0;
unsigned long long pkg_any_gfxe_c0;
unsigned long long pkg_both_core_gfxe_c0;
tools/power turbostat: work around RC6 counter wrap Sometimes the rc6 sysfs counter spontaneously resets, causing turbostat prints a very large number as it tries to calcuate % = 100 * (old - new) / interval When we see (old > new), print ***.**% instead of a bogus huge number. Note that this detection is not fool-proof, as the counter could reset several times and still result in new > old. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-06 17:16:00 -04:00
long long gfx_rc6_ms;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 00:37:54 -05:00
unsigned int gfx_mhz;
tools/power turbostat: Add a new GFXAMHz column that exposes gt_act_freq_mhz. The column already present called GFXMHz reads from gt_cur_freq_mhz, which represents the GT frequency that was requested, but power management might not be able to do that. So the new column will display what the actual frequency GT is running at. Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-22 15:02:07 -07:00
unsigned int gfx_act_mhz;
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
long long sam_mc6_ms;
unsigned int sam_mhz;
unsigned int sam_act_mhz;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
unsigned int package_id;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
struct rapl_counter energy_pkg; /* MSR_PKG_ENERGY_STATUS */
struct rapl_counter energy_dram; /* MSR_DRAM_ENERGY_STATUS */
struct rapl_counter energy_cores; /* MSR_PP0_ENERGY_STATUS */
struct rapl_counter energy_gfx; /* MSR_PP1_ENERGY_STATUS */
struct rapl_counter rapl_pkg_perf_status; /* MSR_PKG_PERF_STATUS */
struct rapl_counter rapl_dram_perf_status; /* MSR_DRAM_PERF_STATUS */
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
unsigned int pkg_temp_c;
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
unsigned int uncore_mhz;
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
unsigned long long die_c6;
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
unsigned long long counter[MAX_ADDED_PACKAGE_COUNTERS];
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
unsigned long long perf_counter[MAX_ADDED_PACKAGE_COUNTERS];
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
unsigned long long pmt_counter[PMT_MAX_ADDED_PACKAGE_COUNTERS];
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
} *package_even, *package_odd;
#define ODD_COUNTERS thread_odd, core_odd, package_odd
#define EVEN_COUNTERS thread_even, core_even, package_even
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
#define GET_THREAD(thread_base, thread_no, core_no, node_no, pkg_no) \
((thread_base) + \
((pkg_no) * \
topo.nodes_per_pkg * topo.cores_per_node * topo.threads_per_core) + \
((node_no) * topo.cores_per_node * topo.threads_per_core) + \
((core_no) * topo.threads_per_core) + \
(thread_no))
#define GET_CORE(core_base, core_no, node_no, pkg_no) \
((core_base) + \
((pkg_no) * topo.nodes_per_pkg * topo.cores_per_node) + \
((node_no) * topo.cores_per_node) + \
(core_no))
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
#define GET_PKG(pkg_base, pkg_no) (pkg_base + pkg_no)
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
/*
* The accumulated sum of MSR is defined as a monotonic
* increasing MSR, it will be accumulated periodically,
* despite its register's bit width.
*/
enum {
IDX_PKG_ENERGY,
IDX_DRAM_ENERGY,
IDX_PP0_ENERGY,
IDX_PP1_ENERGY,
IDX_PKG_PERF,
IDX_DRAM_PERF,
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
IDX_PSYS_ENERGY,
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
IDX_COUNT,
};
int get_msr_sum(int cpu, off_t offset, unsigned long long *msr);
struct msr_sum_array {
/* get_msr_sum() = sum + (get_msr() - last) */
struct {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
/*The accumulated MSR value is updated by the timer */
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
unsigned long long sum;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
/*The MSR footprint recorded in last timer */
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
unsigned long long last;
} entries[IDX_COUNT];
};
/* The percpu MSR sum array.*/
struct msr_sum_array *per_cpu_msr_sum;
tools/power turbostat: Fix offset overflow issue in index converting The idx_to_offset() function returns type int (32-bit signed), but MSR_PKG_ENERGY_STAT is u32 and would be interpreted as a negative number. The end result is that it hits the if (offset < 0) check in update_msr_sum() which prevents the timer callback from updating the stat in the background when long durations are used. The similar issue exists in offset_to_idx() and update_msr_sum(). Fix this issue by converting the 'int' to 'off_t' accordingly. Fixes: 9972d5d84d76 ("tools/power turbostat: Enable accumulate RAPL display") Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-28 17:09:16 +08:00
off_t idx_to_offset(int idx)
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
{
tools/power turbostat: Fix offset overflow issue in index converting The idx_to_offset() function returns type int (32-bit signed), but MSR_PKG_ENERGY_STAT is u32 and would be interpreted as a negative number. The end result is that it hits the if (offset < 0) check in update_msr_sum() which prevents the timer callback from updating the stat in the background when long durations are used. The similar issue exists in offset_to_idx() and update_msr_sum(). Fix this issue by converting the 'int' to 'off_t' accordingly. Fixes: 9972d5d84d76 ("tools/power turbostat: Enable accumulate RAPL display") Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-28 17:09:16 +08:00
off_t offset;
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
switch (idx) {
case IDX_PKG_ENERGY:
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (platform->rapl_msrs & RAPL_AMD_F17H)
tools/power/turbostat: Fix turbostat for AMD Zen CPUs It was reported that on Zen+ system turbostat started exiting, which was tracked down to the MSR_PKG_ENERGY_STAT read failing because offset_to_idx wasn't returning a non-negative index. This patch combined the modification from Bingsong Si and Bas Nieuwenhuizen and addd the MSR to the index system as alternative for MSR_PKG_ENERGY_STATUS. Fixes: 9972d5d84d76 ("tools/power turbostat: Enable accumulate RAPL display") Reported-by: youling257 <youling257@gmail.com> Tested-by: youling257 <youling257@gmail.com> Tested-by: Kurt Garloff <kurt@garloff.de> Tested-by: Bingsong Si <owen.si@ucloud.cn> Tested-by: Artem S. Tashkinov <aros@gmx.com> Co-developed-by: Bingsong Si <owen.si@ucloud.cn> Co-developed-by: Terry Bowman <terry.bowman@amd.com> Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl> Reviewed-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-28 17:09:03 +08:00
offset = MSR_PKG_ENERGY_STAT;
else
offset = MSR_PKG_ENERGY_STATUS;
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
break;
case IDX_DRAM_ENERGY:
offset = MSR_DRAM_ENERGY_STATUS;
break;
case IDX_PP0_ENERGY:
offset = MSR_PP0_ENERGY_STATUS;
break;
case IDX_PP1_ENERGY:
offset = MSR_PP1_ENERGY_STATUS;
break;
case IDX_PKG_PERF:
offset = MSR_PKG_PERF_STATUS;
break;
case IDX_DRAM_PERF:
offset = MSR_DRAM_PERF_STATUS;
break;
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
case IDX_PSYS_ENERGY:
offset = MSR_PLATFORM_ENERGY_STATUS;
break;
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
default:
offset = -1;
}
return offset;
}
tools/power turbostat: Fix offset overflow issue in index converting The idx_to_offset() function returns type int (32-bit signed), but MSR_PKG_ENERGY_STAT is u32 and would be interpreted as a negative number. The end result is that it hits the if (offset < 0) check in update_msr_sum() which prevents the timer callback from updating the stat in the background when long durations are used. The similar issue exists in offset_to_idx() and update_msr_sum(). Fix this issue by converting the 'int' to 'off_t' accordingly. Fixes: 9972d5d84d76 ("tools/power turbostat: Enable accumulate RAPL display") Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-28 17:09:16 +08:00
int offset_to_idx(off_t offset)
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
{
int idx;
switch (offset) {
case MSR_PKG_ENERGY_STATUS:
tools/power/turbostat: Fix turbostat for AMD Zen CPUs It was reported that on Zen+ system turbostat started exiting, which was tracked down to the MSR_PKG_ENERGY_STAT read failing because offset_to_idx wasn't returning a non-negative index. This patch combined the modification from Bingsong Si and Bas Nieuwenhuizen and addd the MSR to the index system as alternative for MSR_PKG_ENERGY_STATUS. Fixes: 9972d5d84d76 ("tools/power turbostat: Enable accumulate RAPL display") Reported-by: youling257 <youling257@gmail.com> Tested-by: youling257 <youling257@gmail.com> Tested-by: Kurt Garloff <kurt@garloff.de> Tested-by: Bingsong Si <owen.si@ucloud.cn> Tested-by: Artem S. Tashkinov <aros@gmx.com> Co-developed-by: Bingsong Si <owen.si@ucloud.cn> Co-developed-by: Terry Bowman <terry.bowman@amd.com> Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl> Reviewed-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-28 17:09:03 +08:00
case MSR_PKG_ENERGY_STAT:
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
idx = IDX_PKG_ENERGY;
break;
case MSR_DRAM_ENERGY_STATUS:
idx = IDX_DRAM_ENERGY;
break;
case MSR_PP0_ENERGY_STATUS:
idx = IDX_PP0_ENERGY;
break;
case MSR_PP1_ENERGY_STATUS:
idx = IDX_PP1_ENERGY;
break;
case MSR_PKG_PERF_STATUS:
idx = IDX_PKG_PERF;
break;
case MSR_DRAM_PERF_STATUS:
idx = IDX_DRAM_PERF;
break;
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
case MSR_PLATFORM_ENERGY_STATUS:
idx = IDX_PSYS_ENERGY;
break;
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
default:
idx = -1;
}
return idx;
}
int idx_valid(int idx)
{
switch (idx) {
case IDX_PKG_ENERGY:
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
return platform->rapl_msrs & (RAPL_PKG | RAPL_AMD_F17H);
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
case IDX_DRAM_ENERGY:
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
return platform->rapl_msrs & RAPL_DRAM;
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
case IDX_PP0_ENERGY:
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
return platform->rapl_msrs & RAPL_CORE_ENERGY_STATUS;
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
case IDX_PP1_ENERGY:
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
return platform->rapl_msrs & RAPL_GFX;
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
case IDX_PKG_PERF:
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
return platform->rapl_msrs & RAPL_PKG_PERF_STATUS;
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
case IDX_DRAM_PERF:
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
return platform->rapl_msrs & RAPL_DRAM_PERF_STATUS;
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
case IDX_PSYS_ENERGY:
return platform->rapl_msrs & RAPL_PSYS;
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
default:
return 0;
}
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
struct sys_counters {
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
/* MSR added counters */
tools/power turbostat: fix bugs in --add option When --add was used more than once, overflowed buffers caused some counters to be stored on top of others, corrupting the results. Simplify the code by simply reserving space for up to 16 added counters per each cpu, core, package. Per-cpu added counters were being printed only per-core. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 22:22:13 -05:00
unsigned int added_thread_counters;
unsigned int added_core_counters;
unsigned int added_package_counters;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
struct msr_counter *tp;
struct msr_counter *cp;
struct msr_counter *pp;
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
/* perf added counters */
unsigned int added_thread_perf_counters;
unsigned int added_core_perf_counters;
unsigned int added_package_perf_counters;
struct perf_counter_info *perf_tp;
struct perf_counter_info *perf_cp;
struct perf_counter_info *perf_pp;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
struct pmt_counter *pmt_tp;
struct pmt_counter *pmt_cp;
struct pmt_counter *pmt_pp;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
} sys;
tools/power turbostat: Enable non-privileged users to read sysfs counters A group of counters called "sysfs" displays software C-state request counts and resulting perceived C-state residency. They are not built-in counters that turbostat knows about ahead of time, rather they are discovered in sysfs when turbostat starts. Thus, they are added dynamically, using the same interface as user-added MSR counters. When turbostat enters "no-msr" mode, such as when running as a non-privileged user, it clears all added counters. Updating that to clear only actual MSR added counters allows regular users to see the sysfs counters. [lenb: commit message] Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-25 17:54:18 +02:00
static size_t free_msr_counters_(struct msr_counter **pp)
tools/power turbostat: Clear added counters when in no-msr mode If user request --no-msr or is not able to access the MSRs, turbostat should clear all the counters added with --add. Because MSR access permission checks are done after the cmdline is parsed, the decision has to be defered up until the transition into no-msr mode happen. Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-14 11:36:55 +01:00
{
tools/power turbostat: Enable non-privileged users to read sysfs counters A group of counters called "sysfs" displays software C-state request counts and resulting perceived C-state residency. They are not built-in counters that turbostat knows about ahead of time, rather they are discovered in sysfs when turbostat starts. Thus, they are added dynamically, using the same interface as user-added MSR counters. When turbostat enters "no-msr" mode, such as when running as a non-privileged user, it clears all added counters. Updating that to clear only actual MSR added counters allows regular users to see the sysfs counters. [lenb: commit message] Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-25 17:54:18 +02:00
struct msr_counter *p = NULL;
size_t num_freed = 0;
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
tools/power turbostat: Enable non-privileged users to read sysfs counters A group of counters called "sysfs" displays software C-state request counts and resulting perceived C-state residency. They are not built-in counters that turbostat knows about ahead of time, rather they are discovered in sysfs when turbostat starts. Thus, they are added dynamically, using the same interface as user-added MSR counters. When turbostat enters "no-msr" mode, such as when running as a non-privileged user, it clears all added counters. Updating that to clear only actual MSR added counters allows regular users to see the sysfs counters. [lenb: commit message] Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-25 17:54:18 +02:00
while (*pp) {
p = *pp;
tools/power turbostat: Clear added counters when in no-msr mode If user request --no-msr or is not able to access the MSRs, turbostat should clear all the counters added with --add. Because MSR access permission checks are done after the cmdline is parsed, the decision has to be defered up until the transition into no-msr mode happen. Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-14 11:36:55 +01:00
tools/power turbostat: Enable non-privileged users to read sysfs counters A group of counters called "sysfs" displays software C-state request counts and resulting perceived C-state residency. They are not built-in counters that turbostat knows about ahead of time, rather they are discovered in sysfs when turbostat starts. Thus, they are added dynamically, using the same interface as user-added MSR counters. When turbostat enters "no-msr" mode, such as when running as a non-privileged user, it clears all added counters. Updating that to clear only actual MSR added counters allows regular users to see the sysfs counters. [lenb: commit message] Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-25 17:54:18 +02:00
if (p->msr_num != 0) {
*pp = p->next;
free(p);
++num_freed;
tools/power turbostat: Clear added counters when in no-msr mode If user request --no-msr or is not able to access the MSRs, turbostat should clear all the counters added with --add. Because MSR access permission checks are done after the cmdline is parsed, the decision has to be defered up until the transition into no-msr mode happen. Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-14 11:36:55 +01:00
tools/power turbostat: Enable non-privileged users to read sysfs counters A group of counters called "sysfs" displays software C-state request counts and resulting perceived C-state residency. They are not built-in counters that turbostat knows about ahead of time, rather they are discovered in sysfs when turbostat starts. Thus, they are added dynamically, using the same interface as user-added MSR counters. When turbostat enters "no-msr" mode, such as when running as a non-privileged user, it clears all added counters. Updating that to clear only actual MSR added counters allows regular users to see the sysfs counters. [lenb: commit message] Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-25 17:54:18 +02:00
continue;
}
pp = &p->next;
tools/power turbostat: Clear added counters when in no-msr mode If user request --no-msr or is not able to access the MSRs, turbostat should clear all the counters added with --add. Because MSR access permission checks are done after the cmdline is parsed, the decision has to be defered up until the transition into no-msr mode happen. Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-14 11:36:55 +01:00
}
tools/power turbostat: Enable non-privileged users to read sysfs counters A group of counters called "sysfs" displays software C-state request counts and resulting perceived C-state residency. They are not built-in counters that turbostat knows about ahead of time, rather they are discovered in sysfs when turbostat starts. Thus, they are added dynamically, using the same interface as user-added MSR counters. When turbostat enters "no-msr" mode, such as when running as a non-privileged user, it clears all added counters. Updating that to clear only actual MSR added counters allows regular users to see the sysfs counters. [lenb: commit message] Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-25 17:54:18 +02:00
return num_freed;
}
/*
* Free all added counters accessed via msr.
*/
static void free_sys_msr_counters(void)
{
/* Thread counters */
sys.added_thread_counters -= free_msr_counters_(&sys.tp);
/* Core counters */
sys.added_core_counters -= free_msr_counters_(&sys.cp);
/* Package counters */
sys.added_package_counters -= free_msr_counters_(&sys.pp);
tools/power turbostat: Clear added counters when in no-msr mode If user request --no-msr or is not able to access the MSRs, turbostat should clear all the counters added with --add. Because MSR access permission checks are done after the cmdline is parsed, the decision has to be defered up until the transition into no-msr mode happen. Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-14 11:36:55 +01:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
struct system_summary {
struct thread_data threads;
struct core_data cores;
struct pkg_data packages;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
} average;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
struct platform_counters {
struct rapl_counter energy_psys; /* MSR_PLATFORM_ENERGY_STATUS */
} platform_counters_odd, platform_counters_even;
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
struct cpu_topology {
int physical_package_id;
tools/power turbostat: Add Die column If the system has more than one software visible die per package, print a Die column. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-14 19:17:40 -08:00
int die_id;
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
int logical_cpu_id;
tools/power turbostat: Calculate additional node information for a package The code currently assumes each package has exactly one node. This is not the case for AMD systems and Intel systems with COD. AMD systems also may re-enumerate each node's core IDs starting at 0 (for example, an AMD processor may have two nodes, each with core IDs from 0 to 7). In order to properly enumerate the cores we need to track both the physical and logical node IDs. Add physical_node_id to track the node ID assigned by the kernel, and logical_node_id used by turbostat to track the nodes per package ie) a 0-based count within the package. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:30 -04:00
int physical_node_id;
int logical_node_id; /* 0-based count within the package */
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
int physical_core_id;
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
int thread_id;
tools/power turbostat: Allow using cpu device in perf counters on hybrid platforms Intel hybrid platforms expose different perf devices for P and E cores. Instead of one, "/sys/bus/event_source/devices/cpu" device, there are "/sys/bus/event_source/devices/{cpu_core,cpu_atom}". This, however makes it more complicated for the user, because most of the counters are available on both and had to be handled manually. This patch allows users to use "virtual" cpu device that is seemingly translated to cpu_core and cpu_atom perf devices, depending on the type of a CPU we are opening the counter for. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-20 18:47:59 +02:00
int type;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
cpu_set_t *put_ids; /* Processing Unit/Thread IDs */
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
} *cpus;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
struct topo_params {
int num_packages;
tools/power turbostat: Add Die column If the system has more than one software visible die per package, print a Die column. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-14 19:17:40 -08:00
int num_die;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
int num_cpus;
int num_cores;
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
int allowed_packages;
int allowed_cpus;
int allowed_cores;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
int max_cpu_num;
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs Save the highest core and package id when parsing topology to allocate enough memory when get_rapl_counters() is called with a core or a package id as a domain. Note that RAPL domains are per-package on Intel, but per-core on AMD. Thus, the RAPL code effectively runs in different modes on those two product lines. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-06 15:39:08 +02:00
int max_core_id;
int max_package_id;
tools/power turbostat: Remember global max_die_id This is necessary to gracefully handle sparse die_id's. no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 11:56:48 -04:00
int max_die_id;
tools/power turbostat: Calculate additional node information for a package The code currently assumes each package has exactly one node. This is not the case for AMD systems and Intel systems with COD. AMD systems also may re-enumerate each node's core IDs starting at 0 (for example, an AMD processor may have two nodes, each with core IDs from 0 to 7). In order to properly enumerate the cores we need to track both the physical and logical node IDs. Add physical_node_id to track the node ID assigned by the kernel, and logical_node_id used by turbostat to track the nodes per package ie) a 0-based count within the package. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:30 -04:00
int max_node_num;
tools/power turbostat: remove num_ from cpu_topology struct Cleanup, remove num_ from num_nodes_per_pkg, num_cores_per_node, and num_threads_per_node. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:33 -04:00
int nodes_per_pkg;
int cores_per_node;
int threads_per_core;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
} topo;
struct timeval tv_even, tv_odd, tv_delta;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int *irq_column_2_cpu; /* /proc/interrupts column numbers */
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
int *irqs_per_cpu; /* indexed by cpu_num */
tools/power/turbostat: Handle offlined CPUs in cpu_subset It is possible that the cpu_subset contains offlined CPUs. If this happens during start, exit immediately because this is likely an operator error that is best fixed by re-invoking. If this happens at runtime, give a warning only because turbostat should do its best effort to continue running. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 14:20:47 +08:00
void setup_all_buffers(bool startup);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: Fix missing SYS_LPI counter on some Chromebooks Some Chromebook BIOS' do not export an ACPI LPIT, which is how Linux finds the residency counter for CPU and SYSTEM low power states, that is exports in /sys/devices/system/cpu/cpuidle/*residency_us When these sysfs attributes are missing, check the debugfs attrubte from the pmc_core driver, which accesses the same counter value. Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 18:26:05 -04:00
char *sys_lpi_file;
char *sys_lpi_file_sysfs = "/sys/devices/system/cpu/cpuidle/low_power_idle_system_residency_us";
char *sys_lpi_file_debugfs = "/sys/kernel/debug/pmc_core/slp_s0_residency_usec";
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
int cpu_is_not_present(int cpu)
tools/power turbostat: fix un-intended affinity of forked program Linux 3.4 included a modification to turbostat to lower cross-call overhead by using scheduler affinity: 15aaa34654831e98dd76f7738b6c7f5d05a66430 (tools turbostat: reduce measurement overhead due to IPIs) In the use-case where turbostat forks a child program, that change had the un-intended side-effect of binding the child to the last cpu in the system. This change removed the binding before forking the child. This is a back-port of a fix already included in turbostat v2. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-03 23:24:00 -04:00
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return !CPU_ISSET_S(cpu, cpu_present_setsize, cpu_present_set);
tools/power turbostat: fix un-intended affinity of forked program Linux 3.4 included a modification to turbostat to lower cross-call overhead by using scheduler affinity: 15aaa34654831e98dd76f7738b6c7f5d05a66430 (tools turbostat: reduce measurement overhead due to IPIs) In the use-case where turbostat forks a child program, that change had the un-intended side-effect of binding the child to the last cpu in the system. This change removed the binding before forking the child. This is a back-port of a fix already included in turbostat v2. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-03 23:24:00 -04:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power/turbostat: Introduce cpu_allowed_set Turbostat supports "-c" parameter which limits output to system summary plus the specified cpu-set. But some code still uses cpu_present_set to read and dump the counters. Introduce cpu_allowed_set for code that should obey the specified cpu-set. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 11:01:17 +08:00
int cpu_is_not_allowed(int cpu)
{
return !CPU_ISSET_S(cpu, cpu_allowed_setsize, cpu_allowed_set);
}
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 21:44:40 -04:00
/*
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
* run func(thread, core, package) in topology order
* skip non-present cpus
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 21:44:40 -04:00
*/
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int for_all_cpus(int (func) (struct thread_data *, struct core_data *, struct pkg_data *),
struct thread_data *thread_base, struct core_data *core_base, struct pkg_data *pkg_base)
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 21:44:40 -04:00
{
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
int retval, pkg_no, core_no, thread_no, node_no;
tools/power turbostat: fix un-intended affinity of forked program Linux 3.4 included a modification to turbostat to lower cross-call overhead by using scheduler affinity: 15aaa34654831e98dd76f7738b6c7f5d05a66430 (tools turbostat: reduce measurement overhead due to IPIs) In the use-case where turbostat forks a child program, that change had the un-intended side-effect of binding the child to the last cpu in the system. This change removed the binding before forking the child. This is a back-port of a fix already included in turbostat v2. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-03 23:24:00 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
for (pkg_no = 0; pkg_no < topo.num_packages; ++pkg_no) {
tools/power turbostat: Restore ability to execute in topology-order turbostat executes on CPUs in "topology order". This is an optimization for measuring profoundly idle systems -- as the closest hardware is woken next... Fix a typo that was added with the sub-die-node support, that broke topology ordering on multi-node systems. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-15 21:58:23 -08:00
for (node_no = 0; node_no < topo.nodes_per_pkg; node_no++) {
for (core_no = 0; core_no < topo.cores_per_node; ++core_no) {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
for (thread_no = 0; thread_no < topo.threads_per_core; ++thread_no) {
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
struct thread_data *t;
struct core_data *c;
struct pkg_data *p;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
t = GET_THREAD(thread_base, thread_no, core_no, node_no, pkg_no);
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
tools/power/turbostat: Obey allowed CPUs when accessing CPU counters for_all_cpus/for_all_cpus_2 are used for accessing the per CPU counters, and they should follow the cpu_allowed_set instead of cpu_present_set. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 13:52:02 +08:00
if (cpu_is_not_allowed(t->cpu_id))
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
continue;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
c = GET_CORE(core_base, core_no, node_no, pkg_no);
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
p = GET_PKG(pkg_base, pkg_no);
retval = func(t, c, p);
if (retval)
return retval;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
}
}
return 0;
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 21:44:40 -04:00
}
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
int is_cpu_first_thread_in_core(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
UNUSED(p);
tools/power/turbostat: Obey allowed CPUs for primary thread/core detection Thread_id doesn't tell if a CPU is allowed or not. Detect allowed CPUs only and use the first detected thread/core as the primary thread/core of a core/package. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 09:59:14 +08:00
return ((int)t->cpu_id == c->base_cpu || c->base_cpu < 0);
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
}
int is_cpu_first_core_in_package(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
UNUSED(c);
tools/power/turbostat: Obey allowed CPUs for primary thread/core detection Thread_id doesn't tell if a CPU is allowed or not. Detect allowed CPUs only and use the first detected thread/core as the primary thread/core of a core/package. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 09:59:14 +08:00
return ((int)t->cpu_id == p->base_cpu || p->base_cpu < 0);
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
}
int is_cpu_first_thread_in_package(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
tools/power/turbostat: Obey allowed CPUs for primary thread/core detection Thread_id doesn't tell if a CPU is allowed or not. Detect allowed CPUs only and use the first detected thread/core as the primary thread/core of a core/package. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 09:59:14 +08:00
return is_cpu_first_thread_in_core(t, c, p) && is_cpu_first_core_in_package(t, c, p);
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
}
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 21:44:40 -04:00
int cpu_migrate(int cpu)
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
CPU_ZERO_S(cpu_affinity_setsize, cpu_affinity_set);
CPU_SET_S(cpu, cpu_affinity_setsize, cpu_affinity_set);
if (sched_setaffinity(0, cpu_affinity_setsize, cpu_affinity_set) == -1)
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 21:44:40 -04:00
return -1;
else
return 0;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
int get_msr_fd(int cpu)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
char pathname[32];
int fd;
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
fd = fd_percpu[cpu];
if (fd)
return fd;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
sprintf(pathname, "/dev/cpu/%d/msr", cpu);
fd = open(pathname, O_RDONLY);
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 22:19:58 -04:00
if (fd < 0)
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
err(-1, "%s open failed, try chown or chmod +r /dev/cpu/*/msr, "
"or run with --no-msr, or run as root", pathname);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
fd_percpu[cpu] = fd;
return fd;
}
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
static void bic_disable_msr_access(void)
{
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
const unsigned long bic_msrs = BIC_Mod_c6 | BIC_CoreTmp |
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
BIC_Totl_c0 | BIC_Any_c0 | BIC_GFX_c0 | BIC_CPUGFX | BIC_PkgTmp;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
bic_enabled &= ~bic_msrs;
tools/power turbostat: Clear added counters when in no-msr mode If user request --no-msr or is not able to access the MSRs, turbostat should clear all the counters added with --add. Because MSR access permission checks are done after the cmdline is parsed, the decision has to be defered up until the transition into no-msr mode happen. Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-14 11:36:55 +01:00
tools/power turbostat: Enable non-privileged users to read sysfs counters A group of counters called "sysfs" displays software C-state request counts and resulting perceived C-state residency. They are not built-in counters that turbostat knows about ahead of time, rather they are discovered in sysfs when turbostat starts. Thus, they are added dynamically, using the same interface as user-added MSR counters. When turbostat enters "no-msr" mode, such as when running as a non-privileged user, it clears all added counters. Updating that to clear only actual MSR added counters allows regular users to see the sysfs counters. [lenb: commit message] Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-25 17:54:18 +02:00
free_sys_msr_counters();
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
}
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
static long perf_event_open(struct perf_event_attr *hw_event, pid_t pid, int cpu, int group_fd, unsigned long flags)
{
tools/power turbostat: Add --no-perf option Add the --no-perf option to allow users to run turbostat without accessing perf. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:58:02 +01:00
assert(!no_perf);
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
return syscall(__NR_perf_event_open, hw_event, pid, cpu, group_fd, flags);
}
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
static long open_perf_counter(int cpu, unsigned int type, unsigned int config, int group_fd, __u64 read_format)
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
{
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
struct perf_event_attr attr;
const pid_t pid = -1;
const unsigned long flags = 0;
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
assert(!no_perf);
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
memset(&attr, 0, sizeof(struct perf_event_attr));
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
attr.type = type;
attr.size = sizeof(struct perf_event_attr);
attr.config = config;
attr.disabled = 0;
attr.sample_type = PERF_SAMPLE_IDENTIFIER;
attr.read_format = read_format;
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
const int fd = perf_event_open(&attr, pid, cpu, group_fd, flags);
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
return fd;
}
int get_instr_count_fd(int cpu)
{
if (fd_instr_count_percpu[cpu])
return fd_instr_count_percpu[cpu];
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
fd_instr_count_percpu[cpu] = open_perf_counter(cpu, PERF_TYPE_HARDWARE, PERF_COUNT_HW_INSTRUCTIONS, -1, 0);
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
return fd_instr_count_percpu[cpu];
}
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
int get_msr(int cpu, off_t offset, unsigned long long *msr)
{
ssize_t retval;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
assert(!no_msr);
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
retval = pread(get_msr_fd(cpu), msr, sizeof(*msr), offset);
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 22:19:58 -04:00
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 00:36:50 -04:00
if (retval != sizeof *msr)
tools/power turbostat: BYT does not have MSR_MISC_PWR_MGMT and so --debug fails with: turbostat: msr 1 offset 0x1aa read failed: Input/output error It seems that baytrail, and airmont do not have this MSR. It is included in subsequent Goldmont Atom. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-01 13:08:33 -05:00
err(-1, "cpu%d: msr offset 0x%llx read failed", cpu, (unsigned long long)offset);
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 22:19:58 -04:00
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
int probe_msr(int cpu, off_t offset)
{
ssize_t retval;
unsigned long long dummy;
assert(!no_msr);
retval = pread(get_msr_fd(cpu), &dummy, sizeof(dummy), offset);
if (retval != sizeof(dummy))
return 1;
return 0;
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
/* Convert CPU ID to domain ID for given added perf counter. */
unsigned int cpu_to_domain(const struct perf_counter_info *pc, int cpu)
{
switch (pc->scope) {
case SCOPE_CPU:
return cpu;
case SCOPE_CORE:
return cpus[cpu].physical_core_id;
case SCOPE_PACKAGE:
return cpus[cpu].physical_package_id;
}
__builtin_unreachable();
}
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
#define MAX_DEFERRED 16
tools/power turbostat: Allow -e for all names. Currently, there are a number of variables which are displayed by default, enabled with -e all, and listed by --list, but which you can not give to --enable/-e. So you can enable CPU0c1 (in the bic array), but you can't enable C1 or C1% (not in the bic array, but exists in sysfs). This runs counter to both the documentation and user expectations, and it's just not very user friendly. As such, the mechanism used by --hide has been duplicated, and is now also used by --enable, so we can handle unknown names gracefully. Note: One impact of this is that truly unknown fields given to --enable will no longer generate errors, they will be silently ignored, as --hide does. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:38 -07:00
char *deferred_add_names[MAX_DEFERRED];
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
char *deferred_skip_names[MAX_DEFERRED];
tools/power turbostat: Allow -e for all names. Currently, there are a number of variables which are displayed by default, enabled with -e all, and listed by --list, but which you can not give to --enable/-e. So you can enable CPU0c1 (in the bic array), but you can't enable C1 or C1% (not in the bic array, but exists in sysfs). This runs counter to both the documentation and user expectations, and it's just not very user friendly. As such, the mechanism used by --hide has been duplicated, and is now also used by --enable, so we can handle unknown names gracefully. Note: One impact of this is that truly unknown fields given to --enable will no longer generate errors, they will be silently ignored, as --hide does. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:38 -07:00
int deferred_add_index;
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
int deferred_skip_index;
/*
* HIDE_LIST - hide this list of counters, show the rest [default]
* SHOW_LIST - show this list of counters, hide the rest
*/
enum show_hide_mode { SHOW_LIST, HIDE_LIST } global_show_hide_mode = HIDE_LIST;
void help(void)
{
fprintf(outf,
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
"Usage: turbostat [OPTIONS][(--interval seconds) | COMMAND ...]\n"
"\n"
"Turbostat forks the specified COMMAND and prints statistics\n"
"when COMMAND completes.\n"
"If no COMMAND is specified, turbostat wakes every 5-seconds\n"
"to print statistics, until interrupted.\n"
" -a, --add add a counter\n"
" eg. --add msr0x10,u64,cpu,delta,MY_TSC\n"
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
" eg. --add perf/cstate_pkg/c2-residency,package,delta,percent,perfPC2\n"
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
" eg. --add pmt,name=XTAL,type=raw,domain=package0,offset=0,lsb=0,msb=63,guid=0x1a067102\n"
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
" -c, --cpu cpu-set limit output to summary plus cpu-set:\n"
" {core | package | j,k,l..m,n-p }\n"
" -d, --debug displays usec, Time_Of_Day_Seconds and more debugging\n"
tools/power turbostat: Move debug prints from stdout to stderr This leaves the stdout cleaner, having only counter data. It makes it easier for programs to parse the output of turbostat, for example selftests. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 11:34:38 +02:00
" debug messages are printed to stderr\n"
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
" -D, --Dump displays the raw counter values\n"
" -e, --enable [all | column]\n"
" shows all or the specified disabled column\n"
" -H, --hide [column|column,column,...]\n"
" hide the specified column(s)\n"
" -i, --interval sec.subsec\n"
" Override default 5-second measurement interval\n"
" -J, --Joules displays energy in Joules instead of Watts\n"
" -l, --list list column headers only\n"
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
" -M, --no-msr Disable all uses of the MSR driver\n"
tools/power turbostat: Add --no-perf option Add the --no-perf option to allow users to run turbostat without accessing perf. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:58:02 +01:00
" -P, --no-perf Disable all uses of the perf API\n"
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
" -n, --num_iterations num\n"
" number of the measurement iterations\n"
tools/power turbostat: Allow printing header every N iterations This gives the ability to reprint the header every N iterations, so you can ensure that a scrolling display always has the header visible somewhere on the screen. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:39 -07:00
" -N, --header_iterations num\n"
" print header every num iterations\n"
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
" -o, --out file\n"
" create or truncate \"file\" for all output\n"
" -q, --quiet skip decoding system configuration header\n"
" -s, --show [column|column,column,...]\n"
" show only the specified column(s)\n"
" -S, --Summary\n"
" limits output to 1-line system summary per interval\n"
" -T, --TCC temperature\n"
" sets the Thermal Control Circuit temperature in\n"
" degrees Celsius\n"
" -h, --help print this help message\n"
" -v, --version print version information\n" "\n" "For more help, run \"man turbostat\"\n");
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
}
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
/*
* bic_lookup
* for all the strings in comma separate name_list,
* set the approprate bit in return value.
*/
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
unsigned long long bic_lookup(char *name_list, enum show_hide_mode mode)
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
{
tools/power turbostat: No build warnings with -Wextra Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 23:45:18 -04:00
unsigned int i;
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
unsigned long long retval = 0;
while (name_list) {
char *comma;
comma = strchr(name_list, ',');
if (comma)
*comma = '\0';
for (i = 0; i < MAX_BIC; ++i) {
if (!strcmp(name_list, bic[i].name)) {
retval |= (1ULL << i);
break;
}
tools/power turbostat: tweak --show and --hide capability allow invocations such as # turbostat --show power,Busy% previously the "Busy%" was ignored Signed-off-by: Len Brown <len.brown@intel.com>
2021-07-16 11:55:03 -04:00
if (!strcmp(name_list, "all")) {
retval |= ~0;
break;
} else if (!strcmp(name_list, "topology")) {
retval |= BIC_TOPOLOGY;
break;
} else if (!strcmp(name_list, "power")) {
retval |= BIC_THERMAL_PWR;
break;
} else if (!strcmp(name_list, "idle")) {
retval |= BIC_IDLE;
break;
} else if (!strcmp(name_list, "frequency")) {
retval |= BIC_FREQUENCY;
break;
} else if (!strcmp(name_list, "other")) {
retval |= BIC_OTHER;
break;
}
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
}
if (i == MAX_BIC) {
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
if (mode == SHOW_LIST) {
tools/power turbostat: Allow -e for all names. Currently, there are a number of variables which are displayed by default, enabled with -e all, and listed by --list, but which you can not give to --enable/-e. So you can enable CPU0c1 (in the bic array), but you can't enable C1 or C1% (not in the bic array, but exists in sysfs). This runs counter to both the documentation and user expectations, and it's just not very user friendly. As such, the mechanism used by --hide has been duplicated, and is now also used by --enable, so we can handle unknown names gracefully. Note: One impact of this is that truly unknown fields given to --enable will no longer generate errors, they will be silently ignored, as --hide does. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:38 -07:00
deferred_add_names[deferred_add_index++] = name_list;
if (deferred_add_index >= MAX_DEFERRED) {
fprintf(stderr, "More than max %d un-recognized --add options '%s'\n",
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
MAX_DEFERRED, name_list);
tools/power turbostat: Allow -e for all names. Currently, there are a number of variables which are displayed by default, enabled with -e all, and listed by --list, but which you can not give to --enable/-e. So you can enable CPU0c1 (in the bic array), but you can't enable C1 or C1% (not in the bic array, but exists in sysfs). This runs counter to both the documentation and user expectations, and it's just not very user friendly. As such, the mechanism used by --hide has been duplicated, and is now also used by --enable, so we can handle unknown names gracefully. Note: One impact of this is that truly unknown fields given to --enable will no longer generate errors, they will be silently ignored, as --hide does. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:38 -07:00
help();
exit(1);
}
} else {
deferred_skip_names[deferred_skip_index++] = name_list;
if (debug)
fprintf(stderr, "deferred \"%s\"\n", name_list);
if (deferred_skip_index >= MAX_DEFERRED) {
fprintf(stderr, "More than max %d un-recognized --skip options '%s'\n",
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
MAX_DEFERRED, name_list);
tools/power turbostat: Allow -e for all names. Currently, there are a number of variables which are displayed by default, enabled with -e all, and listed by --list, but which you can not give to --enable/-e. So you can enable CPU0c1 (in the bic array), but you can't enable C1 or C1% (not in the bic array, but exists in sysfs). This runs counter to both the documentation and user expectations, and it's just not very user friendly. As such, the mechanism used by --hide has been duplicated, and is now also used by --enable, so we can handle unknown names gracefully. Note: One impact of this is that truly unknown fields given to --enable will no longer generate errors, they will be silently ignored, as --hide does. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:38 -07:00
help();
exit(1);
}
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
}
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
}
name_list = comma;
if (name_list)
name_list++;
}
return retval;
}
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 00:55:19 -05:00
tools/power turbostat: Add --list option to show available header names It is handy to know the list of column header names, so that they can be used with --add and --skip The new --list option shows them: sudo ./turbostat --list --hide sysfs ,Core,CPU,Avg_MHz,Busy%,Bzy_MHz,TSC_MHz,IRQ,SMI,CPU%c1,CPU%c3,CPU%c6,CPU%c7,CoreTmp,PkgTmp,GFX%rc6,GFXMHz,PkgWatt,CorWatt,GFXWatt Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 17:15:11 -05:00
void print_header(char *delim)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
struct msr_counter *mp;
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
struct perf_counter_info *pp;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
struct pmt_counter *ppmt;
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
int printed = 0;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
if (DO_BIC(BIC_USEC))
outp += sprintf(outp, "%susec", (printed++ ? delim : ""));
if (DO_BIC(BIC_TOD))
outp += sprintf(outp, "%sTime_Of_Day_Seconds", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Package))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sPackage", (printed++ ? delim : ""));
tools/power turbostat: Add Die column If the system has more than one software visible die per package, print a Die column. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-14 19:17:40 -08:00
if (DO_BIC(BIC_Die))
outp += sprintf(outp, "%sDie", (printed++ ? delim : ""));
tools/power turbostat: Add Node in output Output a Node column if there is more than one node/socket. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:35 -04:00
if (DO_BIC(BIC_Node))
outp += sprintf(outp, "%sNode", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Core))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sCore", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_CPU))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sCPU", (printed++ ? delim : ""));
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
if (DO_BIC(BIC_APIC))
outp += sprintf(outp, "%sAPIC", (printed++ ? delim : ""));
if (DO_BIC(BIC_X2APIC))
outp += sprintf(outp, "%sX2APIC", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Avg_MHz))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sAvg_MHz", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Busy))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sBusy%%", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Bzy_MHz))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sBzy_MHz", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_TSC_MHz))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sTSC_MHz", (printed++ ? delim : ""));
tools/power turbostat: simplify default output Casual turbostat users generally just want to know MHz. So by default, just print enough information to make sense of MHz. All the other configuration data and columns for C-states and temperature etc, are printed with the --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-23 00:34:57 -05:00
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
if (DO_BIC(BIC_IPC))
outp += sprintf(outp, "%sIPC", (printed++ ? delim : ""));
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
if (DO_BIC(BIC_IRQ)) {
if (sums_need_wide_columns)
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s IRQ", (printed++ ? delim : ""));
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
else
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sIRQ", (printed++ ? delim : ""));
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
}
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_SMI))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sSMI", (printed++ ? delim : ""));
tools/power turbostat: simplify default output Casual turbostat users generally just want to know MHz. So by default, just print enough information to make sense of MHz. All the other configuration data and columns for C-states and temperature etc, are printed with the --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-23 00:34:57 -05:00
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (mp = sys.tp; mp; mp = mp->next) {
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
if (mp->format == FORMAT_RAW) {
if (mp->width == 64)
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
outp += sprintf(outp, "%s%18.18s", (printed++ ? delim : ""), mp->name);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
else
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
outp += sprintf(outp, "%s%10.10s", (printed++ ? delim : ""), mp->name);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
} else {
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
if ((mp->type == COUNTER_ITEMS) && sums_need_wide_columns)
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
outp += sprintf(outp, "%s%8s", (printed++ ? delim : ""), mp->name);
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
else
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
outp += sprintf(outp, "%s%s", (printed++ ? delim : ""), mp->name);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
for (pp = sys.perf_tp; pp; pp = pp->next) {
if (pp->format == FORMAT_RAW) {
if (pp->width == 64)
outp += sprintf(outp, "%s%18.18s", (printed++ ? delim : ""), pp->name);
else
outp += sprintf(outp, "%s%10.10s", (printed++ ? delim : ""), pp->name);
} else {
if ((pp->type == COUNTER_ITEMS) && sums_need_wide_columns)
outp += sprintf(outp, "%s%8s", (printed++ ? delim : ""), pp->name);
else
outp += sprintf(outp, "%s%s", (printed++ ? delim : ""), pp->name);
}
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
ppmt = sys.pmt_tp;
while (ppmt) {
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
switch (ppmt->type) {
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
case PMT_TYPE_RAW:
if (pmt_counter_get_width(ppmt) <= 32)
outp += sprintf(outp, "%s%10.10s", (printed++ ? delim : ""), ppmt->name);
else
outp += sprintf(outp, "%s%18.18s", (printed++ ? delim : ""), ppmt->name);
break;
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
case PMT_TYPE_XTAL_TIME:
tools/power turbostat: Fix column printing for PMT xtal_time counters If the very first printed column was for a PMT counter of type xtal_time we would misalign the column header, because we were always printing the delimiter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-07 13:43:39 +02:00
outp += sprintf(outp, "%s%s", (printed++ ? delim : ""), ppmt->name);
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
break;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
}
ppmt = ppmt->next;
}
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
if (DO_BIC(BIC_CPU_c1))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sCPU%%c1", (printed++ ? delim : ""));
tools/power turbostat: Cleanup CC3-skip code no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-19 17:52:32 -04:00
if (DO_BIC(BIC_CPU_c3))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sCPU%%c3", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_CPU_c6))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sCPU%%c6", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_CPU_c7))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sCPU%%c7", (printed++ ? delim : ""));
tools/power turbostat: fix bugs in --add option When --add was used more than once, overflowed buffers caused some counters to be stored on top of others, corrupting the results. Simplify the code by simply reserving space for up to 16 added counters per each cpu, core, package. Per-cpu added counters were being printed only per-core. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 22:22:13 -05:00
tools/power turbostat: Baytrail c-state support The Baytrail SOC, with its Silvermont core, has some unique properties: 1. a hardware CC1 residency counter 2. a module-c6 residency counter 3. a package-c6 counter at traditional package-c7 counter address. The SOC does not support c3, pc3, c7 or pc7 counters. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:27:20 -05:00
if (DO_BIC(BIC_Mod_c6))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sMod%%c6", (printed++ ? delim : ""));
tools/power turbostat: fix bugs in --add option When --add was used more than once, overflowed buffers caused some counters to be stored on top of others, corrupting the results. Simplify the code by simply reserving space for up to 16 added counters per each cpu, core, package. Per-cpu added counters were being printed only per-core. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 22:22:13 -05:00
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_CoreTmp))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sCoreTmp", (printed++ ? delim : ""));
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
tools/power turbostat: Support thermal throttle count print The turbostat data is collected by end user for power evaluationit. However it looks like we are missing enough thermal context there. Already a couple of time we found that power management developer asking something like this: grep -r . /sys/devices/system/cpu/cpu*/thermal_throttle/* Print the per core thermal throttle count so as to get suffificent thermal context. turbostat -i 5 -s Core,CPU,CoreThr Core CPU CoreThr - - 104 0 0 61 0 4 1 1 0 1 5 2 2 104 2 6 3 3 7 3 7 Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-11-12 19:51:59 +08:00
if (DO_BIC(BIC_CORE_THROT_CNT))
outp += sprintf(outp, "%sCoreThr", (printed++ ? delim : ""));
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (platform->rapl_msrs && !rapl_joules) {
tools/power/turbostat: Abstract Per Core RAPL support Abstract the support for Per Core RAPL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:37:46 +08:00
if (DO_BIC(BIC_CorWatt) && platform->has_per_core_rapl)
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
outp += sprintf(outp, "%sCorWatt", (printed++ ? delim : ""));
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
} else if (platform->rapl_msrs && rapl_joules) {
tools/power/turbostat: Abstract Per Core RAPL support Abstract the support for Per Core RAPL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:37:46 +08:00
if (DO_BIC(BIC_Cor_J) && platform->has_per_core_rapl)
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
outp += sprintf(outp, "%sCor_J", (printed++ ? delim : ""));
}
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (mp = sys.cp; mp; mp = mp->next) {
if (mp->format == FORMAT_RAW) {
if (mp->width == 64)
tools/power turbostat: Add --list option to show available header names It is handy to know the list of column header names, so that they can be used with --add and --skip The new --list option shows them: sudo ./turbostat --list --hide sysfs ,Core,CPU,Avg_MHz,Busy%,Bzy_MHz,TSC_MHz,IRQ,SMI,CPU%c1,CPU%c3,CPU%c6,CPU%c7,CoreTmp,PkgTmp,GFX%rc6,GFXMHz,PkgWatt,CorWatt,GFXWatt Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 17:15:11 -05:00
outp += sprintf(outp, "%s%18.18s", delim, mp->name);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
else
tools/power turbostat: Add --list option to show available header names It is handy to know the list of column header names, so that they can be used with --add and --skip The new --list option shows them: sudo ./turbostat --list --hide sysfs ,Core,CPU,Avg_MHz,Busy%,Bzy_MHz,TSC_MHz,IRQ,SMI,CPU%c1,CPU%c3,CPU%c6,CPU%c7,CoreTmp,PkgTmp,GFX%rc6,GFXMHz,PkgWatt,CorWatt,GFXWatt Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 17:15:11 -05:00
outp += sprintf(outp, "%s%10.10s", delim, mp->name);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
} else {
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
if ((mp->type == COUNTER_ITEMS) && sums_need_wide_columns)
outp += sprintf(outp, "%s%8s", delim, mp->name);
else
outp += sprintf(outp, "%s%s", delim, mp->name);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
for (pp = sys.perf_cp; pp; pp = pp->next) {
if (pp->format == FORMAT_RAW) {
if (pp->width == 64)
outp += sprintf(outp, "%s%18.18s", (printed++ ? delim : ""), pp->name);
else
outp += sprintf(outp, "%s%10.10s", (printed++ ? delim : ""), pp->name);
} else {
if ((pp->type == COUNTER_ITEMS) && sums_need_wide_columns)
outp += sprintf(outp, "%s%8s", (printed++ ? delim : ""), pp->name);
else
outp += sprintf(outp, "%s%s", (printed++ ? delim : ""), pp->name);
}
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
ppmt = sys.pmt_cp;
while (ppmt) {
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
switch (ppmt->type) {
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
case PMT_TYPE_RAW:
if (pmt_counter_get_width(ppmt) <= 32)
outp += sprintf(outp, "%s%10.10s", (printed++ ? delim : ""), ppmt->name);
else
outp += sprintf(outp, "%s%18.18s", (printed++ ? delim : ""), ppmt->name);
break;
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
case PMT_TYPE_XTAL_TIME:
tools/power turbostat: Fix column printing for PMT xtal_time counters If the very first printed column was for a PMT counter of type xtal_time we would misalign the column header, because we were always printing the delimiter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-07 13:43:39 +02:00
outp += sprintf(outp, "%s%s", (printed++ ? delim : ""), ppmt->name);
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
break;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
}
ppmt = ppmt->next;
}
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_PkgTmp))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sPkgTmp", (printed++ ? delim : ""));
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_GFX_rc6))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sGFX%%rc6", (printed++ ? delim : ""));
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 01:28:12 -05:00
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_GFXMHz))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sGFXMHz", (printed++ ? delim : ""));
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 00:37:54 -05:00
tools/power turbostat: Add a new GFXAMHz column that exposes gt_act_freq_mhz. The column already present called GFXMHz reads from gt_cur_freq_mhz, which represents the GT frequency that was requested, but power management might not be able to do that. So the new column will display what the actual frequency GT is running at. Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-22 15:02:07 -07:00
if (DO_BIC(BIC_GFXACTMHz))
outp += sprintf(outp, "%sGFXAMHz", (printed++ ? delim : ""));
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
if (DO_BIC(BIC_SAM_mc6))
outp += sprintf(outp, "%sSAM%%mc6", (printed++ ? delim : ""));
if (DO_BIC(BIC_SAMMHz))
outp += sprintf(outp, "%sSAMMHz", (printed++ ? delim : ""));
if (DO_BIC(BIC_SAMACTMHz))
outp += sprintf(outp, "%sSAMAMHz", (printed++ ? delim : ""));
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_Totl_c0))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sTotl%%C0", (printed++ ? delim : ""));
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_Any_c0))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sAny%%C0", (printed++ ? delim : ""));
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_GFX_c0))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sGFX%%C0", (printed++ ? delim : ""));
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_CPUGFX))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sCPUGFX%%", (printed++ ? delim : ""));
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc2))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sPkg%%pc2", (printed++ ? delim : ""));
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc3))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sPkg%%pc3", (printed++ ? delim : ""));
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc6))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sPkg%%pc6", (printed++ ? delim : ""));
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc7))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sPkg%%pc7", (printed++ ? delim : ""));
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc8))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sPkg%%pc8", (printed++ ? delim : ""));
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc9))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sPkg%%pc9", (printed++ ? delim : ""));
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc10))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sPk%%pc10", (printed++ ? delim : ""));
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
if (DO_BIC(BIC_Diec6))
outp += sprintf(outp, "%sDie%%c6", (printed++ ? delim : ""));
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
if (DO_BIC(BIC_CPU_LPI))
outp += sprintf(outp, "%sCPU%%LPI", (printed++ ? delim : ""));
if (DO_BIC(BIC_SYS_LPI))
outp += sprintf(outp, "%sSYS%%LPI", (printed++ ? delim : ""));
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (platform->rapl_msrs && !rapl_joules) {
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_PkgWatt))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sPkgWatt", (printed++ ? delim : ""));
tools/power/turbostat: Abstract Per Core RAPL support Abstract the support for Per Core RAPL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:37:46 +08:00
if (DO_BIC(BIC_CorWatt) && !platform->has_per_core_rapl)
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sCorWatt", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_GFXWatt))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sGFXWatt", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_RAMWatt))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sRAMWatt", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_PKG__))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sPKG_%%", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_RAM__))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sRAM_%%", (printed++ ? delim : ""));
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
} else if (platform->rapl_msrs && rapl_joules) {
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Pkg_J))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sPkg_J", (printed++ ? delim : ""));
tools/power/turbostat: Abstract Per Core RAPL support Abstract the support for Per Core RAPL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:37:46 +08:00
if (DO_BIC(BIC_Cor_J) && !platform->has_per_core_rapl)
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sCor_J", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_GFX_J))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sGFX_J", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_RAM_J))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sRAM_J", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_PKG__))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sPKG_%%", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_RAM__))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%sRAM_%%", (printed++ ? delim : ""));
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-16 10:23:41 -08:00
}
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
if (DO_BIC(BIC_UNCORE_MHZ))
outp += sprintf(outp, "%sUncMHz", (printed++ ? delim : ""));
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (mp = sys.pp; mp; mp = mp->next) {
if (mp->format == FORMAT_RAW) {
if (mp->width == 64)
tools/power turbostat: Add --list option to show available header names It is handy to know the list of column header names, so that they can be used with --add and --skip The new --list option shows them: sudo ./turbostat --list --hide sysfs ,Core,CPU,Avg_MHz,Busy%,Bzy_MHz,TSC_MHz,IRQ,SMI,CPU%c1,CPU%c3,CPU%c6,CPU%c7,CoreTmp,PkgTmp,GFX%rc6,GFXMHz,PkgWatt,CorWatt,GFXWatt Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 17:15:11 -05:00
outp += sprintf(outp, "%s%18.18s", delim, mp->name);
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
else if (mp->width == 32)
tools/power turbostat: Add --list option to show available header names It is handy to know the list of column header names, so that they can be used with --add and --skip The new --list option shows them: sudo ./turbostat --list --hide sysfs ,Core,CPU,Avg_MHz,Busy%,Bzy_MHz,TSC_MHz,IRQ,SMI,CPU%c1,CPU%c3,CPU%c6,CPU%c7,CoreTmp,PkgTmp,GFX%rc6,GFXMHz,PkgWatt,CorWatt,GFXWatt Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 17:15:11 -05:00
outp += sprintf(outp, "%s%10.10s", delim, mp->name);
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
else
outp += sprintf(outp, "%s%7.7s", delim, mp->name);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
} else {
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
if ((mp->type == COUNTER_ITEMS) && sums_need_wide_columns)
outp += sprintf(outp, "%s%8s", delim, mp->name);
else
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
outp += sprintf(outp, "%s%7.7s", delim, mp->name);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
for (pp = sys.perf_pp; pp; pp = pp->next) {
if (pp->format == FORMAT_RAW) {
if (pp->width == 64)
outp += sprintf(outp, "%s%18.18s", (printed++ ? delim : ""), pp->name);
else
outp += sprintf(outp, "%s%10.10s", (printed++ ? delim : ""), pp->name);
} else {
if ((pp->type == COUNTER_ITEMS) && sums_need_wide_columns)
outp += sprintf(outp, "%s%8s", (printed++ ? delim : ""), pp->name);
else
outp += sprintf(outp, "%s%s", (printed++ ? delim : ""), pp->name);
}
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
ppmt = sys.pmt_pp;
while (ppmt) {
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
switch (ppmt->type) {
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
case PMT_TYPE_RAW:
if (pmt_counter_get_width(ppmt) <= 32)
outp += sprintf(outp, "%s%10.10s", (printed++ ? delim : ""), ppmt->name);
else
outp += sprintf(outp, "%s%18.18s", (printed++ ? delim : ""), ppmt->name);
break;
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
case PMT_TYPE_XTAL_TIME:
tools/power turbostat: Fix column printing for PMT xtal_time counters If the very first printed column was for a PMT counter of type xtal_time we would misalign the column header, because we were always printing the delimiter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-07 13:43:39 +02:00
outp += sprintf(outp, "%s%s", (printed++ ? delim : ""), ppmt->name);
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
break;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
}
ppmt = ppmt->next;
}
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
if (DO_BIC(BIC_SysWatt))
outp += sprintf(outp, "%sSysWatt", (printed++ ? delim : ""));
if (DO_BIC(BIC_Sys_J))
outp += sprintf(outp, "%sSys_J", (printed++ ? delim : ""));
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
outp += sprintf(outp, "\n");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int dump_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
int i;
struct msr_counter *mp;
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
struct platform_counters *pplat_cnt = p == package_odd ? &platform_counters_odd : &platform_counters_even;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
outp += sprintf(outp, "t %p, c %p, p %p\n", t, c, p);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
if (t) {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "CPU: %d flags 0x%x\n", t->cpu_id, t->flags);
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
outp += sprintf(outp, "TSC: %016llX\n", t->tsc);
outp += sprintf(outp, "aperf: %016llX\n", t->aperf);
outp += sprintf(outp, "mperf: %016llX\n", t->mperf);
outp += sprintf(outp, "c1: %016llX\n", t->c1);
tools/power turbostat: remove obsolete -M, -m, -C, -c options The new --add option has replaced the -M, -m, -C, -c options Eg. -M 0x10 is now --add msr0x10,raw -m 0x10 is now --add msr0x10,raw,u32 -C 0x10 is now --add msr0x10,delta -c 0x10 is now --add msr0x10,delta,u32 The --add option can be repeated to add any number of counters, while the previous options were limited to adding one of each type. In addition, the --add option can accept a column label, and can also display a counter as a percentage of elapsed cycles. Eg. --add msr0x3fe,core,percent,MY_CC3 Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-24 15:18:37 -05:00
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
if (DO_BIC(BIC_IPC))
outp += sprintf(outp, "IPC: %lld\n", t->instr_count);
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_IRQ))
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
outp += sprintf(outp, "IRQ: %lld\n", t->irq_count);
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_SMI))
tools/power turbostat: fix zero IRQ count shown in one-shot command mode The IRQ column has been working for periodic mode, but not in one-shot command mode, it shows only 0. until now. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-14 22:07:52 -05:00
outp += sprintf(outp, "SMI: %d\n", t->smi_count);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.tp; mp; i++, mp = mp->next) {
tools/power turbostat: enhance -D (debug counter dump) output Eliminate redundant debug output for core and package scope counters. Include name and path for all "ADDED" counters. Signed-off-by: Len Brown <len.brown@intel.com>
2024-02-05 15:56:25 -06:00
outp +=
sprintf(outp, "tADDED [%d] %8s msr0x%x: %08llX %s\n", i, mp->name, mp->msr_num,
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
t->counter[i], mp->sp->path);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: enhance -D (debug counter dump) output Eliminate redundant debug output for core and package scope counters. Include name and path for all "ADDED" counters. Signed-off-by: Len Brown <len.brown@intel.com>
2024-02-05 15:56:25 -06:00
if (c && is_cpu_first_thread_in_core(t, c, p)) {
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
outp += sprintf(outp, "core: %d\n", c->core_id);
outp += sprintf(outp, "c3: %016llX\n", c->c3);
outp += sprintf(outp, "c6: %016llX\n", c->c6);
outp += sprintf(outp, "c7: %016llX\n", c->c7);
outp += sprintf(outp, "DTS: %dC\n", c->core_temp_c);
tools/power turbostat: Support thermal throttle count print The turbostat data is collected by end user for power evaluationit. However it looks like we are missing enough thermal context there. Already a couple of time we found that power management developer asking something like this: grep -r . /sys/devices/system/cpu/cpu*/thermal_throttle/* Print the per core thermal throttle count so as to get suffificent thermal context. turbostat -i 5 -s Core,CPU,CoreThr Core CPU CoreThr - - 104 0 0 61 0 4 1 1 0 1 5 2 2 104 2 6 3 3 7 3 7 Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-11-12 19:51:59 +08:00
outp += sprintf(outp, "cpu_throt_count: %016llX\n", c->core_throt_cnt);
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
const unsigned long long energy_value = c->core_energy.raw_value * c->core_energy.scale;
const double energy_scale = c->core_energy.scale;
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
if (c->core_energy.unit == RAPL_UNIT_JOULES)
outp += sprintf(outp, "Joules: %0llX (scale: %lf)\n", energy_value, energy_scale);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) {
tools/power turbostat: enhance -D (debug counter dump) output Eliminate redundant debug output for core and package scope counters. Include name and path for all "ADDED" counters. Signed-off-by: Len Brown <len.brown@intel.com>
2024-02-05 15:56:25 -06:00
outp +=
sprintf(outp, "cADDED [%d] %8s msr0x%x: %08llX %s\n", i, mp->name, mp->msr_num,
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
c->counter[i], mp->sp->path);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
tools/power turbostat: Baytrail c-state support The Baytrail SOC, with its Silvermont core, has some unique properties: 1. a hardware CC1 residency counter 2. a module-c6 residency counter 3. a package-c6 counter at traditional package-c7 counter address. The SOC does not support c3, pc3, c7 or pc7 counters. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:27:20 -05:00
outp += sprintf(outp, "mc6_us: %016llX\n", c->mc6_us);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: enhance -D (debug counter dump) output Eliminate redundant debug output for core and package scope counters. Include name and path for all "ADDED" counters. Signed-off-by: Len Brown <len.brown@intel.com>
2024-02-05 15:56:25 -06:00
if (p && is_cpu_first_core_in_package(t, c, p)) {
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
outp += sprintf(outp, "package: %d\n", p->package_id);
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
outp += sprintf(outp, "Weighted cores: %016llX\n", p->pkg_wtd_core_c0);
outp += sprintf(outp, "Any cores: %016llX\n", p->pkg_any_core_c0);
outp += sprintf(outp, "Any GFX: %016llX\n", p->pkg_any_gfxe_c0);
outp += sprintf(outp, "CPU + GFX: %016llX\n", p->pkg_both_core_gfxe_c0);
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
outp += sprintf(outp, "pc2: %016llX\n", p->pc2);
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc3))
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-09 23:39:45 -05:00
outp += sprintf(outp, "pc3: %016llX\n", p->pc3);
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc6))
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-09 23:39:45 -05:00
outp += sprintf(outp, "pc6: %016llX\n", p->pc6);
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc7))
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-09 23:39:45 -05:00
outp += sprintf(outp, "pc7: %016llX\n", p->pc7);
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
outp += sprintf(outp, "pc8: %016llX\n", p->pc8);
outp += sprintf(outp, "pc9: %016llX\n", p->pc9);
outp += sprintf(outp, "pc10: %016llX\n", p->pc10);
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
outp += sprintf(outp, "cpu_lpi: %016llX\n", p->cpu_lpi);
outp += sprintf(outp, "sys_lpi: %016llX\n", p->sys_lpi);
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
outp += sprintf(outp, "Joules PKG: %0llX\n", p->energy_pkg.raw_value);
outp += sprintf(outp, "Joules COR: %0llX\n", p->energy_cores.raw_value);
outp += sprintf(outp, "Joules GFX: %0llX\n", p->energy_gfx.raw_value);
outp += sprintf(outp, "Joules RAM: %0llX\n", p->energy_dram.raw_value);
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
outp += sprintf(outp, "Joules PSYS: %0llX\n", pplat_cnt->energy_psys.raw_value);
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
outp += sprintf(outp, "Throttle PKG: %0llX\n", p->rapl_pkg_perf_status.raw_value);
outp += sprintf(outp, "Throttle RAM: %0llX\n", p->rapl_dram_perf_status.raw_value);
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
outp += sprintf(outp, "PTM: %dC\n", p->pkg_temp_c);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.pp; mp; i++, mp = mp->next) {
tools/power turbostat: enhance -D (debug counter dump) output Eliminate redundant debug output for core and package scope counters. Include name and path for all "ADDED" counters. Signed-off-by: Len Brown <len.brown@intel.com>
2024-02-05 15:56:25 -06:00
outp +=
sprintf(outp, "pADDED [%d] %8s msr0x%x: %08llX %s\n", i, mp->name, mp->msr_num,
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
p->counter[i], mp->sp->path);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
outp += sprintf(outp, "\n");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
double rapl_counter_get_value(const struct rapl_counter *c, enum rapl_unit desired_unit, double interval)
{
assert(desired_unit != RAPL_UNIT_INVALID);
/*
* For now we don't expect anything other than joules,
* so just simplify the logic.
*/
assert(c->unit == RAPL_UNIT_JOULES);
const double scaled = c->raw_value * c->scale;
if (desired_unit == RAPL_UNIT_WATTS)
return scaled / interval;
return scaled;
}
tools turbostat: add summary option turbostat -s cuts down on the amount of output, per user request. also treak some output whitespace and the man page. Signed-off-by: Len Brown <len.brown@intel.com>
2012-02-06 18:37:16 -05:00
/*
* column formatting convention & formats
*/
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int format_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
static int count;
struct platform_counters *pplat_cnt = NULL;
tools/power turbostat: use tsc_tweak everwhere it is needed The CPU ticks at a rate in the "bus clock" domain. eg. 100 MHz * bus_ratio. On newer processors, the TSC has been moved out of this BCLK domain and into a separate crystal-clock domain. While the TSC ticks "close to" the base frequency, those that look closely at the numbers will notice small errors in calculations that mix units of TSC clocks and bus clocks. "tsc_tweak" was introduced to address the most visible mixing -- the %Busy and the the Busy_MHz calculations. (A simplification as since removed TSC from the BusyMHz calculation) Here we apply the tsc_tweak to everyplace where BCLK and TSC units are mixed. The results is that on a system which is 100% idle, the sum of the C-states are now much more likely to be closer to 100%. Reported-by: Travis Downs <travis.downs@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:29:51 -05:00
double interval_float, tsc;
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 00:55:19 -05:00
char *fmt8;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
int i;
struct msr_counter *mp;
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
struct perf_counter_info *pp;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
struct pmt_counter *ppmt;
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
char *delim = "\t";
int printed = 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
if (t == &average.threads) {
pplat_cnt = count & 1 ? &platform_counters_odd : &platform_counters_even;
++count;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
/* if showing only 1st thread in core and this isn't one, bail out */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (show_core_only && !is_cpu_first_thread_in_core(t, c, p))
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return 0;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
/* if showing only 1st thread in pkg and this isn't one, bail out */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (show_pkg_only && !is_cpu_first_core_in_package(t, c, p))
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return 0;
tools/power turbostat: add --cpu parameter With the --cpu parameter, turbostat prints only lines for the specified set of CPUs: sudo ./turbostat --quiet --show Core,CPU --cpu 0,1,3..5,6-7 Core CPU - - 0 0 0 4 1 1 1 5 2 6 3 3 3 7 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 23:54:15 -05:00
/*if not summary line and --cpu is used */
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
if ((t != &average.threads) && (cpu_subset && !CPU_ISSET_S(t->cpu_id, cpu_subset_size, cpu_subset)))
tools/power turbostat: add --cpu parameter With the --cpu parameter, turbostat prints only lines for the specified set of CPUs: sudo ./turbostat --quiet --show Core,CPU --cpu 0,1,3..5,6-7 Core CPU - - 0 0 0 4 1 1 1 5 2 6 3 3 3 7 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 23:54:15 -05:00
return 0;
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
if (DO_BIC(BIC_USEC)) {
tools/power turbostat: if --debug, print sampling overhead The --debug option now pre-pends each row with the number of micro-seconds [usec] to collect the finishing snapshot for that row. Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-27 21:06:55 -07:00
/* on each row, print how many usec each timestamp took to gather */
struct timeval tv;
timersub(&t->tv_end, &t->tv_begin, &tv);
outp += sprintf(outp, "%5ld\t", tv.tv_sec * 1000000 + tv.tv_usec);
}
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
/* Time_Of_Day_Seconds: on each row, print sec.usec last timestamp taken */
if (DO_BIC(BIC_TOD))
outp += sprintf(outp, "%10ld.%06ld\t", t->tv_end.tv_sec, t->tv_end.tv_usec);
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
interval_float = t->tv_delta.tv_sec + t->tv_delta.tv_usec / 1000000.0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: use tsc_tweak everwhere it is needed The CPU ticks at a rate in the "bus clock" domain. eg. 100 MHz * bus_ratio. On newer processors, the TSC has been moved out of this BCLK domain and into a separate crystal-clock domain. While the TSC ticks "close to" the base frequency, those that look closely at the numbers will notice small errors in calculations that mix units of TSC clocks and bus clocks. "tsc_tweak" was introduced to address the most visible mixing -- the %Busy and the the Busy_MHz calculations. (A simplification as since removed TSC from the BusyMHz calculation) Here we apply the tsc_tweak to everyplace where BCLK and TSC units are mixed. The results is that on a system which is 100% idle, the sum of the C-states are now much more likely to be closer to 100%. Reported-by: Travis Downs <travis.downs@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:29:51 -05:00
tsc = t->tsc * tsc_tweak;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/* topo columns, print blanks on 1st (average) line */
if (t == &average.threads) {
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Package))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s-", (printed++ ? delim : ""));
tools/power turbostat: Add Die column If the system has more than one software visible die per package, print a Die column. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-14 19:17:40 -08:00
if (DO_BIC(BIC_Die))
outp += sprintf(outp, "%s-", (printed++ ? delim : ""));
tools/power turbostat: Add Node in output Output a Node column if there is more than one node/socket. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:35 -04:00
if (DO_BIC(BIC_Node))
outp += sprintf(outp, "%s-", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Core))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s-", (printed++ ? delim : ""));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_CPU))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s-", (printed++ ? delim : ""));
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
if (DO_BIC(BIC_APIC))
outp += sprintf(outp, "%s-", (printed++ ? delim : ""));
if (DO_BIC(BIC_X2APIC))
outp += sprintf(outp, "%s-", (printed++ ? delim : ""));
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
} else {
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Package)) {
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
if (p)
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), p->package_id);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
else
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s-", (printed++ ? delim : ""));
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools/power turbostat: Add Die column If the system has more than one software visible die per package, print a Die column. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-14 19:17:40 -08:00
if (DO_BIC(BIC_Die)) {
if (c)
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), cpus[t->cpu_id].die_id);
else
outp += sprintf(outp, "%s-", (printed++ ? delim : ""));
}
tools/power turbostat: Add Node in output Output a Node column if there is more than one node/socket. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:35 -04:00
if (DO_BIC(BIC_Node)) {
if (t)
outp += sprintf(outp, "%s%d",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
(printed++ ? delim : ""), cpus[t->cpu_id].physical_node_id);
tools/power turbostat: Add Node in output Output a Node column if there is more than one node/socket. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:35 -04:00
else
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s-", (printed++ ? delim : ""));
tools/power turbostat: Add Node in output Output a Node column if there is more than one node/socket. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:35 -04:00
}
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Core)) {
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
if (c)
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), c->core_id);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
else
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s-", (printed++ ? delim : ""));
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_CPU))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), t->cpu_id);
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
if (DO_BIC(BIC_APIC))
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), t->apic_id);
if (DO_BIC(BIC_X2APIC))
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), t->x2apic_id);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 00:55:19 -05:00
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Avg_MHz))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.0f", (printed++ ? delim : ""), 1.0 / units * t->aperf / interval_float);
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 00:55:19 -05:00
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Busy))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * t->mperf / tsc);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Bzy_MHz)) {
tools/power turbostat: simplify Bzy_MHz calculation Bzy_MHz = TSC_delta*tsc_tweak/APERF_delta/MPERF_delta/measurement_interval becomes Bzy_MHz = base_mhz/APERF_delta/MPERF_delta on systems which support MSR_NHM_PLATFORM_INFO. base_mhz is calculated directly from the base_ratio reported in MSR_NHM_PLATFORM_INFO * bclk, and bclk is discovered via MSR or cpuid. This reduces the dependency of Bzy_MHz calculation on the TSC. Previously, there were 4 TSC readings required in each caculation, the raw TSC delta combined with the measurement_interval. This also removes the "tsc_tweak" correction factor used when TSC runs on a different base clock from the CPU's bclk. After this change, tsc_tweak is used only for %Busy. The end-result should be a Bzy_MHz result slightly less prone to jitter. Signed-off-by: Len Brown <len.brown@intel.com>
2015-10-19 22:37:40 -04:00
if (has_base_hz)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp +=
sprintf(outp, "%s%.0f", (printed++ ? delim : ""), base_hz / units * t->aperf / t->mperf);
tools/power turbostat: simplify Bzy_MHz calculation Bzy_MHz = TSC_delta*tsc_tweak/APERF_delta/MPERF_delta/measurement_interval becomes Bzy_MHz = base_mhz/APERF_delta/MPERF_delta on systems which support MSR_NHM_PLATFORM_INFO. base_mhz is calculated directly from the base_ratio reported in MSR_NHM_PLATFORM_INFO * bclk, and bclk is discovered via MSR or cpuid. This reduces the dependency of Bzy_MHz calculation on the TSC. Previously, there were 4 TSC readings required in each caculation, the raw TSC delta combined with the measurement_interval. This also removes the "tsc_tweak" correction factor used when TSC runs on a different base clock from the CPU's bclk. After this change, tsc_tweak is used only for %Busy. The end-result should be a Bzy_MHz result slightly less prone to jitter. Signed-off-by: Len Brown <len.brown@intel.com>
2015-10-19 22:37:40 -04:00
else
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%.0f", (printed++ ? delim : ""),
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tsc / units * t->aperf / t->mperf / interval_float);
tools/power turbostat: simplify Bzy_MHz calculation Bzy_MHz = TSC_delta*tsc_tweak/APERF_delta/MPERF_delta/measurement_interval becomes Bzy_MHz = base_mhz/APERF_delta/MPERF_delta on systems which support MSR_NHM_PLATFORM_INFO. base_mhz is calculated directly from the base_ratio reported in MSR_NHM_PLATFORM_INFO * bclk, and bclk is discovered via MSR or cpuid. This reduces the dependency of Bzy_MHz calculation on the TSC. Previously, there were 4 TSC readings required in each caculation, the raw TSC delta combined with the measurement_interval. This also removes the "tsc_tweak" correction factor used when TSC runs on a different base clock from the CPU's bclk. After this change, tsc_tweak is used only for %Busy. The end-result should be a Bzy_MHz result slightly less prone to jitter. Signed-off-by: Len Brown <len.brown@intel.com>
2015-10-19 22:37:40 -04:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_TSC_MHz))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.0f", (printed++ ? delim : ""), 1.0 * t->tsc / units / interval_float);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
if (DO_BIC(BIC_IPC))
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 1.0 * t->instr_count / t->aperf);
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
/* IRQ */
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
if (DO_BIC(BIC_IRQ)) {
if (sums_need_wide_columns)
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%8lld", (printed++ ? delim : ""), t->irq_count);
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
else
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%lld", (printed++ ? delim : ""), t->irq_count);
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
}
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
tools/power turbostat: simplify default output Casual turbostat users generally just want to know MHz. So by default, just print enough information to make sense of MHz. All the other configuration data and columns for C-states and temperature etc, are printed with the --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-23 00:34:57 -05:00
/* SMI */
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_SMI))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), t->smi_count);
tools/power turbostat: simplify default output Casual turbostat users generally just want to know MHz. So by default, just print enough information to make sense of MHz. All the other configuration data and columns for C-states and temperature etc, are printed with the --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-23 00:34:57 -05:00
tools/power turbostat: fix bugs in --add option When --add was used more than once, overflowed buffers caused some counters to be stored on top of others, corrupting the results. Simplify the code by simply reserving space for up to 16 added counters per each cpu, core, package. Per-cpu added counters were being printed only per-core. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 22:22:13 -05:00
/* Added counters */
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.tp; mp; i++, mp = mp->next) {
if (mp->format == FORMAT_RAW) {
if (mp->width == 32)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp +=
sprintf(outp, "%s0x%08x", (printed++ ? delim : ""), (unsigned int)t->counter[i]);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
else
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s0x%016llx", (printed++ ? delim : ""), t->counter[i]);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
} else if (mp->format == FORMAT_DELTA) {
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
if ((mp->type == COUNTER_ITEMS) && sums_need_wide_columns)
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%8lld", (printed++ ? delim : ""), t->counter[i]);
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
else
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%lld", (printed++ ? delim : ""), t->counter[i]);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
} else if (mp->format == FORMAT_PERCENT) {
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
if (mp->type == COUNTER_USEC)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp +=
sprintf(outp, "%s%.2f", (printed++ ? delim : ""),
t->counter[i] / interval_float / 10000);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
else
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * t->counter[i] / tsc);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
/* Added perf counters */
for (i = 0, pp = sys.perf_tp; pp; ++i, pp = pp->next) {
if (pp->format == FORMAT_RAW) {
if (pp->width == 32)
outp +=
sprintf(outp, "%s0x%08x", (printed++ ? delim : ""),
(unsigned int)t->perf_counter[i]);
else
outp += sprintf(outp, "%s0x%016llx", (printed++ ? delim : ""), t->perf_counter[i]);
} else if (pp->format == FORMAT_DELTA) {
if ((pp->type == COUNTER_ITEMS) && sums_need_wide_columns)
outp += sprintf(outp, "%s%8lld", (printed++ ? delim : ""), t->perf_counter[i]);
else
outp += sprintf(outp, "%s%lld", (printed++ ? delim : ""), t->perf_counter[i]);
} else if (pp->format == FORMAT_PERCENT) {
if (pp->type == COUNTER_USEC)
outp +=
sprintf(outp, "%s%.2f", (printed++ ? delim : ""),
t->perf_counter[i] / interval_float / 10000);
else
outp +=
sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * t->perf_counter[i] / tsc);
}
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
for (i = 0, ppmt = sys.pmt_tp; ppmt; i++, ppmt = ppmt->next) {
tools/power turbostat: fix GCC9 build regression Fix build regression seen when using old gcc-9 compiler. Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Reviewed-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-31 12:24:09 -04:00
const unsigned long value_raw = t->pmt_counter[i];
const double value_converted = 100.0 * value_raw / crystal_hz / interval_float;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
switch (ppmt->type) {
case PMT_TYPE_RAW:
if (pmt_counter_get_width(ppmt) <= 32)
outp += sprintf(outp, "%s0x%08x", (printed++ ? delim : ""),
(unsigned int)t->pmt_counter[i]);
else
outp += sprintf(outp, "%s0x%016llx", (printed++ ? delim : ""), t->pmt_counter[i]);
break;
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
case PMT_TYPE_XTAL_TIME:
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), value_converted);
break;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
}
}
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
/* C1 */
if (DO_BIC(BIC_CPU_c1))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * t->c1 / tsc);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
tools/power turbostat: fix bugs in --add option When --add was used more than once, overflowed buffers caused some counters to be stored on top of others, corrupting the results. Simplify the code by simply reserving space for up to 16 added counters per each cpu, core, package. Per-cpu added counters were being printed only per-core. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 22:22:13 -05:00
/* print per-core data only for 1st thread in core */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (!is_cpu_first_thread_in_core(t, c, p))
tools/power turbostat: fix bugs in --add option When --add was used more than once, overflowed buffers caused some counters to be stored on top of others, corrupting the results. Simplify the code by simply reserving space for up to 16 added counters per each cpu, core, package. Per-cpu added counters were being printed only per-core. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 22:22:13 -05:00
goto done;
tools/power turbostat: Cleanup CC3-skip code no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-19 17:52:32 -04:00
if (DO_BIC(BIC_CPU_c3))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * c->c3 / tsc);
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_CPU_c6))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * c->c6 / tsc);
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_CPU_c7))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * c->c7 / tsc);
tools/power turbostat: fix bugs in --add option When --add was used more than once, overflowed buffers caused some counters to be stored on top of others, corrupting the results. Simplify the code by simply reserving space for up to 16 added counters per each cpu, core, package. Per-cpu added counters were being printed only per-core. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 22:22:13 -05:00
tools/power turbostat: Baytrail c-state support The Baytrail SOC, with its Silvermont core, has some unique properties: 1. a hardware CC1 residency counter 2. a module-c6 residency counter 3. a package-c6 counter at traditional package-c7 counter address. The SOC does not support c3, pc3, c7 or pc7 counters. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:27:20 -05:00
/* Mod%c6 */
if (DO_BIC(BIC_Mod_c6))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * c->mc6_us / tsc);
tools/power turbostat: Baytrail c-state support The Baytrail SOC, with its Silvermont core, has some unique properties: 1. a hardware CC1 residency counter 2. a module-c6 residency counter 3. a package-c6 counter at traditional package-c7 counter address. The SOC does not support c3, pc3, c7 or pc7 counters. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:27:20 -05:00
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_CoreTmp))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), c->core_temp_c);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: Support thermal throttle count print The turbostat data is collected by end user for power evaluationit. However it looks like we are missing enough thermal context there. Already a couple of time we found that power management developer asking something like this: grep -r . /sys/devices/system/cpu/cpu*/thermal_throttle/* Print the per core thermal throttle count so as to get suffificent thermal context. turbostat -i 5 -s Core,CPU,CoreThr Core CPU CoreThr - - 104 0 0 61 0 4 1 1 0 1 5 2 2 104 2 6 3 3 7 3 7 Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-11-12 19:51:59 +08:00
/* Core throttle count */
if (DO_BIC(BIC_CORE_THROT_CNT))
outp += sprintf(outp, "%s%lld", (printed++ ? delim : ""), c->core_throt_cnt);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) {
if (mp->format == FORMAT_RAW) {
if (mp->width == 32)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp +=
sprintf(outp, "%s0x%08x", (printed++ ? delim : ""), (unsigned int)c->counter[i]);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
else
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s0x%016llx", (printed++ ? delim : ""), c->counter[i]);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
} else if (mp->format == FORMAT_DELTA) {
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
if ((mp->type == COUNTER_ITEMS) && sums_need_wide_columns)
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%8lld", (printed++ ? delim : ""), c->counter[i]);
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
else
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%lld", (printed++ ? delim : ""), c->counter[i]);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
} else if (mp->format == FORMAT_PERCENT) {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * c->counter[i] / tsc);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
for (i = 0, pp = sys.perf_cp; pp; i++, pp = pp->next) {
if (pp->format == FORMAT_RAW) {
if (pp->width == 32)
outp +=
sprintf(outp, "%s0x%08x", (printed++ ? delim : ""),
(unsigned int)c->perf_counter[i]);
else
outp += sprintf(outp, "%s0x%016llx", (printed++ ? delim : ""), c->perf_counter[i]);
} else if (pp->format == FORMAT_DELTA) {
if ((pp->type == COUNTER_ITEMS) && sums_need_wide_columns)
outp += sprintf(outp, "%s%8lld", (printed++ ? delim : ""), c->perf_counter[i]);
else
outp += sprintf(outp, "%s%lld", (printed++ ? delim : ""), c->perf_counter[i]);
} else if (pp->format == FORMAT_PERCENT) {
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * c->perf_counter[i] / tsc);
}
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
for (i = 0, ppmt = sys.pmt_cp; ppmt; i++, ppmt = ppmt->next) {
tools/power turbostat: fix GCC9 build regression Fix build regression seen when using old gcc-9 compiler. Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Reviewed-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-31 12:24:09 -04:00
const unsigned long value_raw = c->pmt_counter[i];
const double value_converted = 100.0 * value_raw / crystal_hz / interval_float;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
switch (ppmt->type) {
case PMT_TYPE_RAW:
if (pmt_counter_get_width(ppmt) <= 32)
outp += sprintf(outp, "%s0x%08x", (printed++ ? delim : ""),
(unsigned int)c->pmt_counter[i]);
else
outp += sprintf(outp, "%s0x%016llx", (printed++ ? delim : ""), c->pmt_counter[i]);
break;
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
case PMT_TYPE_XTAL_TIME:
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), value_converted);
break;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
}
}
tools/power turbostat: Enable accumulate RAPL display Enable the accumulated RAPL display by default. Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:32:05 +08:00
fmt8 = "%s%.2f";
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
tools/power/turbostat: Abstract Per Core RAPL support Abstract the support for Per Core RAPL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:37:46 +08:00
if (DO_BIC(BIC_CorWatt) && platform->has_per_core_rapl)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp +=
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
sprintf(outp, fmt8, (printed++ ? delim : ""),
rapl_counter_get_value(&c->core_energy, RAPL_UNIT_WATTS, interval_float));
tools/power/turbostat: Abstract Per Core RAPL support Abstract the support for Per Core RAPL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:37:46 +08:00
if (DO_BIC(BIC_Cor_J) && platform->has_per_core_rapl)
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
outp += sprintf(outp, fmt8, (printed++ ? delim : ""),
rapl_counter_get_value(&c->core_energy, RAPL_UNIT_JOULES, interval_float));
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/* print per-package data only for 1st core in package */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (!is_cpu_first_core_in_package(t, c, p))
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
goto done;
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
/* PkgTmp */
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_PkgTmp))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), p->pkg_temp_c);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 01:28:12 -05:00
/* GFXrc6 */
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_GFX_rc6)) {
tools/power turbostat: Gracefully handle ACPI S3 turbostat gives valid results across suspend to idle, aka freeze, whether invoked in interval mode, or in command mode. Indeed, this can be used to measure suspend to idle: turbostat echo freeze > /sys/power/state But this does not work across suspend to ACPI S3, because the processor counters, including the TSC, are reset on resume. Further, when turbostat detects a problem, it does't forgive the hardware, and interval mode will print *'s from there on out. Instead, upon detecting counters going backwards, simply reset and start over. Interval mode across ACPI S3: (observe TSC going backwards) root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 1 0.06 858 2294 0x0000000000000000 0 0 0.06 847 2294 0x0000002a254b98ac 1 1 0.06 878 2294 0x0000002a254efa3a 2 1 0.07 843 2294 0x0000002a2551df65 3 0 0.05 863 2294 0x0000002a2553fea2 turbostat: re-initialized with num_cpus 4 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 2 0.20 849 2294 0x0000000000000000 0 2 0.26 856 2294 0x0000000449abb60d 1 2 0.20 844 2294 0x0000000449b087ec 2 2 0.21 850 2294 0x0000000449b35d5d 3 1 0.12 839 2294 0x0000000449b5fd5a ^C Command mode across ACPI S3: root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 sleep 10 ./turbostat: Counter reset detected 14.196299 sec Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 20:31:46 -04:00
if (p->gfx_rc6_ms == -1) { /* detect GFX counter reset */
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s**.**", (printed++ ? delim : ""));
tools/power turbostat: work around RC6 counter wrap Sometimes the rc6 sysfs counter spontaneously resets, causing turbostat prints a very large number as it tries to calcuate % = 100 * (old - new) / interval When we see (old > new), print ***.**% instead of a bogus huge number. Note that this detection is not fool-proof, as the counter could reset several times and still result in new > old. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-06 17:16:00 -04:00
} else {
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""),
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
p->gfx_rc6_ms / 10.0 / interval_float);
tools/power turbostat: work around RC6 counter wrap Sometimes the rc6 sysfs counter spontaneously resets, causing turbostat prints a very large number as it tries to calcuate % = 100 * (old - new) / interval When we see (old > new), print ***.**% instead of a bogus huge number. Note that this detection is not fool-proof, as the counter could reset several times and still result in new > old. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-06 17:16:00 -04:00
}
}
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 01:28:12 -05:00
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 00:37:54 -05:00
/* GFXMHz */
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_GFXMHz))
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), p->gfx_mhz);
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 00:37:54 -05:00
tools/power turbostat: Add a new GFXAMHz column that exposes gt_act_freq_mhz. The column already present called GFXMHz reads from gt_cur_freq_mhz, which represents the GT frequency that was requested, but power management might not be able to do that. So the new column will display what the actual frequency GT is running at. Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-22 15:02:07 -07:00
/* GFXACTMHz */
if (DO_BIC(BIC_GFXACTMHz))
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), p->gfx_act_mhz);
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
/* SAMmc6 */
if (DO_BIC(BIC_SAM_mc6)) {
if (p->sam_mc6_ms == -1) { /* detect GFX counter reset */
outp += sprintf(outp, "%s**.**", (printed++ ? delim : ""));
} else {
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""),
p->sam_mc6_ms / 10.0 / interval_float);
}
}
/* SAMMHz */
if (DO_BIC(BIC_SAMMHz))
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), p->sam_mhz);
/* SAMACTMHz */
if (DO_BIC(BIC_SAMACTMHz))
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), p->sam_act_mhz);
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
/* Totl%C0, Any%C0 GFX%C0 CPUGFX% */
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_Totl_c0))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pkg_wtd_core_c0 / tsc);
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_Any_c0))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pkg_any_core_c0 / tsc);
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_GFX_c0))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pkg_any_gfxe_c0 / tsc);
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_CPUGFX))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pkg_both_core_gfxe_c0 / tsc);
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc2))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc2 / tsc);
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc3))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc3 / tsc);
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc6))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc6 / tsc);
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc7))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc7 / tsc);
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc8))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc8 / tsc);
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc9))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc9 / tsc);
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc10))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc10 / tsc);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
if (DO_BIC(BIC_Diec6))
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
outp +=
sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->die_c6 / crystal_hz / interval_float);
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
tools/power turbostat: Do not print negative LPI residency turbostat prints the abnormal SYS%LPI across suspend-to-idle: SYS%LPI = 114479815993277.50 This is reproduced by: Run a freeze cycle, e.g. "sleepgraph -m freeze -rtcwake 15". Then do a reboot. After boot up, launch the suspend-idle-idle and check the SYS%LPI field. The slp_so residence counter is in LPIT table, and BIOS does not clears this register across reset. The PMC expects the OS to calculate the LPI residency based on the delta. However, there is an firmware issue that the LPIT gets cleared to 0 during the second suspend to idle after the reboot, which brings negative delta value. [lenb: updated to print "neg" upon this BIOS failure] Reported-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2023-10-22 13:52:21 +08:00
if (DO_BIC(BIC_CPU_LPI)) {
if (p->cpu_lpi >= 0)
outp +=
sprintf(outp, "%s%.2f", (printed++ ? delim : ""),
100.0 * p->cpu_lpi / 1000000.0 / interval_float);
else
outp += sprintf(outp, "%s(neg)", (printed++ ? delim : ""));
}
if (DO_BIC(BIC_SYS_LPI)) {
if (p->sys_lpi >= 0)
outp +=
sprintf(outp, "%s%.2f", (printed++ ? delim : ""),
100.0 * p->sys_lpi / 1000000.0 / interval_float);
else
outp += sprintf(outp, "%s(neg)", (printed++ ? delim : ""));
}
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_PkgWatt))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp +=
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
sprintf(outp, fmt8, (printed++ ? delim : ""),
rapl_counter_get_value(&p->energy_pkg, RAPL_UNIT_WATTS, interval_float));
tools/power/turbostat: Abstract Per Core RAPL support Abstract the support for Per Core RAPL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:37:46 +08:00
if (DO_BIC(BIC_CorWatt) && !platform->has_per_core_rapl)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp +=
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
sprintf(outp, fmt8, (printed++ ? delim : ""),
rapl_counter_get_value(&p->energy_cores, RAPL_UNIT_WATTS, interval_float));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_GFXWatt))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp +=
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
sprintf(outp, fmt8, (printed++ ? delim : ""),
rapl_counter_get_value(&p->energy_gfx, RAPL_UNIT_WATTS, interval_float));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_RAMWatt))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp +=
sprintf(outp, fmt8, (printed++ ? delim : ""),
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rapl_counter_get_value(&p->energy_dram, RAPL_UNIT_WATTS, interval_float));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_Pkg_J))
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
outp += sprintf(outp, fmt8, (printed++ ? delim : ""),
rapl_counter_get_value(&p->energy_pkg, RAPL_UNIT_JOULES, interval_float));
tools/power/turbostat: Abstract Per Core RAPL support Abstract the support for Per Core RAPL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:37:46 +08:00
if (DO_BIC(BIC_Cor_J) && !platform->has_per_core_rapl)
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
outp += sprintf(outp, fmt8, (printed++ ? delim : ""),
rapl_counter_get_value(&p->energy_cores, RAPL_UNIT_JOULES, interval_float));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_GFX_J))
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
outp += sprintf(outp, fmt8, (printed++ ? delim : ""),
rapl_counter_get_value(&p->energy_gfx, RAPL_UNIT_JOULES, interval_float));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_RAM_J))
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
outp += sprintf(outp, fmt8, (printed++ ? delim : ""),
rapl_counter_get_value(&p->energy_dram, RAPL_UNIT_JOULES, interval_float));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_PKG__))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp +=
sprintf(outp, fmt8, (printed++ ? delim : ""),
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rapl_counter_get_value(&p->rapl_pkg_perf_status, RAPL_UNIT_WATTS, interval_float));
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_RAM__))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp +=
sprintf(outp, fmt8, (printed++ ? delim : ""),
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rapl_counter_get_value(&p->rapl_dram_perf_status, RAPL_UNIT_WATTS, interval_float));
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
/* UncMHz */
if (DO_BIC(BIC_UNCORE_MHZ))
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), p->uncore_mhz);
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.pp; mp; i++, mp = mp->next) {
if (mp->format == FORMAT_RAW) {
if (mp->width == 32)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp +=
sprintf(outp, "%s0x%08x", (printed++ ? delim : ""), (unsigned int)p->counter[i]);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
else
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s0x%016llx", (printed++ ? delim : ""), p->counter[i]);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
} else if (mp->format == FORMAT_DELTA) {
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
if ((mp->type == COUNTER_ITEMS) && sums_need_wide_columns)
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%8lld", (printed++ ? delim : ""), p->counter[i]);
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
else
tools/power turbostat: update --list feature Make it possible to take the entire un-edited output from `turbostat --list` and feed it to "turbostat --show" or "turbostat --hide". To do this, the leading comma was removed (no mater what columns are active) and also they dynamic C-state "C1, C2, C3" etc are replaced by the string "sysfs", which refers to them as a group. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-16 23:07:51 -05:00
outp += sprintf(outp, "%s%lld", (printed++ ? delim : ""), p->counter[i]);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
} else if (mp->format == FORMAT_PERCENT) {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->counter[i] / tsc);
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
} else if (mp->type == COUNTER_K2M)
outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), (unsigned int)p->counter[i] / 1000);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
for (i = 0, pp = sys.perf_pp; pp; i++, pp = pp->next) {
if (pp->format == FORMAT_RAW) {
if (pp->width == 32)
outp +=
sprintf(outp, "%s0x%08x", (printed++ ? delim : ""),
(unsigned int)p->perf_counter[i]);
else
outp += sprintf(outp, "%s0x%016llx", (printed++ ? delim : ""), p->perf_counter[i]);
} else if (pp->format == FORMAT_DELTA) {
if ((pp->type == COUNTER_ITEMS) && sums_need_wide_columns)
outp += sprintf(outp, "%s%8lld", (printed++ ? delim : ""), p->perf_counter[i]);
else
outp += sprintf(outp, "%s%lld", (printed++ ? delim : ""), p->perf_counter[i]);
} else if (pp->format == FORMAT_PERCENT) {
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->perf_counter[i] / tsc);
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
} else if (pp->type == COUNTER_K2M) {
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
outp +=
sprintf(outp, "%s%d", (printed++ ? delim : ""), (unsigned int)p->perf_counter[i] / 1000);
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
}
}
for (i = 0, ppmt = sys.pmt_pp; ppmt; i++, ppmt = ppmt->next) {
tools/power turbostat: fix GCC9 build regression Fix build regression seen when using old gcc-9 compiler. Signed-off-by: Todd Brandt <todd.e.brandt@intel.com> Reviewed-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-31 12:24:09 -04:00
const unsigned long value_raw = p->pmt_counter[i];
const double value_converted = 100.0 * value_raw / crystal_hz / interval_float;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
switch (ppmt->type) {
case PMT_TYPE_RAW:
if (pmt_counter_get_width(ppmt) <= 32)
outp += sprintf(outp, "%s0x%08x", (printed++ ? delim : ""),
(unsigned int)p->pmt_counter[i]);
else
outp += sprintf(outp, "%s0x%016llx", (printed++ ? delim : ""), p->pmt_counter[i]);
break;
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
case PMT_TYPE_XTAL_TIME:
outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), value_converted);
break;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
}
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
if (DO_BIC(BIC_SysWatt) && (t == &average.threads))
outp += sprintf(outp, fmt8, (printed++ ? delim : ""),
rapl_counter_get_value(&pplat_cnt->energy_psys, RAPL_UNIT_WATTS, interval_float));
if (DO_BIC(BIC_Sys_J) && (t == &average.threads))
outp += sprintf(outp, fmt8, (printed++ ? delim : ""),
rapl_counter_get_value(&pplat_cnt->energy_psys, RAPL_UNIT_JOULES, interval_float));
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
done:
tools/power turbostat: remove blank lines When the user reuests to collect and show columns that are not present on every row (eg. for every CPU) turbostat still prints an (empty) line for every CPU. Update so no blank lines are printed. old: # turbostat --quiet --show Pkg%pc6 Pkg%pc6 9.12 9.12 Pkg%pc6 9.12 9.12 new: # turbostat --quiet --show Pkg%pc6 Pkg%pc6 9.12 9.12 Pkg%pc6 9.12 9.12 Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-01-27 22:39:21 -05:00
if (*(outp - 1) != '\n')
outp += sprintf(outp, "\n");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
void flush_output_stdout(void)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
{
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
FILE *filep;
if (outf == stderr)
filep = stdout;
else
filep = outf;
fputs(output_buffer, filep);
fflush(filep);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
outp = output_buffer;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
void flush_output_stderr(void)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
{
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fputs(output_buffer, outf);
fflush(outf);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
outp = output_buffer;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
void format_all_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
tools/power turbostat: Allow printing header every N iterations This gives the ability to reprint the header every N iterations, so you can ensure that a scrolling display always has the header visible somewhere on the screen. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:39 -07:00
static int count;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: Allow printing header every N iterations This gives the ability to reprint the header every N iterations, so you can ensure that a scrolling display always has the header visible somewhere on the screen. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:39 -07:00
if ((!count || (header_iterations && !(count % header_iterations))) || !summary_only)
tools/power turbostat: Add --list option to show available header names It is handy to know the list of column header names, so that they can be used with --add and --skip The new --list option shows them: sudo ./turbostat --list --hide sysfs ,Core,CPU,Avg_MHz,Busy%,Bzy_MHz,TSC_MHz,IRQ,SMI,CPU%c1,CPU%c3,CPU%c6,CPU%c7,CoreTmp,PkgTmp,GFX%rc6,GFXMHz,PkgWatt,CorWatt,GFXWatt Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 17:15:11 -05:00
print_header("\t");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: fix -S on UP systems The -S (system summary) option failed to print any data on a 1-processor system. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-07-20 14:47:03 -04:00
format_counters(&average.threads, &average.cores, &average.packages);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: Allow printing header every N iterations This gives the ability to reprint the header every N iterations, so you can ensure that a scrolling display always has the header visible somewhere on the screen. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:39 -07:00
count++;
tools turbostat: add summary option turbostat -s cuts down on the amount of output, per user request. also treak some output whitespace and the man page. Signed-off-by: Len Brown <len.brown@intel.com>
2012-02-06 18:37:16 -05:00
if (summary_only)
return;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
for_all_cpus(format_counters, t, c, p);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
#define DELTA_WRAP32(new, old) \
tools/power turbostat: Make the energy variable to be 64 bit Change the energy variable from 32bit to 64bit, so that it can record long time duration. After this conversion, adjust the DELTA_WRAP32() accordingly. Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:47 +08:00
old = ((((unsigned long long)new << 32) - ((unsigned long long)old << 32)) >> 32);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int delta_package(struct pkg_data *new, struct pkg_data *old)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
{
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
int i;
struct msr_counter *mp;
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
struct perf_counter_info *pp;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
struct pmt_counter *ppmt;
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_Totl_c0))
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
old->pkg_wtd_core_c0 = new->pkg_wtd_core_c0 - old->pkg_wtd_core_c0;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_Any_c0))
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
old->pkg_any_core_c0 = new->pkg_any_core_c0 - old->pkg_any_core_c0;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_GFX_c0))
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
old->pkg_any_gfxe_c0 = new->pkg_any_gfxe_c0 - old->pkg_any_gfxe_c0;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_CPUGFX))
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
old->pkg_both_core_gfxe_c0 = new->pkg_both_core_gfxe_c0 - old->pkg_both_core_gfxe_c0;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
old->pc2 = new->pc2 - old->pc2;
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc3))
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-09 23:39:45 -05:00
old->pc3 = new->pc3 - old->pc3;
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc6))
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-09 23:39:45 -05:00
old->pc6 = new->pc6 - old->pc6;
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc7))
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-09 23:39:45 -05:00
old->pc7 = new->pc7 - old->pc7;
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 05:22:43 -08:00
old->pc8 = new->pc8 - old->pc8;
old->pc9 = new->pc9 - old->pc9;
old->pc10 = new->pc10 - old->pc10;
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
old->die_c6 = new->die_c6 - old->die_c6;
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
old->cpu_lpi = new->cpu_lpi - old->cpu_lpi;
old->sys_lpi = new->sys_lpi - old->sys_lpi;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
old->pkg_temp_c = new->pkg_temp_c;
tools/power turbostat: work around RC6 counter wrap Sometimes the rc6 sysfs counter spontaneously resets, causing turbostat prints a very large number as it tries to calcuate % = 100 * (old - new) / interval When we see (old > new), print ***.**% instead of a bogus huge number. Note that this detection is not fool-proof, as the counter could reset several times and still result in new > old. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-06 17:16:00 -04:00
/* flag an error when rc6 counter resets/wraps */
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
if (old->gfx_rc6_ms > new->gfx_rc6_ms)
tools/power turbostat: work around RC6 counter wrap Sometimes the rc6 sysfs counter spontaneously resets, causing turbostat prints a very large number as it tries to calcuate % = 100 * (old - new) / interval When we see (old > new), print ***.**% instead of a bogus huge number. Note that this detection is not fool-proof, as the counter could reset several times and still result in new > old. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-06 17:16:00 -04:00
old->gfx_rc6_ms = -1;
else
old->gfx_rc6_ms = new->gfx_rc6_ms - old->gfx_rc6_ms;
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
old->uncore_mhz = new->uncore_mhz;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 00:37:54 -05:00
old->gfx_mhz = new->gfx_mhz;
tools/power turbostat: Add a new GFXAMHz column that exposes gt_act_freq_mhz. The column already present called GFXMHz reads from gt_cur_freq_mhz, which represents the GT frequency that was requested, but power management might not be able to do that. So the new column will display what the actual frequency GT is running at. Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-22 15:02:07 -07:00
old->gfx_act_mhz = new->gfx_act_mhz;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 00:37:54 -05:00
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
/* flag an error when mc6 counter resets/wraps */
if (old->sam_mc6_ms > new->sam_mc6_ms)
old->sam_mc6_ms = -1;
else
old->sam_mc6_ms = new->sam_mc6_ms - old->sam_mc6_ms;
old->sam_mhz = new->sam_mhz;
old->sam_act_mhz = new->sam_act_mhz;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
old->energy_pkg.raw_value = new->energy_pkg.raw_value - old->energy_pkg.raw_value;
old->energy_cores.raw_value = new->energy_cores.raw_value - old->energy_cores.raw_value;
old->energy_gfx.raw_value = new->energy_gfx.raw_value - old->energy_gfx.raw_value;
old->energy_dram.raw_value = new->energy_dram.raw_value - old->energy_dram.raw_value;
old->rapl_pkg_perf_status.raw_value = new->rapl_pkg_perf_status.raw_value - old->rapl_pkg_perf_status.raw_value;
old->rapl_dram_perf_status.raw_value =
new->rapl_dram_perf_status.raw_value - old->rapl_dram_perf_status.raw_value;
tools/power turbostat: Gracefully handle ACPI S3 turbostat gives valid results across suspend to idle, aka freeze, whether invoked in interval mode, or in command mode. Indeed, this can be used to measure suspend to idle: turbostat echo freeze > /sys/power/state But this does not work across suspend to ACPI S3, because the processor counters, including the TSC, are reset on resume. Further, when turbostat detects a problem, it does't forgive the hardware, and interval mode will print *'s from there on out. Instead, upon detecting counters going backwards, simply reset and start over. Interval mode across ACPI S3: (observe TSC going backwards) root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 1 0.06 858 2294 0x0000000000000000 0 0 0.06 847 2294 0x0000002a254b98ac 1 1 0.06 878 2294 0x0000002a254efa3a 2 1 0.07 843 2294 0x0000002a2551df65 3 0 0.05 863 2294 0x0000002a2553fea2 turbostat: re-initialized with num_cpus 4 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 2 0.20 849 2294 0x0000000000000000 0 2 0.26 856 2294 0x0000000449abb60d 1 2 0.20 844 2294 0x0000000449b087ec 2 2 0.21 850 2294 0x0000000449b35d5d 3 1 0.12 839 2294 0x0000000449b5fd5a ^C Command mode across ACPI S3: root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 sleep 10 ./turbostat: Counter reset detected 14.196299 sec Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 20:31:46 -04:00
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.pp; mp; i++, mp = mp->next) {
if (mp->format == FORMAT_RAW)
old->counter[i] = new->counter[i];
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
else if (mp->format == FORMAT_AVERAGE)
old->counter[i] = new->counter[i];
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
else
old->counter[i] = new->counter[i] - old->counter[i];
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
for (i = 0, pp = sys.perf_pp; pp; i++, pp = pp->next) {
if (pp->format == FORMAT_RAW)
old->perf_counter[i] = new->perf_counter[i];
else if (pp->format == FORMAT_AVERAGE)
old->perf_counter[i] = new->perf_counter[i];
else
old->perf_counter[i] = new->perf_counter[i] - old->perf_counter[i];
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
for (i = 0, ppmt = sys.pmt_pp; ppmt; i++, ppmt = ppmt->next) {
if (ppmt->format == FORMAT_RAW)
old->pmt_counter[i] = new->pmt_counter[i];
else
old->pmt_counter[i] = new->pmt_counter[i] - old->pmt_counter[i];
}
tools/power turbostat: Gracefully handle ACPI S3 turbostat gives valid results across suspend to idle, aka freeze, whether invoked in interval mode, or in command mode. Indeed, this can be used to measure suspend to idle: turbostat echo freeze > /sys/power/state But this does not work across suspend to ACPI S3, because the processor counters, including the TSC, are reset on resume. Further, when turbostat detects a problem, it does't forgive the hardware, and interval mode will print *'s from there on out. Instead, upon detecting counters going backwards, simply reset and start over. Interval mode across ACPI S3: (observe TSC going backwards) root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 1 0.06 858 2294 0x0000000000000000 0 0 0.06 847 2294 0x0000002a254b98ac 1 1 0.06 878 2294 0x0000002a254efa3a 2 1 0.07 843 2294 0x0000002a2551df65 3 0 0.05 863 2294 0x0000002a2553fea2 turbostat: re-initialized with num_cpus 4 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 2 0.20 849 2294 0x0000000000000000 0 2 0.26 856 2294 0x0000000449abb60d 1 2 0.20 844 2294 0x0000000449b087ec 2 2 0.21 850 2294 0x0000000449b35d5d 3 1 0.12 839 2294 0x0000000449b5fd5a ^C Command mode across ACPI S3: root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 sleep 10 ./turbostat: Counter reset detected 14.196299 sec Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 20:31:46 -04:00
return 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
void delta_core(struct core_data *new, struct core_data *old)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
int i;
struct msr_counter *mp;
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
struct perf_counter_info *pp;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
struct pmt_counter *ppmt;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
old->c3 = new->c3 - old->c3;
old->c6 = new->c6 - old->c6;
old->c7 = new->c7 - old->c7;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
old->core_temp_c = new->core_temp_c;
tools/power turbostat: Support thermal throttle count print The turbostat data is collected by end user for power evaluationit. However it looks like we are missing enough thermal context there. Already a couple of time we found that power management developer asking something like this: grep -r . /sys/devices/system/cpu/cpu*/thermal_throttle/* Print the per core thermal throttle count so as to get suffificent thermal context. turbostat -i 5 -s Core,CPU,CoreThr Core CPU CoreThr - - 104 0 0 61 0 4 1 1 0 1 5 2 2 104 2 6 3 3 7 3 7 Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-11-12 19:51:59 +08:00
old->core_throt_cnt = new->core_throt_cnt;
tools/power turbostat: Baytrail c-state support The Baytrail SOC, with its Silvermont core, has some unique properties: 1. a hardware CC1 residency counter 2. a module-c6 residency counter 3. a package-c6 counter at traditional package-c7 counter address. The SOC does not support c3, pc3, c7 or pc7 counters. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:27:20 -05:00
old->mc6_us = new->mc6_us - old->mc6_us;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
DELTA_WRAP32(new->core_energy.raw_value, old->core_energy.raw_value);
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) {
if (mp->format == FORMAT_RAW)
old->counter[i] = new->counter[i];
else
old->counter[i] = new->counter[i] - old->counter[i];
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
for (i = 0, pp = sys.perf_cp; pp; i++, pp = pp->next) {
if (pp->format == FORMAT_RAW)
old->perf_counter[i] = new->perf_counter[i];
else
old->perf_counter[i] = new->perf_counter[i] - old->perf_counter[i];
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
for (i = 0, ppmt = sys.pmt_cp; ppmt; i++, ppmt = ppmt->next) {
if (ppmt->format == FORMAT_RAW)
old->pmt_counter[i] = new->pmt_counter[i];
else
old->pmt_counter[i] = new->pmt_counter[i] - old->pmt_counter[i];
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: Fix CPU%C1 display value In some case C1% will be wrong value, when platform doesn't have MSR for C1 residency. For example: Core CPU CPU%c1 - - 100.00 0 0 100.00 0 2 100.00 1 1 100.00 1 3 100.00 But adding Busy% will fix this Core CPU Busy% CPU%c1 - - 99.77 0.23 0 0 99.77 0.23 0 2 99.77 0.23 1 1 99.77 0.23 1 3 99.77 0.23 This issue can be reproduced on most of the recent systems including Broadwell, Skylake and later. This is because if we don't select Busy% or Avg_MHz or Bzy_MHz then mperf value will not be read from MSR, so it will be 0. But this is required for C1% calculation when MSR for C1 residency is not present. Same is true for C3, C6 and C7 column selection. So add another define DO_BIC_READ(), which doesn't depend on user column selection and use for mperf, C3, C6 and C7 related counters. So when there is no platform support for C1 residency counters, we still read these counters, if the CPU has support and user selected display of CPU%c1. Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-08-27 10:57:14 -07:00
int soft_c1_residency_display(int bic)
{
tools/power/turbostat: Abstract MSR_CORE_C1_RES support Abstract the support for MSR_CORE_C1_RES. Delete is_dnv() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:41:30 +08:00
if (!DO_BIC(BIC_CPU_c1) || platform->has_msr_core_c1_res)
tools/power turbostat: Fix CPU%C1 display value In some case C1% will be wrong value, when platform doesn't have MSR for C1 residency. For example: Core CPU CPU%c1 - - 100.00 0 0 100.00 0 2 100.00 1 1 100.00 1 3 100.00 But adding Busy% will fix this Core CPU Busy% CPU%c1 - - 99.77 0.23 0 0 99.77 0.23 0 2 99.77 0.23 1 1 99.77 0.23 1 3 99.77 0.23 This issue can be reproduced on most of the recent systems including Broadwell, Skylake and later. This is because if we don't select Busy% or Avg_MHz or Bzy_MHz then mperf value will not be read from MSR, so it will be 0. But this is required for C1% calculation when MSR for C1 residency is not present. Same is true for C3, C6 and C7 column selection. So add another define DO_BIC_READ(), which doesn't depend on user column selection and use for mperf, C3, C6 and C7 related counters. So when there is no platform support for C1 residency counters, we still read these counters, if the CPU has support and user selected display of CPU%c1. Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-08-27 10:57:14 -07:00
return 0;
return DO_BIC_READ(bic);
}
tools/power: turbostat: fix large c1% issue Under some conditions, c1% was displayed as very large number, much higher than 100%. c1% is not measured, it is derived as "that, which is left over" from other counters. However, the other counters are not collected atomically, and so it is possible for c1% to be calaculagted as a small negative number -- displayed as very large positive. There was a check for mperf vs tsc for this already, but it needed to also include the other counters that are used to calculate c1. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-13 21:31:46 -04:00
/*
* old = new - old
*/
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int delta_thread(struct thread_data *new, struct thread_data *old, struct core_data *core_delta)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
{
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
int i;
struct msr_counter *mp;
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
struct perf_counter_info *pp;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
struct pmt_counter *ppmt;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
/* we run cpuid just the 1st time, copy the results */
if (DO_BIC(BIC_APIC))
new->apic_id = old->apic_id;
if (DO_BIC(BIC_X2APIC))
new->x2apic_id = old->x2apic_id;
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
/*
* the timestamps from start of measurement interval are in "old"
* the timestamp from end of measurement interval are in "new"
* over-write old w/ new so we can print end of interval values
*/
tools/power turbostat: Make interval calculation per thread to reduce jitter Turbostat currently normalizes TSC and other values by dividing by an interval. This interval is the delta between the start of one global (all counters on all CPUs) sampling and the start of another. However, this introduces a lot of jitter into the data. In order to reduce jitter, the interval calculation should be based on timestamps taken per thread and close to the start of the thread's sampling. Define a per thread time value to hold the delta between samples taken on the thread. Use the timestamp taken at the beginning of sampling to calculate the delta. Move the thread's beginning timestamp to after the CPU migration to avoid jitter due to the migration. Use the global time delta for the average time delta. Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-25 17:32:42 +00:00
timersub(&new->tv_begin, &old->tv_begin, &old->tv_delta);
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
old->tv_begin = new->tv_begin;
old->tv_end = new->tv_end;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
old->tsc = new->tsc - old->tsc;
/* check for TSC < 1 Mcycles over interval */
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:18 -07:00
if (old->tsc < (1000 * 1000))
errx(-3, "Insanely slow TSC rate, TSC stops in idle?\n"
"You can disable all c-states by booting with \"idle=poll\"\n"
"or just the deep ones with \"processor.max_cstate=1\"");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
old->c1 = new->c1 - old->c1;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power/turbostat: bugfix "--show IPC" turbostat --show IPC displays "inf" for the IPC column turbostat was missing the explicit dependency of IPC on APERF, and thus neglected to collect APERF when only IPC was requested. typcial use: turbostat --quiet --show CPU,IPC Signed-off-by: Len Brown <len.brown@intel.com>
2023-06-28 09:40:53 -04:00
if (DO_BIC(BIC_Avg_MHz) || DO_BIC(BIC_Busy) || DO_BIC(BIC_Bzy_MHz) || DO_BIC(BIC_IPC)
|| soft_c1_residency_display(BIC_Avg_MHz)) {
tools/power turbostat: relax dependency on APERF_MSR While turbostat is significantly less useful on systems with no APERF_MSR, it seems more friendly to run on such systems and report what we can, rather than refusing to run. Update man page to reflect recent changes. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 01:33:58 -05:00
if ((new->aperf > old->aperf) && (new->mperf > old->mperf)) {
old->aperf = new->aperf - old->aperf;
old->mperf = new->mperf - old->mperf;
} else {
tools/power turbostat: Gracefully handle ACPI S3 turbostat gives valid results across suspend to idle, aka freeze, whether invoked in interval mode, or in command mode. Indeed, this can be used to measure suspend to idle: turbostat echo freeze > /sys/power/state But this does not work across suspend to ACPI S3, because the processor counters, including the TSC, are reset on resume. Further, when turbostat detects a problem, it does't forgive the hardware, and interval mode will print *'s from there on out. Instead, upon detecting counters going backwards, simply reset and start over. Interval mode across ACPI S3: (observe TSC going backwards) root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 1 0.06 858 2294 0x0000000000000000 0 0 0.06 847 2294 0x0000002a254b98ac 1 1 0.06 878 2294 0x0000002a254efa3a 2 1 0.07 843 2294 0x0000002a2551df65 3 0 0.05 863 2294 0x0000002a2553fea2 turbostat: re-initialized with num_cpus 4 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 2 0.20 849 2294 0x0000000000000000 0 2 0.26 856 2294 0x0000000449abb60d 1 2 0.20 844 2294 0x0000000449b087ec 2 2 0.21 850 2294 0x0000000449b35d5d 3 1 0.12 839 2294 0x0000000449b5fd5a ^C Command mode across ACPI S3: root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 sleep 10 ./turbostat: Counter reset detected 14.196299 sec Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 20:31:46 -04:00
return -1;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power/turbostat: Abstract MSR_CORE_C1_RES support Abstract the support for MSR_CORE_C1_RES. Delete is_dnv() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:41:30 +08:00
if (platform->has_msr_core_c1_res) {
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
/*
* Some models have a dedicated C1 residency MSR,
* which should be more accurate than the derivation below.
*/
} else {
/*
* As counter collection is not atomic,
* it is possible for mperf's non-halted cycles + idle states
* to exceed TSC's all cycles: show c1 = 0% in that case.
*/
tools/power turbostat: fix impossibly large CPU%c1 value Most CPUs do not have a hardware c1 counter, and so turbostat derives c1 residency: c1 = TSC - MPERF - other_core_cstate_counters As it is not possible to atomically read these coutners, measurement jitter can case this calcuation to "go negative" when very close to 0. Turbostat detect that case and simply prints c1 = 0.00% But that check neglected to account for systems where the TSC crystal clock domain and the MPERF BCLK domain are differ by a small amount. That allowed very small negative c1 numbers to escape this check and be printed as huge positve numbers. This code begs for a bit of cleanup, but this patch is the minimal change to fix the issue. Signed-off-by: Len Brown <len.brown@intel.com>
2017-04-12 19:44:51 -04:00
if ((old->mperf + core_delta->c3 + core_delta->c6 + core_delta->c7) > (old->tsc * tsc_tweak))
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
old->c1 = 0;
else {
/* normal case, derive c1 */
tools/power turbostat: use tsc_tweak everwhere it is needed The CPU ticks at a rate in the "bus clock" domain. eg. 100 MHz * bus_ratio. On newer processors, the TSC has been moved out of this BCLK domain and into a separate crystal-clock domain. While the TSC ticks "close to" the base frequency, those that look closely at the numbers will notice small errors in calculations that mix units of TSC clocks and bus clocks. "tsc_tweak" was introduced to address the most visible mixing -- the %Busy and the the Busy_MHz calculations. (A simplification as since removed TSC from the BusyMHz calculation) Here we apply the tsc_tweak to everyplace where BCLK and TSC units are mixed. The results is that on a system which is 100% idle, the sum of the C-states are now much more likely to be closer to 100%. Reported-by: Travis Downs <travis.downs@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:29:51 -05:00
old->c1 = (old->tsc * tsc_tweak) - old->mperf - core_delta->c3
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
- core_delta->c6 - core_delta->c7;
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools/power: turbostat: fix large c1% issue Under some conditions, c1% was displayed as very large number, much higher than 100%. c1% is not measured, it is derived as "that, which is left over" from other counters. However, the other counters are not collected atomically, and so it is possible for c1% to be calaculagted as a small negative number -- displayed as very large positive. There was a check for mperf vs tsc for this already, but it needed to also include the other counters that are used to calculate c1. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-13 21:31:46 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
if (old->mperf == 0) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
if (debug > 1)
fprintf(outf, "cpu%d MPERF 0!\n", old->cpu_id);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
old->mperf = 1; /* divide by 0 protection */
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
if (DO_BIC(BIC_IPC))
old->instr_count = new->instr_count - old->instr_count;
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_IRQ))
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
old->irq_count = new->irq_count - old->irq_count;
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_SMI))
tools/power turbostat: display SMI count by default The SMI counter is popular -- so display it by default rather than requiring an option. What the heck, we've blown the 80 column budget on many systems already... Note that the value displayed is the delta during the measurement interval. The absolute value of the counter can still be seen with the generic 32-bit MSR option, ie. -m 0x34 Signed-off-by: Len Brown <len.brown@intel.com>
2013-02-10 17:19:24 -05:00
old->smi_count = new->smi_count - old->smi_count;
tools/power turbostat: Gracefully handle ACPI S3 turbostat gives valid results across suspend to idle, aka freeze, whether invoked in interval mode, or in command mode. Indeed, this can be used to measure suspend to idle: turbostat echo freeze > /sys/power/state But this does not work across suspend to ACPI S3, because the processor counters, including the TSC, are reset on resume. Further, when turbostat detects a problem, it does't forgive the hardware, and interval mode will print *'s from there on out. Instead, upon detecting counters going backwards, simply reset and start over. Interval mode across ACPI S3: (observe TSC going backwards) root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 1 0.06 858 2294 0x0000000000000000 0 0 0.06 847 2294 0x0000002a254b98ac 1 1 0.06 878 2294 0x0000002a254efa3a 2 1 0.07 843 2294 0x0000002a2551df65 3 0 0.05 863 2294 0x0000002a2553fea2 turbostat: re-initialized with num_cpus 4 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 2 0.20 849 2294 0x0000000000000000 0 2 0.26 856 2294 0x0000000449abb60d 1 2 0.20 844 2294 0x0000000449b087ec 2 2 0.21 850 2294 0x0000000449b35d5d 3 1 0.12 839 2294 0x0000000449b5fd5a ^C Command mode across ACPI S3: root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 sleep 10 ./turbostat: Counter reset detected 14.196299 sec Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 20:31:46 -04:00
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.tp; mp; i++, mp = mp->next) {
if (mp->format == FORMAT_RAW)
old->counter[i] = new->counter[i];
else
old->counter[i] = new->counter[i] - old->counter[i];
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
for (i = 0, pp = sys.perf_tp; pp; i++, pp = pp->next) {
if (pp->format == FORMAT_RAW)
old->perf_counter[i] = new->perf_counter[i];
else
old->perf_counter[i] = new->perf_counter[i] - old->perf_counter[i];
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
for (i = 0, ppmt = sys.pmt_tp; ppmt; i++, ppmt = ppmt->next) {
if (ppmt->format == FORMAT_RAW)
old->pmt_counter[i] = new->pmt_counter[i];
else
old->pmt_counter[i] = new->pmt_counter[i] - old->pmt_counter[i];
}
tools/power turbostat: Gracefully handle ACPI S3 turbostat gives valid results across suspend to idle, aka freeze, whether invoked in interval mode, or in command mode. Indeed, this can be used to measure suspend to idle: turbostat echo freeze > /sys/power/state But this does not work across suspend to ACPI S3, because the processor counters, including the TSC, are reset on resume. Further, when turbostat detects a problem, it does't forgive the hardware, and interval mode will print *'s from there on out. Instead, upon detecting counters going backwards, simply reset and start over. Interval mode across ACPI S3: (observe TSC going backwards) root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 1 0.06 858 2294 0x0000000000000000 0 0 0.06 847 2294 0x0000002a254b98ac 1 1 0.06 878 2294 0x0000002a254efa3a 2 1 0.07 843 2294 0x0000002a2551df65 3 0 0.05 863 2294 0x0000002a2553fea2 turbostat: re-initialized with num_cpus 4 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 2 0.20 849 2294 0x0000000000000000 0 2 0.26 856 2294 0x0000000449abb60d 1 2 0.20 844 2294 0x0000000449b087ec 2 2 0.21 850 2294 0x0000000449b35d5d 3 1 0.12 839 2294 0x0000000449b5fd5a ^C Command mode across ACPI S3: root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 sleep 10 ./turbostat: Counter reset detected 14.196299 sec Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 20:31:46 -04:00
return 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
int delta_cpu(struct thread_data *t, struct core_data *c,
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
struct pkg_data *p, struct thread_data *t2, struct core_data *c2, struct pkg_data *p2)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
{
tools/power turbostat: Gracefully handle ACPI S3 turbostat gives valid results across suspend to idle, aka freeze, whether invoked in interval mode, or in command mode. Indeed, this can be used to measure suspend to idle: turbostat echo freeze > /sys/power/state But this does not work across suspend to ACPI S3, because the processor counters, including the TSC, are reset on resume. Further, when turbostat detects a problem, it does't forgive the hardware, and interval mode will print *'s from there on out. Instead, upon detecting counters going backwards, simply reset and start over. Interval mode across ACPI S3: (observe TSC going backwards) root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 1 0.06 858 2294 0x0000000000000000 0 0 0.06 847 2294 0x0000002a254b98ac 1 1 0.06 878 2294 0x0000002a254efa3a 2 1 0.07 843 2294 0x0000002a2551df65 3 0 0.05 863 2294 0x0000002a2553fea2 turbostat: re-initialized with num_cpus 4 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 2 0.20 849 2294 0x0000000000000000 0 2 0.26 856 2294 0x0000000449abb60d 1 2 0.20 844 2294 0x0000000449b087ec 2 2 0.21 850 2294 0x0000000449b35d5d 3 1 0.12 839 2294 0x0000000449b5fd5a ^C Command mode across ACPI S3: root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 sleep 10 ./turbostat: Counter reset detected 14.196299 sec Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 20:31:46 -04:00
int retval = 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/* calculate core delta only for 1st thread in core */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (is_cpu_first_thread_in_core(t, c, p))
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
delta_core(c, c2);
/* always calculate thread delta */
tools/power turbostat: Gracefully handle ACPI S3 turbostat gives valid results across suspend to idle, aka freeze, whether invoked in interval mode, or in command mode. Indeed, this can be used to measure suspend to idle: turbostat echo freeze > /sys/power/state But this does not work across suspend to ACPI S3, because the processor counters, including the TSC, are reset on resume. Further, when turbostat detects a problem, it does't forgive the hardware, and interval mode will print *'s from there on out. Instead, upon detecting counters going backwards, simply reset and start over. Interval mode across ACPI S3: (observe TSC going backwards) root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 1 0.06 858 2294 0x0000000000000000 0 0 0.06 847 2294 0x0000002a254b98ac 1 1 0.06 878 2294 0x0000002a254efa3a 2 1 0.07 843 2294 0x0000002a2551df65 3 0 0.05 863 2294 0x0000002a2553fea2 turbostat: re-initialized with num_cpus 4 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 2 0.20 849 2294 0x0000000000000000 0 2 0.26 856 2294 0x0000000449abb60d 1 2 0.20 844 2294 0x0000000449b087ec 2 2 0.21 850 2294 0x0000000449b35d5d 3 1 0.12 839 2294 0x0000000449b5fd5a ^C Command mode across ACPI S3: root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 sleep 10 ./turbostat: Counter reset detected 14.196299 sec Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 20:31:46 -04:00
retval = delta_thread(t, t2, c2); /* c2 is core delta */
if (retval)
return retval;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/* calculate package delta only for 1st core in package */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (is_cpu_first_core_in_package(t, c, p))
tools/power turbostat: Gracefully handle ACPI S3 turbostat gives valid results across suspend to idle, aka freeze, whether invoked in interval mode, or in command mode. Indeed, this can be used to measure suspend to idle: turbostat echo freeze > /sys/power/state But this does not work across suspend to ACPI S3, because the processor counters, including the TSC, are reset on resume. Further, when turbostat detects a problem, it does't forgive the hardware, and interval mode will print *'s from there on out. Instead, upon detecting counters going backwards, simply reset and start over. Interval mode across ACPI S3: (observe TSC going backwards) root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 1 0.06 858 2294 0x0000000000000000 0 0 0.06 847 2294 0x0000002a254b98ac 1 1 0.06 878 2294 0x0000002a254efa3a 2 1 0.07 843 2294 0x0000002a2551df65 3 0 0.05 863 2294 0x0000002a2553fea2 turbostat: re-initialized with num_cpus 4 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 2 0.20 849 2294 0x0000000000000000 0 2 0.26 856 2294 0x0000000449abb60d 1 2 0.20 844 2294 0x0000000449b087ec 2 2 0.21 850 2294 0x0000000449b35d5d 3 1 0.12 839 2294 0x0000000449b5fd5a ^C Command mode across ACPI S3: root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 sleep 10 ./turbostat: Counter reset detected 14.196299 sec Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 20:31:46 -04:00
retval = delta_package(p, p2);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: Gracefully handle ACPI S3 turbostat gives valid results across suspend to idle, aka freeze, whether invoked in interval mode, or in command mode. Indeed, this can be used to measure suspend to idle: turbostat echo freeze > /sys/power/state But this does not work across suspend to ACPI S3, because the processor counters, including the TSC, are reset on resume. Further, when turbostat detects a problem, it does't forgive the hardware, and interval mode will print *'s from there on out. Instead, upon detecting counters going backwards, simply reset and start over. Interval mode across ACPI S3: (observe TSC going backwards) root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 1 0.06 858 2294 0x0000000000000000 0 0 0.06 847 2294 0x0000002a254b98ac 1 1 0.06 878 2294 0x0000002a254efa3a 2 1 0.07 843 2294 0x0000002a2551df65 3 0 0.05 863 2294 0x0000002a2553fea2 turbostat: re-initialized with num_cpus 4 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 2 0.20 849 2294 0x0000000000000000 0 2 0.26 856 2294 0x0000000449abb60d 1 2 0.20 844 2294 0x0000000449b087ec 2 2 0.21 850 2294 0x0000000449b35d5d 3 1 0.12 839 2294 0x0000000449b5fd5a ^C Command mode across ACPI S3: root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 sleep 10 ./turbostat: Counter reset detected 14.196299 sec Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 20:31:46 -04:00
return retval;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
void delta_platform(struct platform_counters *new, struct platform_counters *old)
{
old->energy_psys.raw_value = new->energy_psys.raw_value - old->energy_psys.raw_value;
}
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
void rapl_counter_clear(struct rapl_counter *c)
{
c->raw_value = 0;
c->scale = 0.0;
c->unit = RAPL_UNIT_INVALID;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
void clear_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
int i;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
struct msr_counter *mp;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
t->tv_begin.tv_sec = 0;
t->tv_begin.tv_usec = 0;
t->tv_end.tv_sec = 0;
t->tv_end.tv_usec = 0;
tools/power turbostat: Make interval calculation per thread to reduce jitter Turbostat currently normalizes TSC and other values by dividing by an interval. This interval is the delta between the start of one global (all counters on all CPUs) sampling and the start of another. However, this introduces a lot of jitter into the data. In order to reduce jitter, the interval calculation should be based on timestamps taken per thread and close to the start of the thread's sampling. Define a per thread time value to hold the delta between samples taken on the thread. Use the timestamp taken at the beginning of sampling to calculate the delta. Move the thread's beginning timestamp to after the CPU migration to avoid jitter due to the migration. Use the global time delta for the average time delta. Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-25 17:32:42 +00:00
t->tv_delta.tv_sec = 0;
t->tv_delta.tv_usec = 0;
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
t->tsc = 0;
t->aperf = 0;
t->mperf = 0;
t->c1 = 0;
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
t->instr_count = 0;
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
t->irq_count = 0;
t->smi_count = 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
c->c3 = 0;
c->c6 = 0;
c->c7 = 0;
tools/power turbostat: Baytrail c-state support The Baytrail SOC, with its Silvermont core, has some unique properties: 1. a hardware CC1 residency counter 2. a module-c6 residency counter 3. a package-c6 counter at traditional package-c7 counter address. The SOC does not support c3, pc3, c7 or pc7 counters. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:27:20 -05:00
c->mc6_us = 0;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
c->core_temp_c = 0;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rapl_counter_clear(&c->core_energy);
tools/power turbostat: Support thermal throttle count print The turbostat data is collected by end user for power evaluationit. However it looks like we are missing enough thermal context there. Already a couple of time we found that power management developer asking something like this: grep -r . /sys/devices/system/cpu/cpu*/thermal_throttle/* Print the per core thermal throttle count so as to get suffificent thermal context. turbostat -i 5 -s Core,CPU,CoreThr Core CPU CoreThr - - 104 0 0 61 0 4 1 1 0 1 5 2 2 104 2 6 3 3 7 3 7 Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-11-12 19:51:59 +08:00
c->core_throt_cnt = 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
p->pkg_wtd_core_c0 = 0;
p->pkg_any_core_c0 = 0;
p->pkg_any_gfxe_c0 = 0;
p->pkg_both_core_gfxe_c0 = 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
p->pc2 = 0;
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc3))
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-09 23:39:45 -05:00
p->pc3 = 0;
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc6))
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-09 23:39:45 -05:00
p->pc6 = 0;
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc7))
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-09 23:39:45 -05:00
p->pc7 = 0;
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 05:22:43 -08:00
p->pc8 = 0;
p->pc9 = 0;
p->pc10 = 0;
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
p->die_c6 = 0;
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
p->cpu_lpi = 0;
p->sys_lpi = 0;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rapl_counter_clear(&p->energy_pkg);
rapl_counter_clear(&p->energy_dram);
rapl_counter_clear(&p->energy_cores);
rapl_counter_clear(&p->energy_gfx);
rapl_counter_clear(&p->rapl_pkg_perf_status);
rapl_counter_clear(&p->rapl_dram_perf_status);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
p->pkg_temp_c = 0;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 00:37:54 -05:00
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 01:28:12 -05:00
p->gfx_rc6_ms = 0;
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
p->uncore_mhz = 0;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 00:37:54 -05:00
p->gfx_mhz = 0;
tools/power turbostat: Add a new GFXAMHz column that exposes gt_act_freq_mhz. The column already present called GFXMHz reads from gt_cur_freq_mhz, which represents the GT frequency that was requested, but power management might not be able to do that. So the new column will display what the actual frequency GT is running at. Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-22 15:02:07 -07:00
p->gfx_act_mhz = 0;
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
p->sam_mc6_ms = 0;
p->sam_mhz = 0;
p->sam_act_mhz = 0;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.tp; mp; i++, mp = mp->next)
t->counter[i] = 0;
for (i = 0, mp = sys.cp; mp; i++, mp = mp->next)
c->counter[i] = 0;
for (i = 0, mp = sys.pp; mp; i++, mp = mp->next)
p->counter[i] = 0;
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
memset(&t->perf_counter[0], 0, sizeof(t->perf_counter));
memset(&c->perf_counter[0], 0, sizeof(c->perf_counter));
memset(&p->perf_counter[0], 0, sizeof(p->perf_counter));
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
memset(&t->pmt_counter[0], 0, ARRAY_SIZE(t->pmt_counter));
memset(&c->pmt_counter[0], 0, ARRAY_SIZE(c->pmt_counter));
memset(&p->pmt_counter[0], 0, ARRAY_SIZE(p->pmt_counter));
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
void rapl_counter_accumulate(struct rapl_counter *dst, const struct rapl_counter *src)
{
/* Copy unit and scale from src if dst is not initialized */
if (dst->unit == RAPL_UNIT_INVALID) {
dst->unit = src->unit;
dst->scale = src->scale;
}
assert(dst->unit == src->unit);
assert(dst->scale == src->scale);
dst->raw_value += src->raw_value;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int sum_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
int i;
struct msr_counter *mp;
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
struct perf_counter_info *pp;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
struct pmt_counter *ppmt;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
/* copy un-changing apic_id's */
if (DO_BIC(BIC_APIC))
average.threads.apic_id = t->apic_id;
if (DO_BIC(BIC_X2APIC))
average.threads.x2apic_id = t->x2apic_id;
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
/* remember first tv_begin */
if (average.threads.tv_begin.tv_sec == 0)
average.threads.tv_begin = t->tv_begin;
/* remember last tv_end */
average.threads.tv_end = t->tv_end;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
average.threads.tsc += t->tsc;
average.threads.aperf += t->aperf;
average.threads.mperf += t->mperf;
average.threads.c1 += t->c1;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
average.threads.instr_count += t->instr_count;
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
average.threads.irq_count += t->irq_count;
average.threads.smi_count += t->smi_count;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.tp; mp; i++, mp = mp->next) {
if (mp->format == FORMAT_RAW)
continue;
average.threads.counter[i] += t->counter[i];
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
for (i = 0, pp = sys.perf_tp; pp; i++, pp = pp->next) {
if (pp->format == FORMAT_RAW)
continue;
average.threads.perf_counter[i] += t->perf_counter[i];
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
for (i = 0, ppmt = sys.pmt_tp; ppmt; i++, ppmt = ppmt->next) {
average.threads.pmt_counter[i] += t->pmt_counter[i];
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/* sum per-core values only for 1st thread in core */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (!is_cpu_first_thread_in_core(t, c, p))
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
average.cores.c3 += c->c3;
average.cores.c6 += c->c6;
average.cores.c7 += c->c7;
tools/power turbostat: Baytrail c-state support The Baytrail SOC, with its Silvermont core, has some unique properties: 1. a hardware CC1 residency counter 2. a module-c6 residency counter 3. a package-c6 counter at traditional package-c7 counter address. The SOC does not support c3, pc3, c7 or pc7 counters. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:27:20 -05:00
average.cores.mc6_us += c->mc6_us;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
average.cores.core_temp_c = MAX(average.cores.core_temp_c, c->core_temp_c);
tools/power turbostat: Support thermal throttle count print The turbostat data is collected by end user for power evaluationit. However it looks like we are missing enough thermal context there. Already a couple of time we found that power management developer asking something like this: grep -r . /sys/devices/system/cpu/cpu*/thermal_throttle/* Print the per core thermal throttle count so as to get suffificent thermal context. turbostat -i 5 -s Core,CPU,CoreThr Core CPU CoreThr - - 104 0 0 61 0 4 1 1 0 1 5 2 2 104 2 6 3 3 7 3 7 Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-11-12 19:51:59 +08:00
average.cores.core_throt_cnt = MAX(average.cores.core_throt_cnt, c->core_throt_cnt);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rapl_counter_accumulate(&average.cores.core_energy, &c->core_energy);
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) {
if (mp->format == FORMAT_RAW)
continue;
average.cores.counter[i] += c->counter[i];
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
for (i = 0, pp = sys.perf_cp; pp; i++, pp = pp->next) {
if (pp->format == FORMAT_RAW)
continue;
average.cores.perf_counter[i] += c->perf_counter[i];
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
for (i = 0, ppmt = sys.pmt_cp; ppmt; i++, ppmt = ppmt->next) {
average.cores.pmt_counter[i] += c->pmt_counter[i];
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/* sum per-pkg values only for 1st core in pkg */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (!is_cpu_first_core_in_package(t, c, p))
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return 0;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_Totl_c0))
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
average.packages.pkg_wtd_core_c0 += p->pkg_wtd_core_c0;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_Any_c0))
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
average.packages.pkg_any_core_c0 += p->pkg_any_core_c0;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_GFX_c0))
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
average.packages.pkg_any_gfxe_c0 += p->pkg_any_gfxe_c0;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_CPUGFX))
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
average.packages.pkg_both_core_gfxe_c0 += p->pkg_both_core_gfxe_c0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
average.packages.pc2 += p->pc2;
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc3))
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-09 23:39:45 -05:00
average.packages.pc3 += p->pc3;
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc6))
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-09 23:39:45 -05:00
average.packages.pc6 += p->pc6;
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc7))
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-09 23:39:45 -05:00
average.packages.pc7 += p->pc7;
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 05:22:43 -08:00
average.packages.pc8 += p->pc8;
average.packages.pc9 += p->pc9;
average.packages.pc10 += p->pc10;
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
average.packages.die_c6 += p->die_c6;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
average.packages.cpu_lpi = p->cpu_lpi;
average.packages.sys_lpi = p->sys_lpi;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rapl_counter_accumulate(&average.packages.energy_pkg, &p->energy_pkg);
rapl_counter_accumulate(&average.packages.energy_dram, &p->energy_dram);
rapl_counter_accumulate(&average.packages.energy_cores, &p->energy_cores);
rapl_counter_accumulate(&average.packages.energy_gfx, &p->energy_gfx);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 01:28:12 -05:00
average.packages.gfx_rc6_ms = p->gfx_rc6_ms;
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
average.packages.uncore_mhz = p->uncore_mhz;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 00:37:54 -05:00
average.packages.gfx_mhz = p->gfx_mhz;
tools/power turbostat: Add a new GFXAMHz column that exposes gt_act_freq_mhz. The column already present called GFXMHz reads from gt_cur_freq_mhz, which represents the GT frequency that was requested, but power management might not be able to do that. So the new column will display what the actual frequency GT is running at. Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-22 15:02:07 -07:00
average.packages.gfx_act_mhz = p->gfx_act_mhz;
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
average.packages.sam_mc6_ms = p->sam_mc6_ms;
average.packages.sam_mhz = p->sam_mhz;
average.packages.sam_act_mhz = p->sam_act_mhz;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 00:37:54 -05:00
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
average.packages.pkg_temp_c = MAX(average.packages.pkg_temp_c, p->pkg_temp_c);
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rapl_counter_accumulate(&average.packages.rapl_pkg_perf_status, &p->rapl_pkg_perf_status);
rapl_counter_accumulate(&average.packages.rapl_dram_perf_status, &p->rapl_dram_perf_status);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.pp; mp; i++, mp = mp->next) {
tools/power turbostat: Fix added raw MSR output When using --Summary mode, added MSRs in raw mode always print zeros. Print the actual register contents. Example, with patch: note the added column: --add msr0x64f,u32,package,raw,REASON Where: 0x64F is MSR_CORE_PERF_LIMIT_REASONS Busy% Bzy_MHz PkgTmp PkgWatt CorWatt REASON 0.00 4800 35 1.42 0.76 0x00000000 0.00 4801 34 1.42 0.76 0x00000000 80.08 4531 66 108.17 107.52 0x08000000 98.69 4530 66 133.21 132.54 0x08000000 99.28 4505 66 128.26 127.60 0x0c000400 99.65 4486 68 124.91 124.25 0x0c000400 99.63 4483 68 124.90 124.25 0x0c000400 79.34 4481 41 99.80 99.13 0x0c000000 0.00 4801 41 1.40 0.73 0x0c000000 Where, for the test processor (i5-10600K): PKG Limit #1: 125.000 Watts, 8.000000 sec MSR bit 26 = log; bit 10 = status PKG Limit #2: 136.000 Watts, 0.002441 sec MSR bit 27 = log; bit 11 = status Example, without patch: Busy% Bzy_MHz PkgTmp PkgWatt CorWatt REASON 0.01 4800 35 1.43 0.77 0x00000000 0.00 4801 35 1.39 0.73 0x00000000 83.49 4531 66 112.71 112.06 0x00000000 98.69 4530 68 133.35 132.69 0x00000000 99.31 4500 67 127.96 127.30 0x00000000 99.63 4483 69 124.91 124.25 0x00000000 99.61 4481 69 124.90 124.25 0x00000000 99.61 4481 71 124.92 124.25 0x00000000 59.35 4479 42 75.03 74.37 0x00000000 0.00 4800 42 1.39 0.73 0x00000000 0.00 4801 42 1.42 0.76 0x00000000 c000000 [lenb: simplified patch to apply only to package scope] Signed-off-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Len Brown <len.brown@intel.com>
2023-04-03 14:11:38 -07:00
if ((mp->format == FORMAT_RAW) && (topo.num_packages == 0))
average.packages.counter[i] = p->counter[i];
else
average.packages.counter[i] += p->counter[i];
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
for (i = 0, pp = sys.perf_pp; pp; i++, pp = pp->next) {
if ((pp->format == FORMAT_RAW) && (topo.num_packages == 0))
average.packages.perf_counter[i] = p->perf_counter[i];
else
average.packages.perf_counter[i] += p->perf_counter[i];
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
for (i = 0, ppmt = sys.pmt_pp; ppmt; i++, ppmt = ppmt->next) {
average.packages.pmt_counter[i] += p->pmt_counter[i];
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return 0;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/*
* sum the counters for all cpus in the system
* compute the weighted average
*/
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
void compute_average(struct thread_data *t, struct core_data *c, struct pkg_data *p)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
{
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
int i;
struct msr_counter *mp;
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
struct perf_counter_info *pp;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
struct pmt_counter *ppmt;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
clear_counters(&average.threads, &average.cores, &average.packages);
for_all_cpus(sum_counters, t, c, p);
tools/power turbostat: Make interval calculation per thread to reduce jitter Turbostat currently normalizes TSC and other values by dividing by an interval. This interval is the delta between the start of one global (all counters on all CPUs) sampling and the start of another. However, this introduces a lot of jitter into the data. In order to reduce jitter, the interval calculation should be based on timestamps taken per thread and close to the start of the thread's sampling. Define a per thread time value to hold the delta between samples taken on the thread. Use the timestamp taken at the beginning of sampling to calculate the delta. Move the thread's beginning timestamp to after the CPU migration to avoid jitter due to the migration. Use the global time delta for the average time delta. Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-25 17:32:42 +00:00
/* Use the global time delta for the average. */
average.threads.tv_delta = tv_delta;
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.threads.tsc /= topo.allowed_cpus;
average.threads.aperf /= topo.allowed_cpus;
average.threads.mperf /= topo.allowed_cpus;
average.threads.instr_count /= topo.allowed_cpus;
average.threads.c1 /= topo.allowed_cpus;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
if (average.threads.irq_count > 9999999)
sums_need_wide_columns = 1;
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.cores.c3 /= topo.allowed_cores;
average.cores.c6 /= topo.allowed_cores;
average.cores.c7 /= topo.allowed_cores;
average.cores.mc6_us /= topo.allowed_cores;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_Totl_c0))
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.packages.pkg_wtd_core_c0 /= topo.allowed_packages;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_Any_c0))
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.packages.pkg_any_core_c0 /= topo.allowed_packages;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_GFX_c0))
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.packages.pkg_any_gfxe_c0 /= topo.allowed_packages;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_CPUGFX))
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.packages.pkg_both_core_gfxe_c0 /= topo.allowed_packages;
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.packages.pc2 /= topo.allowed_packages;
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc3))
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.packages.pc3 /= topo.allowed_packages;
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc6))
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.packages.pc6 /= topo.allowed_packages;
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
if (DO_BIC(BIC_Pkgpc7))
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.packages.pc7 /= topo.allowed_packages;
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 05:22:43 -08:00
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.packages.pc8 /= topo.allowed_packages;
average.packages.pc9 /= topo.allowed_packages;
average.packages.pc10 /= topo.allowed_packages;
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
average.packages.die_c6 /= topo.allowed_packages;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.tp; mp; i++, mp = mp->next) {
if (mp->format == FORMAT_RAW)
continue;
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
if (mp->type == COUNTER_ITEMS) {
if (average.threads.counter[i] > 9999999)
sums_need_wide_columns = 1;
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
continue;
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
}
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.threads.counter[i] /= topo.allowed_cpus;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) {
if (mp->format == FORMAT_RAW)
continue;
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
if (mp->type == COUNTER_ITEMS) {
if (average.cores.counter[i] > 9999999)
sums_need_wide_columns = 1;
}
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.cores.counter[i] /= topo.allowed_cores;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
for (i = 0, mp = sys.pp; mp; i++, mp = mp->next) {
if (mp->format == FORMAT_RAW)
continue;
tools/power turbostat: use wide columns to display large numbers When a counter overlfows 7 columns, it shifts the remaining columns to the right, so they no longer line up under their column header. Update turbostat to dectect when it is handling large numbers, and switch to wider columns where, necessary. Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 21:45:40 -05:00
if (mp->type == COUNTER_ITEMS) {
if (average.packages.counter[i] > 9999999)
sums_need_wide_columns = 1;
}
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
average.packages.counter[i] /= topo.allowed_packages;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
for (i = 0, pp = sys.perf_tp; pp; i++, pp = pp->next) {
if (pp->format == FORMAT_RAW)
continue;
if (pp->type == COUNTER_ITEMS) {
if (average.threads.perf_counter[i] > 9999999)
sums_need_wide_columns = 1;
continue;
}
average.threads.perf_counter[i] /= topo.allowed_cpus;
}
for (i = 0, pp = sys.perf_cp; pp; i++, pp = pp->next) {
if (pp->format == FORMAT_RAW)
continue;
if (pp->type == COUNTER_ITEMS) {
if (average.cores.perf_counter[i] > 9999999)
sums_need_wide_columns = 1;
}
average.cores.perf_counter[i] /= topo.allowed_cores;
}
for (i = 0, pp = sys.perf_pp; pp; i++, pp = pp->next) {
if (pp->format == FORMAT_RAW)
continue;
if (pp->type == COUNTER_ITEMS) {
if (average.packages.perf_counter[i] > 9999999)
sums_need_wide_columns = 1;
}
average.packages.perf_counter[i] /= topo.allowed_packages;
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
for (i = 0, ppmt = sys.pmt_tp; ppmt; i++, ppmt = ppmt->next) {
average.threads.pmt_counter[i] /= topo.allowed_cpus;
}
for (i = 0, ppmt = sys.pmt_cp; ppmt; i++, ppmt = ppmt->next) {
average.cores.pmt_counter[i] /= topo.allowed_cores;
}
for (i = 0, ppmt = sys.pmt_pp; ppmt; i++, ppmt = ppmt->next) {
average.packages.pmt_counter[i] /= topo.allowed_packages;
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
static unsigned long long rdtsc(void)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
unsigned int low, high;
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 22:19:58 -04:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
asm volatile ("rdtsc":"=a" (low), "=d"(high));
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 22:19:58 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return low | ((unsigned long long)high) << 32;
}
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 22:19:58 -04:00
tools/power turbostat: extend --add option to accept /sys path Previously, the --add option could specify only an MSR. Here is is extended so an arbitrary /sys attribute, as specified by an absolute file path name. sudo ./turbostat --add /sys/devices/system/cpu/cpu0/cpuidle/state5/usage Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-08 02:41:51 -05:00
/*
* Open a file, and exit on failure
*/
FILE *fopen_or_die(const char *path, const char *mode)
{
FILE *filep = fopen(path, mode);
if (!filep)
err(1, "%s: open failed", path);
return filep;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: extend --add option to accept /sys path Previously, the --add option could specify only an MSR. Here is is extended so an arbitrary /sys attribute, as specified by an absolute file path name. sudo ./turbostat --add /sys/devices/system/cpu/cpu0/cpuidle/state5/usage Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-08 02:41:51 -05:00
/*
* snapshot_sysfs_counter()
*
* return snapshot of given counter
*/
unsigned long long snapshot_sysfs_counter(char *path)
{
FILE *fp;
int retval;
unsigned long long counter;
fp = fopen_or_die(path, "r");
retval = fscanf(fp, "%lld", &counter);
if (retval != 1)
err(1, "snapshot_sysfs_counter(%s)", path);
fclose(fp);
return counter;
}
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
int get_mp(int cpu, struct msr_counter *mp, unsigned long long *counterp, char *counter_path)
tools/power turbostat: extend --add option to accept /sys path Previously, the --add option could specify only an MSR. Here is is extended so an arbitrary /sys attribute, as specified by an absolute file path name. sudo ./turbostat --add /sys/devices/system/cpu/cpu0/cpuidle/state5/usage Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-08 02:41:51 -05:00
{
if (mp->msr_num != 0) {
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
assert(!no_msr);
tools/power turbostat: extend --add option to accept /sys path Previously, the --add option could specify only an MSR. Here is is extended so an arbitrary /sys attribute, as specified by an absolute file path name. sudo ./turbostat --add /sys/devices/system/cpu/cpu0/cpuidle/state5/usage Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-08 02:41:51 -05:00
if (get_msr(cpu, mp->msr_num, counterp))
return -1;
} else {
tools/power turbostat: fix possible sprintf buffer overflow Signed-off-by: Len Brown <len.brown@intel.com>
2017-12-08 17:38:17 -05:00
char path[128 + PATH_BYTES];
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
if (mp->flags & SYSFS_PERCPU) {
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
sprintf(path, "/sys/devices/system/cpu/cpu%d/%s", cpu, mp->sp->path);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
*counterp = snapshot_sysfs_counter(path);
} else {
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
*counterp = snapshot_sysfs_counter(counter_path);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
}
tools/power turbostat: extend --add option to accept /sys path Previously, the --add option could specify only an MSR. Here is is extended so an arbitrary /sys attribute, as specified by an absolute file path name. sudo ./turbostat --add /sys/devices/system/cpu/cpu0/cpuidle/state5/usage Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-08 02:41:51 -05:00
}
return 0;
}
tools/power turbostat: Survive sparse die_id Turbostat assumed that every package had a die_id = 0. When this assumption was violated, it exited when looking for the package uncore frequency: turbostat: /sys/.../intel_uncore_frequency/package_01_die_00/current_freq_khz: open failed: No such file or directory Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 12:02:24 -04:00
unsigned long long get_legacy_uncore_mhz(int package)
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
{
char path[128];
tools/power turbostat: Survive sparse die_id Turbostat assumed that every package had a die_id = 0. When this assumption was violated, it exited when looking for the package uncore frequency: turbostat: /sys/.../intel_uncore_frequency/package_01_die_00/current_freq_khz: open failed: No such file or directory Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 12:02:24 -04:00
int die;
static int warn_once;
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
tools/power turbostat: Survive sparse die_id Turbostat assumed that every package had a die_id = 0. When this assumption was violated, it exited when looking for the package uncore frequency: turbostat: /sys/.../intel_uncore_frequency/package_01_die_00/current_freq_khz: open failed: No such file or directory Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 12:02:24 -04:00
/*
* for this package, use the first die_id that exists
*/
for (die = 0; die <= topo.max_die_id; ++die) {
sprintf(path, "/sys/devices/system/cpu/intel_uncore_frequency/package_%02d_die_%02d/current_freq_khz",
package, die);
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
tools/power turbostat: Survive sparse die_id Turbostat assumed that every package had a die_id = 0. When this assumption was violated, it exited when looking for the package uncore frequency: turbostat: /sys/.../intel_uncore_frequency/package_01_die_00/current_freq_khz: open failed: No such file or directory Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 12:02:24 -04:00
if (access(path, R_OK) == 0)
return (snapshot_sysfs_counter(path) / 1000);
}
if (!warn_once) {
warnx("BUG: %s: No %s", __func__, path);
warn_once = 1;
}
return 0;
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
}
tools/power/turbostat: Read energy_perf_bias from sysfs ... instead of poking at the MSR directly. Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Len Brown <lenb@kernel.org> Cc: linux-pm@vger.kernel.org Link: https://lkml.kernel.org/r/20201029190259.3476-3-bp@alien8.de
2020-10-15 14:50:16 +02:00
int get_epb(int cpu)
{
char path[128 + PATH_BYTES];
tools/power/turbostat: Fallback to an MSR read for EPB Commit 6d6501d912a9 ("tools/power/turbostat: Read energy_perf_bias from sysfs") converted turbostat to read the energy_perf_bias value from sysfs. However, older kernels which do not have that file yet, would fail. For those, fall back to the MSR reading. Fixes: 6d6501d912a9 ("tools/power/turbostat: Read energy_perf_bias from sysfs") Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Borislav Petkov <bp@suse.de> Tested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Link: https://lkml.kernel.org/r/20210127132444.981120-1-dedekind1@gmail.com
2021-01-28 20:28:56 +01:00
unsigned long long msr;
tools/power/turbostat: Read energy_perf_bias from sysfs ... instead of poking at the MSR directly. Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Len Brown <lenb@kernel.org> Cc: linux-pm@vger.kernel.org Link: https://lkml.kernel.org/r/20201029190259.3476-3-bp@alien8.de
2020-10-15 14:50:16 +02:00
int ret, epb = -1;
FILE *fp;
sprintf(path, "/sys/devices/system/cpu/cpu%d/power/energy_perf_bias", cpu);
tools/power/turbostat: Fallback to an MSR read for EPB Commit 6d6501d912a9 ("tools/power/turbostat: Read energy_perf_bias from sysfs") converted turbostat to read the energy_perf_bias value from sysfs. However, older kernels which do not have that file yet, would fail. For those, fall back to the MSR reading. Fixes: 6d6501d912a9 ("tools/power/turbostat: Read energy_perf_bias from sysfs") Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Borislav Petkov <bp@suse.de> Tested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Link: https://lkml.kernel.org/r/20210127132444.981120-1-dedekind1@gmail.com
2021-01-28 20:28:56 +01:00
fp = fopen(path, "r");
if (!fp)
goto msr_fallback;
tools/power/turbostat: Read energy_perf_bias from sysfs ... instead of poking at the MSR directly. Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Len Brown <lenb@kernel.org> Cc: linux-pm@vger.kernel.org Link: https://lkml.kernel.org/r/20201029190259.3476-3-bp@alien8.de
2020-10-15 14:50:16 +02:00
ret = fscanf(fp, "%d", &epb);
if (ret != 1)
err(1, "%s(%s)", __func__, path);
fclose(fp);
return epb;
tools/power/turbostat: Fallback to an MSR read for EPB Commit 6d6501d912a9 ("tools/power/turbostat: Read energy_perf_bias from sysfs") converted turbostat to read the energy_perf_bias value from sysfs. However, older kernels which do not have that file yet, would fail. For those, fall back to the MSR reading. Fixes: 6d6501d912a9 ("tools/power/turbostat: Read energy_perf_bias from sysfs") Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Borislav Petkov <bp@suse.de> Tested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Link: https://lkml.kernel.org/r/20210127132444.981120-1-dedekind1@gmail.com
2021-01-28 20:28:56 +01:00
msr_fallback:
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (no_msr)
return -1;
tools/power/turbostat: Fallback to an MSR read for EPB Commit 6d6501d912a9 ("tools/power/turbostat: Read energy_perf_bias from sysfs") converted turbostat to read the energy_perf_bias value from sysfs. However, older kernels which do not have that file yet, would fail. For those, fall back to the MSR reading. Fixes: 6d6501d912a9 ("tools/power/turbostat: Read energy_perf_bias from sysfs") Reported-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Borislav Petkov <bp@suse.de> Tested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Link: https://lkml.kernel.org/r/20210127132444.981120-1-dedekind1@gmail.com
2021-01-28 20:28:56 +01:00
get_msr(cpu, MSR_IA32_ENERGY_PERF_BIAS, &msr);
return msr & 0xf;
tools/power/turbostat: Read energy_perf_bias from sysfs ... instead of poking at the MSR directly. Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Len Brown <lenb@kernel.org> Cc: linux-pm@vger.kernel.org Link: https://lkml.kernel.org/r/20201029190259.3476-3-bp@alien8.de
2020-10-15 14:50:16 +02:00
}
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
void get_apic_id(struct thread_data *t)
{
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
unsigned int eax, ebx, ecx, edx;
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
if (DO_BIC(BIC_APIC)) {
eax = ebx = ecx = edx = 0;
__cpuid(1, eax, ebx, ecx, edx);
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
t->apic_id = (ebx >> 24) & 0xff;
}
if (!DO_BIC(BIC_X2APIC))
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
return;
tools/power turbostat: Add support for Hygon Fam 18h (Dhyana) RAPL Commit 9392bd98bba760be96ee ("tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL") and the commit 3316f99a9f1b68c578c5 ("tools/power turbostat: Also read package power on AMD F17h (Zen)") add AMD Fam 17h RAPL support. Hygon Family 18h(Dhyana) support RAPL in bit 14 of CPUID 0x80000007 EDX, and has MSRs RAPL_PWR_UNIT/CORE_ENERGY_STAT/PKG_ENERGY_STAT. So add Hygon Dhyana Family 18h support for RAPL. Already tested on Hygon multi-node systems and it shows correct per-core energy usage and the total package power. Signed-off-by: Pu Wen <puwen@hygon.cn> Reviewed-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2019-08-31 10:20:31 +08:00
if (authentic_amd || hygon_genuine) {
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
unsigned int topology_extensions;
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
if (max_extended_level < 0x8000001e)
return;
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
eax = ebx = ecx = edx = 0;
__cpuid(0x80000001, eax, ebx, ecx, edx);
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
topology_extensions = ecx & (1 << 22);
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
if (topology_extensions == 0)
return;
eax = ebx = ecx = edx = 0;
__cpuid(0x8000001e, eax, ebx, ecx, edx);
t->x2apic_id = eax;
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
return;
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
}
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
if (!genuine_intel)
return;
if (max_level < 0xb)
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
return;
ecx = 0;
__cpuid(0xb, eax, ebx, ecx, edx);
t->x2apic_id = edx;
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
if (debug && (t->apic_id != (t->x2apic_id & 0xff)))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "cpu%d: BIOS BUG: apic 0x%x x2apic 0x%x\n", t->cpu_id, t->apic_id, t->x2apic_id);
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
}
tools/power turbostat: Support thermal throttle count print The turbostat data is collected by end user for power evaluationit. However it looks like we are missing enough thermal context there. Already a couple of time we found that power management developer asking something like this: grep -r . /sys/devices/system/cpu/cpu*/thermal_throttle/* Print the per core thermal throttle count so as to get suffificent thermal context. turbostat -i 5 -s Core,CPU,CoreThr Core CPU CoreThr - - 104 0 0 61 0 4 1 1 0 1 5 2 2 104 2 6 3 3 7 3 7 Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-11-12 19:51:59 +08:00
int get_core_throt_cnt(int cpu, unsigned long long *cnt)
{
char path[128 + PATH_BYTES];
unsigned long long tmp;
FILE *fp;
int ret;
sprintf(path, "/sys/devices/system/cpu/cpu%d/thermal_throttle/core_throttle_count", cpu);
fp = fopen(path, "r");
if (!fp)
return -1;
ret = fscanf(fp, "%lld", &tmp);
tools/power turbostat: Fix file pointer leak Currently if a fscanf fails then an early return leaks an open file pointer. Fix this by fclosing the file before the return. Detected using static analysis with cppcheck: tools/power/x86/turbostat/turbostat.c:2039:3: error: Resource leak: fp [resourceLeak] Fixes: eae97e053fe3 ("tools/power turbostat: Support thermal throttle count print") Signed-off-by: Colin Ian King <colin.i.king@gmail.com> Acked-by: Chen Yu <yu.c.chen@intel.com> Reviewed-by: Tom Rix <trix@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-26 14:16:07 +01:00
fclose(fp);
tools/power turbostat: Support thermal throttle count print The turbostat data is collected by end user for power evaluationit. However it looks like we are missing enough thermal context there. Already a couple of time we found that power management developer asking something like this: grep -r . /sys/devices/system/cpu/cpu*/thermal_throttle/* Print the per core thermal throttle count so as to get suffificent thermal context. turbostat -i 5 -s Core,CPU,CoreThr Core CPU CoreThr - - 104 0 0 61 0 4 1 1 0 1 5 2 2 104 2 6 3 3 7 3 7 Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-11-12 19:51:59 +08:00
if (ret != 1)
return -1;
*cnt = tmp;
return 0;
}
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
struct amperf_group_fd {
int aperf; /* Also the group descriptor */
int mperf;
};
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
static int read_perf_counter_info(const char *const path, const char *const parse_format, void *value_ptr)
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
{
int fdmt;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
int bytes_read;
char buf[64];
int ret = -1;
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
fdmt = open(path, O_RDONLY, 0);
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
if (fdmt == -1) {
if (debug)
fprintf(stderr, "Failed to parse perf counter info %s\n", path);
ret = -1;
goto cleanup_and_exit;
}
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
bytes_read = read(fdmt, buf, sizeof(buf) - 1);
if (bytes_read <= 0 || bytes_read >= (int)sizeof(buf)) {
if (debug)
fprintf(stderr, "Failed to parse perf counter info %s\n", path);
ret = -1;
goto cleanup_and_exit;
}
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
buf[bytes_read] = '\0';
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
if (sscanf(buf, parse_format, value_ptr) != 1) {
if (debug)
fprintf(stderr, "Failed to parse perf counter info %s\n", path);
ret = -1;
goto cleanup_and_exit;
}
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
ret = 0;
cleanup_and_exit:
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
close(fdmt);
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
return ret;
}
static unsigned int read_perf_counter_info_n(const char *const path, const char *const parse_format)
{
unsigned int v;
int status;
status = read_perf_counter_info(path, parse_format, &v);
if (status)
v = -1;
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
return v;
}
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
static unsigned int read_perf_type(const char *subsys)
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
{
const char *const path_format = "/sys/bus/event_source/devices/%s/type";
const char *const format = "%u";
char path[128];
snprintf(path, sizeof(path), path_format, subsys);
return read_perf_counter_info_n(path, format);
}
tools/power turbostat: Include umask=%x in perf counter's config Some counters, like cpu/cache-misses/, expose and require umask=%x parameter alongside event=%x in the sysfs perf counter's event file. This change make sure we parse and use it when opening user added counters. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-24 13:17:30 +02:00
static unsigned int read_perf_config(const char *subsys, const char *event_name)
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
{
const char *const path_format = "/sys/bus/event_source/devices/%s/events/%s";
tools/power turbostat: Include umask=%x in perf counter's config Some counters, like cpu/cache-misses/, expose and require umask=%x parameter alongside event=%x in the sysfs perf counter's event file. This change make sure we parse and use it when opening user added counters. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-24 13:17:30 +02:00
FILE *fconfig = NULL;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
char path[128];
tools/power turbostat: Include umask=%x in perf counter's config Some counters, like cpu/cache-misses/, expose and require umask=%x parameter alongside event=%x in the sysfs perf counter's event file. This change make sure we parse and use it when opening user added counters. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-24 13:17:30 +02:00
char config_str[64];
unsigned int config;
unsigned int umask;
bool has_config = false;
bool has_umask = false;
unsigned int ret = -1;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
snprintf(path, sizeof(path), path_format, subsys, event_name);
tools/power turbostat: Include umask=%x in perf counter's config Some counters, like cpu/cache-misses/, expose and require umask=%x parameter alongside event=%x in the sysfs perf counter's event file. This change make sure we parse and use it when opening user added counters. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-24 13:17:30 +02:00
fconfig = fopen(path, "r");
if (!fconfig)
return -1;
if (fgets(config_str, ARRAY_SIZE(config_str), fconfig) != config_str)
goto cleanup_and_exit;
for (char *pconfig_str = &config_str[0]; pconfig_str;) {
if (sscanf(pconfig_str, "event=%x", &config) == 1) {
has_config = true;
goto next;
}
if (sscanf(pconfig_str, "umask=%x", &umask) == 1) {
has_umask = true;
goto next;
}
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
next:
tools/power turbostat: Include umask=%x in perf counter's config Some counters, like cpu/cache-misses/, expose and require umask=%x parameter alongside event=%x in the sysfs perf counter's event file. This change make sure we parse and use it when opening user added counters. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-24 13:17:30 +02:00
pconfig_str = strchr(pconfig_str, ',');
if (pconfig_str) {
*pconfig_str = '\0';
++pconfig_str;
}
}
if (!has_umask)
umask = 0;
if (has_config)
ret = (umask << 8) | config;
cleanup_and_exit:
fclose(fconfig);
return ret;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
}
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
static unsigned int read_perf_rapl_unit(const char *subsys, const char *event_name)
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
{
const char *const path_format = "/sys/bus/event_source/devices/%s/events/%s.unit";
const char *const format = "%s";
char path[128];
char unit_buffer[16];
snprintf(path, sizeof(path), path_format, subsys, event_name);
read_perf_counter_info(path, format, &unit_buffer);
if (strcmp("Joules", unit_buffer) == 0)
return RAPL_UNIT_JOULES;
return RAPL_UNIT_INVALID;
}
tools/power turbostat: Include umask=%x in perf counter's config Some counters, like cpu/cache-misses/, expose and require umask=%x parameter alongside event=%x in the sysfs perf counter's event file. This change make sure we parse and use it when opening user added counters. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-24 13:17:30 +02:00
static double read_perf_scale(const char *subsys, const char *event_name)
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
{
const char *const path_format = "/sys/bus/event_source/devices/%s/events/%s.scale";
const char *const format = "%lf";
char path[128];
double scale;
snprintf(path, sizeof(path), path_format, subsys, event_name);
if (read_perf_counter_info(path, format, &scale))
return 0.0;
return scale;
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
}
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
size_t rapl_counter_info_count_perf(const struct rapl_counter_info_t *rci)
{
size_t ret = 0;
for (int i = 0; i < NUM_RAPL_COUNTERS; ++i)
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
if (rci->source[i] == COUNTER_SOURCE_PERF)
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
++ret;
return ret;
}
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
static size_t cstate_counter_info_count_perf(const struct cstate_counter_info_t *cci)
{
size_t ret = 0;
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
for (int i = 0; i < NUM_CSTATE_COUNTERS; ++i)
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
if (cci->source[i] == COUNTER_SOURCE_PERF)
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
++ret;
return ret;
}
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
void write_rapl_counter(struct rapl_counter *rc, struct rapl_counter_info_t *rci, unsigned int idx)
{
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
if (rci->source[idx] == COUNTER_SOURCE_NONE)
return;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rc->raw_value = rci->data[idx];
rc->unit = rci->unit[idx];
rc->scale = rci->scale[idx];
}
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs Save the highest core and package id when parsing topology to allocate enough memory when get_rapl_counters() is called with a core or a package id as a domain. Note that RAPL domains are per-package on Intel, but per-core on AMD. Thus, the RAPL code effectively runs in different modes on those two product lines. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-06 15:39:08 +02:00
int get_rapl_counters(int cpu, unsigned int domain, struct core_data *c, struct pkg_data *p)
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
{
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
struct platform_counters *pplat_cnt = p == package_odd ? &platform_counters_odd : &platform_counters_even;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
unsigned long long perf_data[NUM_RAPL_COUNTERS + 1];
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs Save the highest core and package id when parsing topology to allocate enough memory when get_rapl_counters() is called with a core or a package id as a domain. Note that RAPL domains are per-package on Intel, but per-core on AMD. Thus, the RAPL code effectively runs in different modes on those two product lines. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-06 15:39:08 +02:00
struct rapl_counter_info_t *rci;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
tools/power turbostat: Move verbose counter messages to level 2 Printing information about the source and value during initialization and reading of the counter for each cpu, while useful when debugging, results in too verbose output. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 14:28:03 +02:00
if (debug >= 2)
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
fprintf(stderr, "%s: cpu%d domain%d\n", __func__, cpu, domain);
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
assert(rapl_counter_info_perdomain);
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs Save the highest core and package id when parsing topology to allocate enough memory when get_rapl_counters() is called with a core or a package id as a domain. Note that RAPL domains are per-package on Intel, but per-core on AMD. Thus, the RAPL code effectively runs in different modes on those two product lines. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-06 15:39:08 +02:00
assert(domain < rapl_counter_info_perdomain_size);
rci = &rapl_counter_info_perdomain[domain];
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
/*
* If we have any perf counters to read, read them all now, in bulk
*/
if (rci->fd_perf != -1) {
size_t num_perf_counters = rapl_counter_info_count_perf(rci);
const ssize_t expected_read_size = (num_perf_counters + 1) * sizeof(unsigned long long);
const ssize_t actual_read_size = read(rci->fd_perf, &perf_data[0], sizeof(perf_data));
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
if (actual_read_size != expected_read_size)
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
err(-1, "%s: failed to read perf_data (%zu %zu)", __func__, expected_read_size,
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
actual_read_size);
}
for (unsigned int i = 0, pi = 1; i < NUM_RAPL_COUNTERS; ++i) {
switch (rci->source[i]) {
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
case COUNTER_SOURCE_NONE:
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
rci->data[i] = 0;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
break;
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
case COUNTER_SOURCE_PERF:
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
assert(pi < ARRAY_SIZE(perf_data));
assert(rci->fd_perf != -1);
tools/power turbostat: Move verbose counter messages to level 2 Printing information about the source and value during initialization and reading of the counter for each cpu, while useful when debugging, results in too verbose output. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 14:28:03 +02:00
if (debug >= 2)
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
fprintf(stderr, "Reading rapl counter via perf at %u (%llu %e %lf)\n",
i, perf_data[pi], rci->scale[i], perf_data[pi] * rci->scale[i]);
rci->data[i] = perf_data[pi];
++pi;
break;
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
case COUNTER_SOURCE_MSR:
tools/power turbostat: Move verbose counter messages to level 2 Printing information about the source and value during initialization and reading of the counter for each cpu, while useful when debugging, results in too verbose output. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 14:28:03 +02:00
if (debug >= 2)
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
fprintf(stderr, "Reading rapl counter via msr at %u\n", i);
assert(!no_msr);
if (rci->flags[i] & RAPL_COUNTER_FLAG_USE_MSR_SUM) {
if (get_msr_sum(cpu, rci->msr[i], &rci->data[i]))
return -13 - i;
} else {
if (get_msr(cpu, rci->msr[i], &rci->data[i]))
return -13 - i;
}
rci->data[i] &= rci->msr_mask[i];
if (rci->msr_shift[i] >= 0)
rci->data[i] >>= abs(rci->msr_shift[i]);
else
rci->data[i] <<= abs(rci->msr_shift[i]);
break;
}
}
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
BUILD_BUG_ON(NUM_RAPL_COUNTERS != 8);
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
write_rapl_counter(&p->energy_pkg, rci, RAPL_RCI_INDEX_ENERGY_PKG);
write_rapl_counter(&p->energy_cores, rci, RAPL_RCI_INDEX_ENERGY_CORES);
write_rapl_counter(&p->energy_dram, rci, RAPL_RCI_INDEX_DRAM);
write_rapl_counter(&p->energy_gfx, rci, RAPL_RCI_INDEX_GFX);
write_rapl_counter(&p->rapl_pkg_perf_status, rci, RAPL_RCI_INDEX_PKG_PERF_STATUS);
write_rapl_counter(&p->rapl_dram_perf_status, rci, RAPL_RCI_INDEX_DRAM_PERF_STATUS);
write_rapl_counter(&c->core_energy, rci, RAPL_RCI_INDEX_CORE_ENERGY);
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
write_rapl_counter(&pplat_cnt->energy_psys, rci, RAPL_RCI_INDEX_ENERGY_PLATFORM);
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
return 0;
}
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
char *find_sysfs_path_by_id(struct sysfs_path *sp, int id)
{
while (sp) {
if (sp->id == id)
return (sp->path);
sp = sp->next;
}
if (debug)
warnx("%s: id%d not found", __func__, id);
return NULL;
}
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
int get_cstate_counters(unsigned int cpu, struct thread_data *t, struct core_data *c, struct pkg_data *p)
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
{
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
/*
* Overcommit memory a little bit here,
* but skip calculating exact sizes for the buffers.
*/
unsigned long long perf_data[NUM_CSTATE_COUNTERS];
unsigned long long perf_data_core[NUM_CSTATE_COUNTERS + 1];
unsigned long long perf_data_pkg[NUM_CSTATE_COUNTERS + 1];
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
struct cstate_counter_info_t *cci;
tools/power turbostat: Move verbose counter messages to level 2 Printing information about the source and value during initialization and reading of the counter for each cpu, while useful when debugging, results in too verbose output. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 14:28:03 +02:00
if (debug >= 2)
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
fprintf(stderr, "%s: cpu%d\n", __func__, cpu);
assert(ccstate_counter_info);
assert(cpu <= ccstate_counter_info_size);
tools/power turbostat: Add ZERO_ARRAY for zero initializing builtin array It makes it harder to shoot yourself in the foot, by using additional __must_be_array() check. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 13:04:33 +02:00
ZERO_ARRAY(perf_data);
ZERO_ARRAY(perf_data_core);
ZERO_ARRAY(perf_data_pkg);
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
cci = &ccstate_counter_info[cpu];
/*
* If we have any perf counters to read, read them all now, in bulk
*/
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
const size_t num_perf_counters = cstate_counter_info_count_perf(cci);
ssize_t expected_read_size = num_perf_counters * sizeof(unsigned long long);
ssize_t actual_read_size_core = 0, actual_read_size_pkg = 0;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
if (cci->fd_perf_core != -1) {
/* Each descriptor read begins with number of counters read. */
expected_read_size += sizeof(unsigned long long);
actual_read_size_core = read(cci->fd_perf_core, &perf_data_core[0], sizeof(perf_data_core));
if (actual_read_size_core <= 0)
err(-1, "%s: read perf %s: %ld", __func__, "core", actual_read_size_core);
}
if (cci->fd_perf_pkg != -1) {
/* Each descriptor read begins with number of counters read. */
expected_read_size += sizeof(unsigned long long);
actual_read_size_pkg = read(cci->fd_perf_pkg, &perf_data_pkg[0], sizeof(perf_data_pkg));
if (actual_read_size_pkg <= 0)
err(-1, "%s: read perf %s: %ld", __func__, "pkg", actual_read_size_pkg);
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
}
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
const ssize_t actual_read_size_total = actual_read_size_core + actual_read_size_pkg;
if (actual_read_size_total != expected_read_size)
err(-1, "%s: failed to read perf_data (%zu %zu)", __func__, expected_read_size, actual_read_size_total);
/*
* Copy ccstate and pcstate data into unified buffer.
*
* Skip first element from core and pkg buffers.
* Kernel puts there how many counters were read.
*/
const size_t num_core_counters = perf_data_core[0];
const size_t num_pkg_counters = perf_data_pkg[0];
assert(num_perf_counters == num_core_counters + num_pkg_counters);
/* Copy ccstate perf data */
memcpy(&perf_data[0], &perf_data_core[1], num_core_counters * sizeof(unsigned long long));
/* Copy pcstate perf data */
memcpy(&perf_data[num_core_counters], &perf_data_pkg[1], num_pkg_counters * sizeof(unsigned long long));
for (unsigned int i = 0, pi = 0; i < NUM_CSTATE_COUNTERS; ++i) {
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
switch (cci->source[i]) {
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
case COUNTER_SOURCE_NONE:
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
break;
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
case COUNTER_SOURCE_PERF:
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
assert(pi < ARRAY_SIZE(perf_data));
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
assert(cci->fd_perf_core != -1 || cci->fd_perf_pkg != -1);
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
tools/power turbostat: Move verbose counter messages to level 2 Printing information about the source and value during initialization and reading of the counter for each cpu, while useful when debugging, results in too verbose output. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 14:28:03 +02:00
if (debug >= 2)
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
fprintf(stderr, "cstate via %s %u: %llu\n", "perf", i, perf_data[pi]);
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
cci->data[i] = perf_data[pi];
++pi;
break;
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
case COUNTER_SOURCE_MSR:
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
assert(!no_msr);
if (get_msr(cpu, cci->msr[i], &cci->data[i]))
return -13 - i;
tools/power turbostat: Move verbose counter messages to level 2 Printing information about the source and value during initialization and reading of the counter for each cpu, while useful when debugging, results in too verbose output. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 14:28:03 +02:00
if (debug >= 2)
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
fprintf(stderr, "cstate via %s0x%llx %u: %llu\n", "msr", cci->msr[i], i, cci->data[i]);
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
break;
}
}
/*
* Helper to write the data only if the source of
* the counter for the current cpu is not none.
*
* Otherwise we would overwrite core data with 0 (default value),
* when invoked for the thread sibling.
*/
#define PERF_COUNTER_WRITE_DATA(out_counter, index) do { \
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
if (cci->source[index] != COUNTER_SOURCE_NONE) \
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
out_counter = cci->data[index]; \
} while (0)
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
BUILD_BUG_ON(NUM_CSTATE_COUNTERS != 11);
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
PERF_COUNTER_WRITE_DATA(t->c1, CCSTATE_RCI_INDEX_C1_RESIDENCY);
PERF_COUNTER_WRITE_DATA(c->c3, CCSTATE_RCI_INDEX_C3_RESIDENCY);
PERF_COUNTER_WRITE_DATA(c->c6, CCSTATE_RCI_INDEX_C6_RESIDENCY);
PERF_COUNTER_WRITE_DATA(c->c7, CCSTATE_RCI_INDEX_C7_RESIDENCY);
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
PERF_COUNTER_WRITE_DATA(p->pc2, PCSTATE_RCI_INDEX_C2_RESIDENCY);
PERF_COUNTER_WRITE_DATA(p->pc3, PCSTATE_RCI_INDEX_C3_RESIDENCY);
PERF_COUNTER_WRITE_DATA(p->pc6, PCSTATE_RCI_INDEX_C6_RESIDENCY);
PERF_COUNTER_WRITE_DATA(p->pc7, PCSTATE_RCI_INDEX_C7_RESIDENCY);
PERF_COUNTER_WRITE_DATA(p->pc8, PCSTATE_RCI_INDEX_C8_RESIDENCY);
PERF_COUNTER_WRITE_DATA(p->pc9, PCSTATE_RCI_INDEX_C9_RESIDENCY);
PERF_COUNTER_WRITE_DATA(p->pc10, PCSTATE_RCI_INDEX_C10_RESIDENCY);
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
#undef PERF_COUNTER_WRITE_DATA
return 0;
}
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
size_t msr_counter_info_count_perf(const struct msr_counter_info_t *mci)
{
size_t ret = 0;
for (int i = 0; i < NUM_MSR_COUNTERS; ++i)
if (mci->source[i] == COUNTER_SOURCE_PERF)
++ret;
return ret;
}
int get_smi_aperf_mperf(unsigned int cpu, struct thread_data *t)
{
unsigned long long perf_data[NUM_MSR_COUNTERS + 1];
struct msr_counter_info_t *mci;
tools/power turbostat: Move verbose counter messages to level 2 Printing information about the source and value during initialization and reading of the counter for each cpu, while useful when debugging, results in too verbose output. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 14:28:03 +02:00
if (debug >= 2)
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
fprintf(stderr, "%s: cpu%d\n", __func__, cpu);
assert(msr_counter_info);
assert(cpu <= msr_counter_info_size);
mci = &msr_counter_info[cpu];
ZERO_ARRAY(perf_data);
ZERO_ARRAY(mci->data);
if (mci->fd_perf != -1) {
const size_t num_perf_counters = msr_counter_info_count_perf(mci);
const ssize_t expected_read_size = (num_perf_counters + 1) * sizeof(unsigned long long);
const ssize_t actual_read_size = read(mci->fd_perf, &perf_data[0], sizeof(perf_data));
if (actual_read_size != expected_read_size)
err(-1, "%s: failed to read perf_data (%zu %zu)", __func__, expected_read_size,
actual_read_size);
}
for (unsigned int i = 0, pi = 1; i < NUM_MSR_COUNTERS; ++i) {
switch (mci->source[i]) {
case COUNTER_SOURCE_NONE:
break;
case COUNTER_SOURCE_PERF:
assert(pi < ARRAY_SIZE(perf_data));
assert(mci->fd_perf != -1);
tools/power turbostat: Move verbose counter messages to level 2 Printing information about the source and value during initialization and reading of the counter for each cpu, while useful when debugging, results in too verbose output. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 14:28:03 +02:00
if (debug >= 2)
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
fprintf(stderr, "Reading msr counter via perf at %u: %llu\n", i, perf_data[pi]);
mci->data[i] = perf_data[pi];
++pi;
break;
case COUNTER_SOURCE_MSR:
assert(!no_msr);
if (get_msr(cpu, mci->msr[i], &mci->data[i]))
return -2 - i;
mci->data[i] &= mci->msr_mask[i];
tools/power turbostat: Move verbose counter messages to level 2 Printing information about the source and value during initialization and reading of the counter for each cpu, while useful when debugging, results in too verbose output. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 14:28:03 +02:00
if (debug >= 2)
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
fprintf(stderr, "Reading msr counter via msr at %u: %llu\n", i, mci->data[i]);
break;
}
}
BUILD_BUG_ON(NUM_MSR_COUNTERS != 3);
t->aperf = mci->data[MSR_RCI_INDEX_APERF];
t->mperf = mci->data[MSR_RCI_INDEX_MPERF];
t->smi_count = mci->data[MSR_RCI_INDEX_SMI];
return 0;
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
int perf_counter_info_read_values(struct perf_counter_info *pp, int cpu, unsigned long long *out, size_t out_size)
{
unsigned int domain;
unsigned long long value;
int fd_counter;
for (size_t i = 0; pp; ++i, pp = pp->next) {
domain = cpu_to_domain(pp, cpu);
assert(domain < pp->num_domains);
fd_counter = pp->fd_perf_per_domain[domain];
if (fd_counter == -1)
continue;
if (read(fd_counter, &value, sizeof(value)) != sizeof(value))
return 1;
assert(i < out_size);
out[i] = value * pp->scale;
}
return 0;
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
unsigned long pmt_gen_value_mask(unsigned int lsb, unsigned int msb)
{
unsigned long mask;
if (msb == 63)
mask = 0xffffffffffffffff;
else
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
mask = ((1 << (msb + 1)) - 1);
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
mask -= (1 << lsb) - 1;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
return mask;
}
unsigned long pmt_read_counter(struct pmt_counter *ppmt, unsigned int domain_id)
{
assert(domain_id < ppmt->num_domains);
const unsigned long *pmmio = ppmt->domains[domain_id].pcounter;
const unsigned long value = pmmio ? *pmmio : 0;
const unsigned long value_mask = pmt_gen_value_mask(ppmt->lsb, ppmt->msb);
const unsigned long value_shift = ppmt->lsb;
return (value & value_mask) >> value_shift;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/*
* get_counters(...)
* migrate to cpu
* acquire and record local counters for that cpu
*/
int get_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
int cpu = t->cpu_id;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
unsigned long long msr;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
struct msr_counter *mp;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
struct pmt_counter *pp;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
int i;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
int status;
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 21:44:40 -04:00
tools/power turbostat: prevent infinite loop on migration error path Turbostat assumed if it can't migrate to a CPU, then the CPU must have gone off-line and turbostat should re-initialize with the new topology. But if turbostat can not migrate because it is restricted by a cpuset, then it will fail to migrate even after re-initialization, resulting in an infinite loop. Spit out a warning when we can't migrate and endure only 2 re-initialize cycles in a row before giving up and exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 22:38:05 -05:00
if (cpu_migrate(cpu)) {
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
fprintf(outf, "%s: Could not migrate to CPU %d\n", __func__, cpu);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return -1;
tools/power turbostat: prevent infinite loop on migration error path Turbostat assumed if it can't migrate to a CPU, then the CPU must have gone off-line and turbostat should re-initialize with the new topology. But if turbostat can not migrate because it is restricted by a cpuset, then it will fail to migrate even after re-initialization, resulting in an infinite loop. Spit out a warning when we can't migrate and endure only 2 re-initialize cycles in a row before giving up and exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 22:38:05 -05:00
}
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 22:19:58 -04:00
tools/power turbostat: Make interval calculation per thread to reduce jitter Turbostat currently normalizes TSC and other values by dividing by an interval. This interval is the delta between the start of one global (all counters on all CPUs) sampling and the start of another. However, this introduces a lot of jitter into the data. In order to reduce jitter, the interval calculation should be based on timestamps taken per thread and close to the start of the thread's sampling. Define a per thread time value to hold the delta between samples taken on the thread. Use the timestamp taken at the beginning of sampling to calculate the delta. Move the thread's beginning timestamp to after the CPU migration to avoid jitter due to the migration. Use the global time delta for the average time delta. Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-25 17:32:42 +00:00
gettimeofday(&t->tv_begin, (struct timezone *)NULL);
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
if (first_counter_read)
get_apic_id(t);
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
t->tsc = rdtsc(); /* we are running on local CPU of interest */
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
get_smi_aperf_mperf(cpu, t);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
if (DO_BIC(BIC_IPC))
if (read(get_instr_count_fd(cpu), &t->instr_count, sizeof(long long)) != sizeof(long long))
return -4;
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_IRQ))
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
t->irq_count = irqs_per_cpu[cpu];
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
get_cstate_counters(cpu, t, c, p);
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.tp; mp; i++, mp = mp->next) {
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
if (get_mp(cpu, mp, &t->counter[i], mp->sp->path))
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
return -10;
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
if (perf_counter_info_read_values(sys.perf_tp, cpu, t->perf_counter, MAX_ADDED_THREAD_COUNTERS))
return -10;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
for (i = 0, pp = sys.pmt_tp; pp; i++, pp = pp->next)
t->pmt_counter[i] = pmt_read_counter(pp, t->cpu_id);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/* collect core counters only for 1st thread in core */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (!is_cpu_first_thread_in_core(t, c, p))
tools/power turbostat: if --debug, print sampling overhead The --debug option now pre-pends each row with the number of micro-seconds [usec] to collect the finishing snapshot for that row. Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-27 21:06:55 -07:00
goto done;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
if (platform->has_per_core_rapl) {
status = get_rapl_counters(cpu, c->core_id, c, p);
if (status != 0)
return status;
}
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
if (DO_BIC(BIC_CPU_c7) && t->is_atom) {
/*
* For Atom CPUs that has core cstate deeper than c6,
* MSR_CORE_C6_RESIDENCY returns residency of cc6 and deeper.
* Minus CC7 (and deeper cstates) residency to get
* accturate cc6 residency.
*/
c->c6 -= c->c7;
tools/power turbostat: Fix Core C6 residency on Atom CPUs For Atom CPUs that have core cstate deeper than C6, MSR_CORE_C6_RESIDENCY actually returns the residency for both CC6 and deeper Core cstates. Thus, the real Core C6 residency should be the subtraction of MSR_CORE_C6_RESIDENCY return value and MSR_CORE_C6_RESIDENCY return value. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-21 22:22:47 +08:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: Baytrail c-state support The Baytrail SOC, with its Silvermont core, has some unique properties: 1. a hardware CC1 residency counter 2. a module-c6 residency counter 3. a package-c6 counter at traditional package-c7 counter address. The SOC does not support c3, pc3, c7 or pc7 counters. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:27:20 -05:00
if (DO_BIC(BIC_Mod_c6))
if (get_msr(cpu, MSR_MODULE_C6_RES_MS, &c->mc6_us))
return -8;
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_CoreTmp)) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
if (get_msr(cpu, MSR_IA32_THERM_STATUS, &msr))
return -9;
tools/power turbostat: rename tcc variables There are two TCC activation temeprature. One is the default TCC activation temperature, also known as TJ_MAX. Another one is the effective TCC activation temperature, which is the subtraction of default TCC activation temperature and TCC offset. The name of variable tcc_activation_temp might be misleading here. Thus rename tcc_activation_temp to tj_max, and use tcc_default and tcc_offset to calculate the effective TCC activation temperature. No functional change in this patch. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-26 18:49:26 +08:00
c->core_temp_c = tj_max - ((msr >> 16) & 0x7F);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
}
tools/power turbostat: Support thermal throttle count print The turbostat data is collected by end user for power evaluationit. However it looks like we are missing enough thermal context there. Already a couple of time we found that power management developer asking something like this: grep -r . /sys/devices/system/cpu/cpu*/thermal_throttle/* Print the per core thermal throttle count so as to get suffificent thermal context. turbostat -i 5 -s Core,CPU,CoreThr Core CPU CoreThr - - 104 0 0 61 0 4 1 1 0 1 5 2 2 104 2 6 3 3 7 3 7 Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-11-12 19:51:59 +08:00
if (DO_BIC(BIC_CORE_THROT_CNT))
get_core_throt_cnt(cpu, &c->core_throt_cnt);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) {
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
if (get_mp(cpu, mp, &c->counter[i], mp->sp->path))
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
return -10;
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
if (perf_counter_info_read_values(sys.perf_cp, cpu, c->perf_counter, MAX_ADDED_CORE_COUNTERS))
return -10;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
for (i = 0, pp = sys.pmt_cp; pp; i++, pp = pp->next)
c->pmt_counter[i] = pmt_read_counter(pp, c->core_id);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/* collect package counters only for 1st core in package */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (!is_cpu_first_core_in_package(t, c, p))
tools/power turbostat: if --debug, print sampling overhead The --debug option now pre-pends each row with the number of micro-seconds [usec] to collect the finishing snapshot for that row. Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-27 21:06:55 -07:00
goto done;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
if (DO_BIC(BIC_Totl_c0)) {
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
if (get_msr(cpu, MSR_PKG_WEIGHTED_CORE_C0_RES, &p->pkg_wtd_core_c0))
return -10;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
}
if (DO_BIC(BIC_Any_c0)) {
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
if (get_msr(cpu, MSR_PKG_ANY_CORE_C0_RES, &p->pkg_any_core_c0))
return -11;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
}
if (DO_BIC(BIC_GFX_c0)) {
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
if (get_msr(cpu, MSR_PKG_ANY_GFXE_C0_RES, &p->pkg_any_gfxe_c0))
return -12;
tools/power turbostat: hide SKL counters, when not requested Skylake has some new counters, and they were erroneously exempt from --show and --hide eg. turbostat --quiet --show CPU CPU Totl%C0 Any%C0 GFX%C0 CPUGFX% - 116.73 90.56 85.69 79.00 0 117.78 91.38 86.47 79.71 2 1 3 is now CPU - 0 2 1 3 Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-20 20:11:55 -04:00
}
if (DO_BIC(BIC_CPUGFX)) {
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
if (get_msr(cpu, MSR_PKG_BOTH_CORE_GFXE_C0_RES, &p->pkg_both_core_gfxe_c0))
return -13;
}
tools/power turbostat: bug fixes to --add, --show/--hide features Fix a bug with --add, where the title of the column is un-initialized if not specified by the user. The initial implementation of --show and --hide neglected to handle the pc8/pc9/pc10 counters. Fix a bug where "--show Core" only worked with --debug Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-27 00:50:45 -05:00
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
if (DO_BIC(BIC_CPU_LPI))
p->cpu_lpi = cpuidle_cur_cpu_lpi_us;
if (DO_BIC(BIC_SYS_LPI))
p->sys_lpi = cpuidle_cur_sys_lpi_us;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
if (!platform->has_per_core_rapl) {
status = get_rapl_counters(cpu, p->package_id, c, p);
if (status != 0)
return status;
tools/power turbostat: Also read package power on AMD F17h (Zen) The package power can also be read from an MSR. It's not clear exactly what is included, and whether it's aggregated over all nodes or reported separately. It does look like this is reported separately per CCX (I get a single value on the Ryzen R7 1700), but it might be reported separately per- die (node?) on larger processors. If that's the case, it would have to be recorded per node and aggregated for the socket. Note that although Zen has these MSRs reporting power, it looks like the actual RAPL configuration (power limits, configured TDP) is done through PCI configuration space. I have not yet found any public documentation for this. Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:42 -04:00
}
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_PkgTmp)) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
if (get_msr(cpu, MSR_IA32_PACKAGE_THERM_STATUS, &msr))
return -17;
tools/power turbostat: rename tcc variables There are two TCC activation temeprature. One is the default TCC activation temperature, also known as TJ_MAX. Another one is the effective TCC activation temperature, which is the subtraction of default TCC activation temperature and TCC offset. The name of variable tcc_activation_temp might be misleading here. Thus rename tcc_activation_temp to tj_max, and use tcc_default and tcc_offset to calculate the effective TCC activation temperature. No functional change in this patch. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-26 18:49:26 +08:00
p->pkg_temp_c = tj_max - ((msr >> 16) & 0x7F);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
}
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 01:28:12 -05:00
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
if (DO_BIC(BIC_UNCORE_MHZ))
tools/power turbostat: Survive sparse die_id Turbostat assumed that every package had a die_id = 0. When this assumption was violated, it exited when looking for the package uncore frequency: turbostat: /sys/.../intel_uncore_frequency/package_01_die_00/current_freq_khz: open failed: No such file or directory Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 12:02:24 -04:00
p->uncore_mhz = get_legacy_uncore_mhz(p->package_id);
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
if (DO_BIC(BIC_GFX_rc6))
p->gfx_rc6_ms = gfx_info[GFX_rc6].val_ull;
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_GFXMHz))
tools/power/turbostat: Unify graphics sysfs snapshots Graphics sysfs snapshots share similar logic. Combine them into one function to avoid code duplication. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-13 10:12:19 +08:00
p->gfx_mhz = gfx_info[GFX_MHz].val;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 00:37:54 -05:00
tools/power turbostat: Add a new GFXAMHz column that exposes gt_act_freq_mhz. The column already present called GFXMHz reads from gt_cur_freq_mhz, which represents the GT frequency that was requested, but power management might not be able to do that. So the new column will display what the actual frequency GT is running at. Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-22 15:02:07 -07:00
if (DO_BIC(BIC_GFXACTMHz))
tools/power/turbostat: Unify graphics sysfs snapshots Graphics sysfs snapshots share similar logic. Combine them into one function to avoid code duplication. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-13 10:12:19 +08:00
p->gfx_act_mhz = gfx_info[GFX_ACTMHz].val;
tools/power turbostat: Add a new GFXAMHz column that exposes gt_act_freq_mhz. The column already present called GFXMHz reads from gt_cur_freq_mhz, which represents the GT frequency that was requested, but power management might not be able to do that. So the new column will display what the actual frequency GT is running at. Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-22 15:02:07 -07:00
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
if (DO_BIC(BIC_SAM_mc6))
p->sam_mc6_ms = gfx_info[SAM_mc6].val_ull;
if (DO_BIC(BIC_SAMMHz))
p->sam_mhz = gfx_info[SAM_MHz].val;
if (DO_BIC(BIC_SAMACTMHz))
p->sam_act_mhz = gfx_info[SAM_ACTMHz].val;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
for (i = 0, mp = sys.pp; mp; i++, mp = mp->next) {
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
char *path = NULL;
if (mp->msr_num == 0) {
path = find_sysfs_path_by_id(mp->sp, p->package_id);
if (path == NULL) {
warnx("%s: package_id %d not found", __func__, p->package_id);
return -10;
}
}
if (get_mp(cpu, mp, &p->counter[i], path))
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
return -10;
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
if (perf_counter_info_read_values(sys.perf_pp, cpu, p->perf_counter, MAX_ADDED_PACKAGE_COUNTERS))
return -10;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
for (i = 0, pp = sys.pmt_pp; pp; i++, pp = pp->next)
p->pmt_counter[i] = pmt_read_counter(pp, p->package_id);
tools/power turbostat: if --debug, print sampling overhead The --debug option now pre-pends each row with the number of micro-seconds [usec] to collect the finishing snapshot for that row. Signed-off-by: Len Brown <len.brown@intel.com>
2017-05-27 21:06:55 -07:00
done:
gettimeofday(&t->tv_end, (struct timezone *)NULL);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 22:19:58 -04:00
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-09 23:39:45 -05:00
int pkg_cstate_limit = PCLUKN;
tools/power turbostat: Fix order of strings in pkg_cstate_limit_strings Change the order so that it matches the indexes defined in: Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-09 12:24:02 +02:00
char *pkg_cstate_limit_strings[] = { "unknown", "reserved", "pc0", "pc1", "pc2",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
"pc3", "pc4", "pc6", "pc6n", "pc6r", "pc7", "pc7s", "pc8", "pc9", "pc10", "unlimited"
};
tools/power turbostat: SKL: Adjust for TSC difference from base frequency On a Skylake with 1500MHz base frequency, the TSC runs at 1512MHz. This is because the TSC is no longer in the n*100 MHz BCLK domain, but is now in the m*24MHz crystal clock domain. (24 MHz * 63 = 1512 MHz) This adds error to several calculations in turbostat, unless the TSC sample sizes are adjusted for this difference. Note that calculations in the time domain are immune from this issue, as the timing sub-system has already calibrated the TSC against a known wall clock. AVG_MHz = APERF_delta/measurement_interval need no adjustment. APERF_delta is in the BCLK domain, and measurement_interval is in the time domain. TSC_MHz = TSC_delta/measurement_interval needs no adjustment -- as we really do want to report the actual measured TSC delta here, and measurement_interval is in the accurate time domain. %Busy = MPERF_delta/TSC_delta needs adjustment to use TSC_BCLK_DOMAIN_delta. TSC_BCLK_DOMAIN_delta = TSC_delta * base_hz / tsc_hz Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/measurement_interval need adjustment as above. No other metrics in turbostat need to be adjusted. Before: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 24.84 2216 1512 0 2191 98.73 2219 1514 2 0 0.01 2130 1512 1 9 0.43 2016 1512 3 2 0.08 2016 1512 After: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 25.05 2198 1512 0 2190 99.62 2199 1512 2 0 0.01 2152 1512 1 9 0.46 2000 1512 3 2 0.10 2000 1512 Note that in this example, the "Before" Bzy_MHz was reported as exceeding the 2200 max turbo rate. Also, even a pinned spin loop would not be reported as over 99% busy. Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-26 00:12:38 -04:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int nhm_pkg_cstate_limits[16] =
{ PCL__0, PCL__1, PCL__3, PCL__6, PCL__7, PCLRSV, PCLRSV, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV,
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
PCLRSV, PCLRSV
};
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int snb_pkg_cstate_limits[16] =
{ PCL__0, PCL__2, PCL_6N, PCL_6R, PCL__7, PCL_7S, PCLRSV, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV,
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
PCLRSV, PCLRSV
};
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int hsw_pkg_cstate_limits[16] =
{ PCL__0, PCL__2, PCL__3, PCL__6, PCL__7, PCL_7S, PCL__8, PCL__9, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV,
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
PCLRSV, PCLRSV
};
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int slv_pkg_cstate_limits[16] =
{ PCL__0, PCL__1, PCLRSV, PCLRSV, PCL__4, PCLRSV, PCL__6, PCL__7, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV,
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
PCL__6, PCL__7
};
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int amt_pkg_cstate_limits[16] =
{ PCLUNL, PCL__1, PCL__2, PCLRSV, PCLRSV, PCLRSV, PCL__6, PCL__7, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV,
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
PCLRSV, PCLRSV
};
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int phi_pkg_cstate_limits[16] =
{ PCL__0, PCL__2, PCL_6N, PCL_6R, PCLRSV, PCLRSV, PCLRSV, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV,
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
PCLRSV, PCLRSV
};
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int glm_pkg_cstate_limits[16] =
{ PCLUNL, PCL__1, PCL__3, PCL__6, PCL__7, PCL_7S, PCL__8, PCL__9, PCL_10, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV,
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
PCLRSV, PCLRSV
};
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int skx_pkg_cstate_limits[16] =
{ PCL__0, PCL__2, PCL_6N, PCL_6R, PCLRSV, PCLRSV, PCLRSV, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV,
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
PCLRSV, PCLRSV
};
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int icx_pkg_cstate_limits[16] =
turbostat: fix PC6 displaying on some systems 'MSR_PKG_CST_CONFIG_CONTROL' encodes the deepest allowed package C-state limit, and turbostat decodes it. Before this patch: turbostat does not recognize value "3" on Ice Lake Xeon (ICX) and Sapphire Rapids Xeon (SPR), treats it as "unknown", and does not display any package C-states in the results table. After this patch: turbostat recognizes value 3 on ICX and SPR, treats it as "PC6", and correctly displays package C-states in the results table. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2022-03-08 13:41:01 +02:00
{ PCL__0, PCL__2, PCL__6, PCL__6, PCLRSV, PCLRSV, PCLRSV, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV,
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
PCLRSV, PCLRSV
};
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
void probe_cst_limit(void)
{
unsigned long long msr;
int *pkg_cstate_limits;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (!platform->has_nhm_msrs || no_msr)
tools/power/turbostat: Abstract Package cstate limit decoding support Abstract the support for decoding package cstate limit from MSR_PKG_CST_CONFIG_CONTROL. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:06 +08:00
return;
switch (platform->cst_limit) {
case CST_LIMIT_NHM:
pkg_cstate_limits = nhm_pkg_cstate_limits;
break;
case CST_LIMIT_SNB:
pkg_cstate_limits = snb_pkg_cstate_limits;
break;
case CST_LIMIT_HSW:
pkg_cstate_limits = hsw_pkg_cstate_limits;
break;
case CST_LIMIT_SKX:
pkg_cstate_limits = skx_pkg_cstate_limits;
break;
case CST_LIMIT_ICX:
pkg_cstate_limits = icx_pkg_cstate_limits;
break;
case CST_LIMIT_SLV:
pkg_cstate_limits = slv_pkg_cstate_limits;
break;
case CST_LIMIT_AMT:
pkg_cstate_limits = amt_pkg_cstate_limits;
break;
case CST_LIMIT_KNL:
pkg_cstate_limits = phi_pkg_cstate_limits;
break;
case CST_LIMIT_GMT:
pkg_cstate_limits = glm_pkg_cstate_limits;
break;
default:
return;
}
get_msr(base_cpu, MSR_PKG_CST_CONFIG_CONTROL, &msr);
pkg_cstate_limit = pkg_cstate_limits[msr & 0xF];
}
tools/power/turbostat: Rename some functions Rename dump_nhm_platform_info() and dump_nhm_cst_cfg() to dump_platform_info() and dump_cst_cfg() because these MSRs' behavior is consistent when they're available. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-03-27 14:58:35 +08:00
static void dump_platform_info(void)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
unsigned long long msr;
unsigned int ratio;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (!platform->has_nhm_msrs || no_msr)
tools/power/turbostat: Relocate pstate probing code Introduce probe_pstates() and move all pstate probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 15:58:10 +08:00
return;
tools/power turbostat: use new name for MSR_PLATFORM_INFO MSR_PLATFORM_INFO is the new name for MSR_NHM_PLATFORM_INFO no functional change Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-11-12 02:42:31 -05:00
get_msr(base_cpu, MSR_PLATFORM_INFO, &msr);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_PLATFORM_INFO: 0x%08llx\n", base_cpu, msr);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 00:01:31 -04:00
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
ratio = (msr >> 40) & 0xFF;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max efficiency frequency\n", ratio, bclk, ratio * bclk);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
ratio = (msr >> 8) & 0xFF;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz base frequency\n", ratio, bclk, ratio * bclk);
tools/power/turbostat: Relocate cstate probing code Move all cstate probing related code into probe_cstates(). Note that dump_platform_info() actually dumps both MSR_PLATFORM_INFO and MSR_IA32_POWER_CTL. MSR_PLATFORM_INFO is for pstate and MSR_IA32_POWER_CTL is for cstate. So split dump_platform_info() and dump MSR_IA32_POWER_CTL in probe_cstates(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-29 13:47:58 +08:00
}
static void dump_power_ctl(void)
{
unsigned long long msr;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (!platform->has_nhm_msrs || no_msr)
tools/power/turbostat: Relocate cstate probing code Move all cstate probing related code into probe_cstates(). Note that dump_platform_info() actually dumps both MSR_PLATFORM_INFO and MSR_IA32_POWER_CTL. MSR_PLATFORM_INFO is for pstate and MSR_IA32_POWER_CTL is for cstate. So split dump_platform_info() and dump MSR_IA32_POWER_CTL in probe_cstates(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-29 13:47:58 +08:00
return;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 08:34:28 -04:00
get_msr(base_cpu, MSR_IA32_POWER_CTL, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_IA32_POWER_CTL: 0x%08llx (C1E auto-promotion: %sabled)\n",
tools/power turbostat: cpu0 is no longer hard-coded, so update output The --debug option reads a number of per-package MSRs. Previously we explicitly read them on cpu0, but recently turbostat changed to read them on the current "base_cpu". Update the print-out to reflect base_cpu, rather than the hard-coded cpu0. Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 12:27:21 -04:00
base_cpu, msr, msr & 0x2 ? "EN" : "DIS");
tools/power turbostat: decode MSR_IA32_POWER_CTL When verbose is enabled, print the C1E-Enable bit in MSR_IA32_POWER_CTL. also delete some redundant tests on the verbose variable. Signed-off-by: Len Brown <len.brown@intel.com>
2013-01-31 15:22:15 -05:00
tools/power turbostat: Print the C-state Pre-wake settings C-state pre-wake setting[1] is an optimization for some Intel CPUs to be woken up from deep C-states in order to reduce latency. According to the spec, the BIT30 is the C-state Pre-wake Disable. Expose this setting accordingly. Sample output from turbostat: ... cpu51: MSR_IA32_POWER_CTL: 0x1a00a40059 (C1E auto-promotion: DISabled) C-state Pre-wake: ENabled cpu51: MSR_TURBO_RATIO_LIMIT: 0x2021212121212224 ... [1] https://intel.github.io/wult/#c-state-pre-wake Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-28 12:18:12 +08:00
/* C-state Pre-wake Disable (CSTATE_PREWAKE_DISABLE) */
tools/power/turbostat: Abstract cstate prewake bit support Abstract cstate prewake bit support. Delete is_icx()/is_spr() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 15:57:44 +08:00
if (platform->has_cst_prewake_bit)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "C-state Pre-wake: %sabled\n", msr & 0x40000000 ? "DIS" : "EN");
tools/power turbostat: Print the C-state Pre-wake settings C-state pre-wake setting[1] is an optimization for some Intel CPUs to be woken up from deep C-states in order to reduce latency. According to the spec, the BIT30 is the C-state Pre-wake Disable. Expose this setting accordingly. Sample output from turbostat: ... cpu51: MSR_IA32_POWER_CTL: 0x1a00a40059 (C1E auto-promotion: DISabled) C-state Pre-wake: ENabled cpu51: MSR_TURBO_RATIO_LIMIT: 0x2021212121212224 ... [1] https://intel.github.io/wult/#c-state-pre-wake Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-28 12:18:12 +08:00
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
return;
}
tools/power/turbostat: Rename some TRL functions Rename dump_hsw_turbo_ratio_limits() and dump_ivt_turbo_ratio_limits() to dump_turbo_ratio_limit2() and dump_turbo_ratio_limit1() because they dump MSR_TURBO_RATIO_LIMIT1/LIMIT2, and the MSRs' behavior is consistent when they are available. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-07-30 13:54:25 +08:00
static void dump_turbo_ratio_limit2(void)
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
{
unsigned long long msr;
unsigned int ratio;
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 08:34:28 -04:00
get_msr(base_cpu, MSR_TURBO_RATIO_LIMIT2, &msr);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_TURBO_RATIO_LIMIT2: 0x%08llx\n", base_cpu, msr);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
ratio = (msr >> 8) & 0xFF;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 18 active cores\n", ratio, bclk, ratio * bclk);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
ratio = (msr >> 0) & 0xFF;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 17 active cores\n", ratio, bclk, ratio * bclk);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
return;
}
tools/power/turbostat: Rename some TRL functions Rename dump_hsw_turbo_ratio_limits() and dump_ivt_turbo_ratio_limits() to dump_turbo_ratio_limit2() and dump_turbo_ratio_limit1() because they dump MSR_TURBO_RATIO_LIMIT1/LIMIT2, and the MSRs' behavior is consistent when they are available. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-07-30 13:54:25 +08:00
static void dump_turbo_ratio_limit1(void)
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
{
unsigned long long msr;
unsigned int ratio;
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 00:01:31 -04:00
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 08:34:28 -04:00
get_msr(base_cpu, MSR_TURBO_RATIO_LIMIT1, &msr);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 00:01:31 -04:00
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_TURBO_RATIO_LIMIT1: 0x%08llx\n", base_cpu, msr);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 00:01:31 -04:00
ratio = (msr >> 56) & 0xFF;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 16 active cores\n", ratio, bclk, ratio * bclk);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 00:01:31 -04:00
ratio = (msr >> 48) & 0xFF;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 15 active cores\n", ratio, bclk, ratio * bclk);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 00:01:31 -04:00
ratio = (msr >> 40) & 0xFF;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 14 active cores\n", ratio, bclk, ratio * bclk);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 00:01:31 -04:00
ratio = (msr >> 32) & 0xFF;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 13 active cores\n", ratio, bclk, ratio * bclk);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 00:01:31 -04:00
ratio = (msr >> 24) & 0xFF;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 12 active cores\n", ratio, bclk, ratio * bclk);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 00:01:31 -04:00
ratio = (msr >> 16) & 0xFF;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 11 active cores\n", ratio, bclk, ratio * bclk);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 00:01:31 -04:00
ratio = (msr >> 8) & 0xFF;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 10 active cores\n", ratio, bclk, ratio * bclk);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 00:01:31 -04:00
ratio = (msr >> 0) & 0xFF;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 9 active cores\n", ratio, bclk, ratio * bclk);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
return;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
static void dump_turbo_ratio_limits(int trl_msr_offset)
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
{
tools/power turbostat: fix decoding for GLM, DNV, SKX turbo-ratio limits Newer processors do not hard-code the the number of cpus in each bin to {1, 2, 3, 4, 5, 6, 7, 8} Rather, they can specify any number of CPUS in each of the 8 bins: eg. ... 37 * 100.0 = 3600.0 MHz max turbo 4 active cores 38 * 100.0 = 3700.0 MHz max turbo 3 active cores 39 * 100.0 = 3800.0 MHz max turbo 2 active cores 39 * 100.0 = 3900.0 MHz max turbo 1 active cores could now look something like this: ... 37 * 100.0 = 3600.0 MHz max turbo 16 active cores 38 * 100.0 = 3700.0 MHz max turbo 8 active cores 39 * 100.0 = 3800.0 MHz max turbo 4 active cores 39 * 100.0 = 3900.0 MHz max turbo 2 active cores Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-31 23:07:49 -05:00
unsigned long long msr, core_counts;
tools/power turbostat: simplify dump_turbo_ratio_limits() code cleanup only. no functional change. Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-31 17:21:06 -10:00
int shift;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: dump secondary Turbo-Ratio-Limit Intel Performance Hybrid processors have a 2nd MSR describing the turbo limits enforced on the Ecores. Note, TRL and Secondary-TRL are usually R/O information, but on overclock-capable parts, they can be written. Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-31 17:29:13 -10:00
get_msr(base_cpu, trl_msr_offset, &msr);
fprintf(outf, "cpu%d: MSR_%sTURBO_RATIO_LIMIT: 0x%08llx\n",
tools/power turbostat: update dump of SECONDARY_TURBO_RATIO_LIMIT cosmetic only (but useful if you copy/paste) Signed-off-by: Len Brown <len.brown@intel.com>
2022-10-13 12:42:29 +02:00
base_cpu, trl_msr_offset == MSR_SECONDARY_TURBO_RATIO_LIMIT ? "SECONDARY_" : "", msr);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 00:01:31 -04:00
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
if (platform->trl_msrs & TRL_CORECOUNT) {
tools/power turbostat: fix decoding for GLM, DNV, SKX turbo-ratio limits Newer processors do not hard-code the the number of cpus in each bin to {1, 2, 3, 4, 5, 6, 7, 8} Rather, they can specify any number of CPUS in each of the 8 bins: eg. ... 37 * 100.0 = 3600.0 MHz max turbo 4 active cores 38 * 100.0 = 3700.0 MHz max turbo 3 active cores 39 * 100.0 = 3800.0 MHz max turbo 2 active cores 39 * 100.0 = 3900.0 MHz max turbo 1 active cores could now look something like this: ... 37 * 100.0 = 3600.0 MHz max turbo 16 active cores 38 * 100.0 = 3700.0 MHz max turbo 8 active cores 39 * 100.0 = 3800.0 MHz max turbo 4 active cores 39 * 100.0 = 3900.0 MHz max turbo 2 active cores Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-31 23:07:49 -05:00
get_msr(base_cpu, MSR_TURBO_RATIO_LIMIT1, &core_counts);
fprintf(outf, "cpu%d: MSR_TURBO_RATIO_LIMIT1: 0x%08llx\n", base_cpu, core_counts);
} else {
core_counts = 0x0807060504030201;
}
tools/power turbostat: simplify dump_turbo_ratio_limits() code cleanup only. no functional change. Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-31 17:21:06 -10:00
for (shift = 56; shift >= 0; shift -= 8) {
unsigned int ratio, group_size;
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 00:01:31 -04:00
tools/power turbostat: simplify dump_turbo_ratio_limits() code cleanup only. no functional change. Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-31 17:21:06 -10:00
ratio = (msr >> shift) & 0xFF;
group_size = (core_counts >> shift) & 0xFF;
if (ratio)
fprintf(outf, "%d * %.1f = %.1f MHz max turbo %d active cores\n",
ratio, bclk, ratio * bclk, group_size);
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
return;
}
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
static void dump_atom_turbo_ratio_limits(void)
tools/power turbostat: dump Atom P-states correctly Turbostat dumps MSR_TURBO_RATIO_LIMIT on Core Architecture. But Atom Architecture uses MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-12 23:49:18 -05:00
{
unsigned long long msr;
unsigned int ratio;
get_msr(base_cpu, MSR_ATOM_CORE_RATIOS, &msr);
fprintf(outf, "cpu%d: MSR_ATOM_CORE_RATIOS: 0x%08llx\n", base_cpu, msr & 0xFFFFFFFF);
ratio = (msr >> 0) & 0x3F;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz minimum operating frequency\n", ratio, bclk, ratio * bclk);
tools/power turbostat: dump Atom P-states correctly Turbostat dumps MSR_TURBO_RATIO_LIMIT on Core Architecture. But Atom Architecture uses MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-12 23:49:18 -05:00
ratio = (msr >> 8) & 0x3F;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz low frequency mode (LFM)\n", ratio, bclk, ratio * bclk);
tools/power turbostat: dump Atom P-states correctly Turbostat dumps MSR_TURBO_RATIO_LIMIT on Core Architecture. But Atom Architecture uses MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-12 23:49:18 -05:00
ratio = (msr >> 16) & 0x3F;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz base frequency\n", ratio, bclk, ratio * bclk);
tools/power turbostat: dump Atom P-states correctly Turbostat dumps MSR_TURBO_RATIO_LIMIT on Core Architecture. But Atom Architecture uses MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-12 23:49:18 -05:00
get_msr(base_cpu, MSR_ATOM_CORE_TURBO_RATIOS, &msr);
fprintf(outf, "cpu%d: MSR_ATOM_CORE_TURBO_RATIOS: 0x%08llx\n", base_cpu, msr & 0xFFFFFFFF);
ratio = (msr >> 24) & 0x3F;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 4 active cores\n", ratio, bclk, ratio * bclk);
tools/power turbostat: dump Atom P-states correctly Turbostat dumps MSR_TURBO_RATIO_LIMIT on Core Architecture. But Atom Architecture uses MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-12 23:49:18 -05:00
ratio = (msr >> 16) & 0x3F;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 3 active cores\n", ratio, bclk, ratio * bclk);
tools/power turbostat: dump Atom P-states correctly Turbostat dumps MSR_TURBO_RATIO_LIMIT on Core Architecture. But Atom Architecture uses MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-12 23:49:18 -05:00
ratio = (msr >> 8) & 0x3F;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 2 active cores\n", ratio, bclk, ratio * bclk);
tools/power turbostat: dump Atom P-states correctly Turbostat dumps MSR_TURBO_RATIO_LIMIT on Core Architecture. But Atom Architecture uses MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-12 23:49:18 -05:00
ratio = (msr >> 0) & 0x3F;
if (ratio)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%d * %.1f = %.1f MHz max turbo 1 active core\n", ratio, bclk, ratio * bclk);
tools/power turbostat: dump Atom P-states correctly Turbostat dumps MSR_TURBO_RATIO_LIMIT on Core Architecture. But Atom Architecture uses MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-12 23:49:18 -05:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
static void dump_knl_turbo_ratio_limits(void)
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-20 09:49:34 -07:00
{
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 14:55:22 +01:00
const unsigned int buckets_no = 7;
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-20 09:49:34 -07:00
unsigned long long msr;
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 14:55:22 +01:00
int delta_cores, delta_ratio;
int i, b_nr;
unsigned int cores[buckets_no];
unsigned int ratio[buckets_no];
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-20 09:49:34 -07:00
tools/power turbostat: Replace MSR_NHM_TURBO_RATIO_LIMIT Replace MSR_NHM_TURBO_RATIO_LIMIT with MSR_TURBO_RATIO_LIMIT. Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Acked-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-07-06 16:07:56 -07:00
get_msr(base_cpu, MSR_TURBO_RATIO_LIMIT, &msr);
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-20 09:49:34 -07:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "cpu%d: MSR_TURBO_RATIO_LIMIT: 0x%08llx\n", base_cpu, msr);
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-20 09:49:34 -07:00
tools/power turbostat: unmark non-kernel-doc comment Do not mark a comment as kernel-doc notation when it is not meant to be in kernel-doc notation. Signed-off-by: Randy Dunlap <rdunlap@infradead.org> Signed-off-by: Len Brown <len.brown@intel.com>
2021-03-25 13:13:33 -07:00
/*
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-20 09:49:34 -07:00
* Turbo encoding in KNL is as follows:
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 14:55:22 +01:00
* [0] -- Reserved
* [7:1] -- Base value of number of active cores of bucket 1.
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-20 09:49:34 -07:00
* [15:8] -- Base value of freq ratio of bucket 1.
* [20:16] -- +ve delta of number of active cores of bucket 2.
* i.e. active cores of bucket 2 =
* active cores of bucket 1 + delta
* [23:21] -- Negative delta of freq ratio of bucket 2.
* i.e. freq ratio of bucket 2 =
* freq ratio of bucket 1 - delta
* [28:24]-- +ve delta of number of active cores of bucket 3.
* [31:29]-- -ve delta of freq ratio of bucket 3.
* [36:32]-- +ve delta of number of active cores of bucket 4.
* [39:37]-- -ve delta of freq ratio of bucket 4.
* [44:40]-- +ve delta of number of active cores of bucket 5.
* [47:45]-- -ve delta of freq ratio of bucket 5.
* [52:48]-- +ve delta of number of active cores of bucket 6.
* [55:53]-- -ve delta of freq ratio of bucket 6.
* [60:56]-- +ve delta of number of active cores of bucket 7.
* [63:61]-- -ve delta of freq ratio of bucket 7.
*/
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 14:55:22 +01:00
b_nr = 0;
cores[b_nr] = (msr & 0xFF) >> 1;
ratio[b_nr] = (msr >> 8) & 0xFF;
for (i = 16; i < 64; i += 8) {
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-20 09:49:34 -07:00
delta_cores = (msr >> i) & 0x1F;
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 14:55:22 +01:00
delta_ratio = (msr >> (i + 5)) & 0x7;
cores[b_nr + 1] = cores[b_nr] + delta_cores;
ratio[b_nr + 1] = ratio[b_nr] - delta_ratio;
b_nr++;
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-20 09:49:34 -07:00
}
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 14:55:22 +01:00
for (i = buckets_no - 1; i >= 0; i--)
if (i > 0 ? ratio[i] != ratio[i - 1] : 1)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf,
tools/power turbostat: add precision to --debug frequency output Add a digit of precision to the --debug output for frequency range. This is useful when BCLK is not an integer. old: 6 * 83 = 500 MHz max efficiency frequency 26 * 83 = 2166 MHz base frequency new: 6 * 83.3 = 499.8 MHz max efficiency frequency 26 * 83.3 = 2165.8 MHz base frequency Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-11 22:12:25 -05:00
"%d * %.1f = %.1f MHz max turbo %d active cores\n",
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 14:55:22 +01:00
ratio[i], bclk, ratio[i] * bclk, cores[i]);
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-20 09:49:34 -07:00
}
tools/power/turbostat: Rename some functions Rename dump_nhm_platform_info() and dump_nhm_cst_cfg() to dump_platform_info() and dump_cst_cfg() because these MSRs' behavior is consistent when they're available. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-03-27 14:58:35 +08:00
static void dump_cst_cfg(void)
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
{
unsigned long long msr;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (!platform->has_nhm_msrs || no_msr)
tools/power/turbostat: Relocate cstate probing code Move all cstate probing related code into probe_cstates(). Note that dump_platform_info() actually dumps both MSR_PLATFORM_INFO and MSR_IA32_POWER_CTL. MSR_PLATFORM_INFO is for pstate and MSR_IA32_POWER_CTL is for cstate. So split dump_platform_info() and dump MSR_IA32_POWER_CTL in probe_cstates(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-29 13:47:58 +08:00
return;
tools/power turbostat: use new name for MSR_PKG_CST_CONFIG_CONTROL Previously called MSR_NHM_SNB_PKG_CST_CFG_CTL Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-07 23:24:57 -05:00
get_msr(base_cpu, MSR_PKG_CST_CONFIG_CONTROL, &msr);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
tools/power turbostat: use new name for MSR_PKG_CST_CONFIG_CONTROL Previously called MSR_NHM_SNB_PKG_CST_CFG_CTL Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-07 23:24:57 -05:00
fprintf(outf, "cpu%d: MSR_PKG_CST_CONFIG_CONTROL: 0x%08llx", base_cpu, msr);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
tools/power turbostat: a small C-states dump readability immprovement Improve readability a little bit by changing this output: MSR_PKG_CST_CONFIG_CONTROL: 0x00008407 (locked: pkg-cstate-limit=7: unlimited, automatic-c-state-conversion=off) with this output: MSR_PKG_CST_CONFIG_CONTROL: 0x00008407 (locked, pkg-cstate-limit=7 (unlimited), automatic-c-state-conversion=off) Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-09-05 15:25:42 +03:00
fprintf(outf, " (%s%s%s%s%slocked, pkg-cstate-limit=%d (%s)",
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
(msr & SNB_C3_AUTO_UNDEMOTE) ? "UNdemote-C3, " : "",
(msr & SNB_C1_AUTO_UNDEMOTE) ? "UNdemote-C1, " : "",
(msr & NHM_C3_AUTO_DEMOTE) ? "demote-C3, " : "",
(msr & NHM_C1_AUTO_DEMOTE) ? "demote-C1, " : "",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
(msr & (1 << 15)) ? "" : "UN", (unsigned int)msr & 0xF, pkg_cstate_limit_strings[pkg_cstate_limit]);
tools/power turbostat: dump BDX, SKX automatic C-state conversion bit BDX and SKX have a bit that tells them to PROMOTE shallow C-states requests to MWAIT(C6). It is generally a BIOS bug if this bit is set. As we have encountered that BIOS bug, let's print this bit in turbostat debug output. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-09-05 15:14:08 +03:00
#define AUTOMATIC_CSTATE_CONVERSION (1UL << 16)
tools/power/turbostat: Abstract Automatic Cstate Conversion support Abstract the support for AUTOMATIC_CSTATE_CONVERSION bit in MSR_PKG_CST_CONFIG_CONTROL. Delete automatic_cstate_conversion_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 11:29:18 +08:00
if (platform->has_cst_auto_convension) {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, ", automatic c-state conversion=%s", (msr & AUTOMATIC_CSTATE_CONVERSION) ? "on" : "off");
tools/power turbostat: dump BDX, SKX automatic C-state conversion bit BDX and SKX have a bit that tells them to PROMOTE shallow C-states requests to MWAIT(C6). It is generally a BIOS bug if this bit is set. As we have encountered that BIOS bug, let's print this bit in turbostat debug output. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-09-05 15:14:08 +03:00
}
fprintf(outf, ")\n");
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
return;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
static void dump_config_tdp(void)
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
{
unsigned long long msr;
get_msr(base_cpu, MSR_CONFIG_TDP_NOMINAL, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_CONFIG_TDP_NOMINAL: 0x%08llx", base_cpu, msr);
tools/power turbostat: bugfix: TDP MSRs print bits fixing MSR_CONFIG_TDP_NOMINAL: should print all 8 bits of base_ratio (bit 0:7) 0xFF MSR_CONFIG_TDP_LEVEL_1: should print all 15 bits of PKG_MIN_PWR_LVL1 (bit 48:62) 0x7FFF should print all 15 bits of PKG_MAX_PWR_LVL1 (bit 32:46) 0x7FFF should print all 8 bits of LVL1_RATIO (bit 16:23) 0xFF should print all 15 bits of PKG_TDP_LVL1 (bit 0:14) 0x7FFF And the same modification to MSR_CONFIG_TDP_LEVEL_2. MSR_TURBO_ACTIVATION_RATIO: should print all 8 bits of MAX_NON_TURBO_RATIO (bit 0:7) 0xFF Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-13 21:09:31 +08:00
fprintf(outf, " (base_ratio=%d)\n", (unsigned int)msr & 0xFF);
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
get_msr(base_cpu, MSR_CONFIG_TDP_LEVEL_1, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_CONFIG_TDP_LEVEL_1: 0x%08llx (", base_cpu, msr);
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
if (msr) {
tools/power turbostat: bugfix: TDP MSRs print bits fixing MSR_CONFIG_TDP_NOMINAL: should print all 8 bits of base_ratio (bit 0:7) 0xFF MSR_CONFIG_TDP_LEVEL_1: should print all 15 bits of PKG_MIN_PWR_LVL1 (bit 48:62) 0x7FFF should print all 15 bits of PKG_MAX_PWR_LVL1 (bit 32:46) 0x7FFF should print all 8 bits of LVL1_RATIO (bit 16:23) 0xFF should print all 15 bits of PKG_TDP_LVL1 (bit 0:14) 0x7FFF And the same modification to MSR_CONFIG_TDP_LEVEL_2. MSR_TURBO_ACTIVATION_RATIO: should print all 8 bits of MAX_NON_TURBO_RATIO (bit 0:7) 0xFF Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-13 21:09:31 +08:00
fprintf(outf, "PKG_MIN_PWR_LVL1=%d ", (unsigned int)(msr >> 48) & 0x7FFF);
fprintf(outf, "PKG_MAX_PWR_LVL1=%d ", (unsigned int)(msr >> 32) & 0x7FFF);
fprintf(outf, "LVL1_RATIO=%d ", (unsigned int)(msr >> 16) & 0xFF);
fprintf(outf, "PKG_TDP_LVL1=%d", (unsigned int)(msr) & 0x7FFF);
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
}
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, ")\n");
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
get_msr(base_cpu, MSR_CONFIG_TDP_LEVEL_2, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_CONFIG_TDP_LEVEL_2: 0x%08llx (", base_cpu, msr);
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
if (msr) {
tools/power turbostat: bugfix: TDP MSRs print bits fixing MSR_CONFIG_TDP_NOMINAL: should print all 8 bits of base_ratio (bit 0:7) 0xFF MSR_CONFIG_TDP_LEVEL_1: should print all 15 bits of PKG_MIN_PWR_LVL1 (bit 48:62) 0x7FFF should print all 15 bits of PKG_MAX_PWR_LVL1 (bit 32:46) 0x7FFF should print all 8 bits of LVL1_RATIO (bit 16:23) 0xFF should print all 15 bits of PKG_TDP_LVL1 (bit 0:14) 0x7FFF And the same modification to MSR_CONFIG_TDP_LEVEL_2. MSR_TURBO_ACTIVATION_RATIO: should print all 8 bits of MAX_NON_TURBO_RATIO (bit 0:7) 0xFF Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-13 21:09:31 +08:00
fprintf(outf, "PKG_MIN_PWR_LVL2=%d ", (unsigned int)(msr >> 48) & 0x7FFF);
fprintf(outf, "PKG_MAX_PWR_LVL2=%d ", (unsigned int)(msr >> 32) & 0x7FFF);
fprintf(outf, "LVL2_RATIO=%d ", (unsigned int)(msr >> 16) & 0xFF);
fprintf(outf, "PKG_TDP_LVL2=%d", (unsigned int)(msr) & 0x7FFF);
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
}
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, ")\n");
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
get_msr(base_cpu, MSR_CONFIG_TDP_CONTROL, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_CONFIG_TDP_CONTROL: 0x%08llx (", base_cpu, msr);
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
if ((msr) & 0x3)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "TDP_LEVEL=%d ", (unsigned int)(msr) & 0x3);
fprintf(outf, " lock=%d", (unsigned int)(msr >> 31) & 1);
fprintf(outf, ")\n");
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
get_msr(base_cpu, MSR_TURBO_ACTIVATION_RATIO, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_TURBO_ACTIVATION_RATIO: 0x%08llx (", base_cpu, msr);
tools/power turbostat: bugfix: TDP MSRs print bits fixing MSR_CONFIG_TDP_NOMINAL: should print all 8 bits of base_ratio (bit 0:7) 0xFF MSR_CONFIG_TDP_LEVEL_1: should print all 15 bits of PKG_MIN_PWR_LVL1 (bit 48:62) 0x7FFF should print all 15 bits of PKG_MAX_PWR_LVL1 (bit 32:46) 0x7FFF should print all 8 bits of LVL1_RATIO (bit 16:23) 0xFF should print all 15 bits of PKG_TDP_LVL1 (bit 0:14) 0x7FFF And the same modification to MSR_CONFIG_TDP_LEVEL_2. MSR_TURBO_ACTIVATION_RATIO: should print all 8 bits of MAX_NON_TURBO_RATIO (bit 0:7) 0xFF Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-13 21:09:31 +08:00
fprintf(outf, "MAX_NON_TURBO_RATIO=%d", (unsigned int)(msr) & 0xFF);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, " lock=%d", (unsigned int)(msr >> 31) & 1);
fprintf(outf, ")\n");
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
}
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-06 17:15:55 -04:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
unsigned int irtl_time_units[] = { 1, 32, 1024, 32768, 1048576, 33554432, 0, 0 };
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-06 17:15:55 -04:00
void print_irtl(void)
{
unsigned long long msr;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (!platform->has_irtl_msrs || no_msr)
tools/power/turbostat: Abstract IRTL support Abstract the support for MSR_PKGC3/PKGC6/PKGC7/PKGC8/PKGC9/PKGC10_IRTL. Delete has_snb_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:18:51 +08:00
return;
tools/power/turbostat: Use fine grained IRTL output It is pointless to dump the IRTL register for a package cstate that is not supported by the platform. Print IRTL only for states that are available in platform->supported_cstates. Delete has_c8910_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:27:26 +08:00
if (platform->supported_cstates & PC3) {
get_msr(base_cpu, MSR_PKGC3_IRTL, &msr);
fprintf(outf, "cpu%d: MSR_PKGC3_IRTL: 0x%08llx (", base_cpu, msr);
fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT",
(msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]);
}
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-06 17:15:55 -04:00
tools/power/turbostat: Use fine grained IRTL output It is pointless to dump the IRTL register for a package cstate that is not supported by the platform. Print IRTL only for states that are available in platform->supported_cstates. Delete has_c8910_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:27:26 +08:00
if (platform->supported_cstates & PC6) {
get_msr(base_cpu, MSR_PKGC6_IRTL, &msr);
fprintf(outf, "cpu%d: MSR_PKGC6_IRTL: 0x%08llx (", base_cpu, msr);
fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT",
(msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]);
}
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-06 17:15:55 -04:00
tools/power/turbostat: Use fine grained IRTL output It is pointless to dump the IRTL register for a package cstate that is not supported by the platform. Print IRTL only for states that are available in platform->supported_cstates. Delete has_c8910_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:27:26 +08:00
if (platform->supported_cstates & PC7) {
get_msr(base_cpu, MSR_PKGC7_IRTL, &msr);
fprintf(outf, "cpu%d: MSR_PKGC7_IRTL: 0x%08llx (", base_cpu, msr);
fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT",
(msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]);
}
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-06 17:15:55 -04:00
tools/power/turbostat: Use fine grained IRTL output It is pointless to dump the IRTL register for a package cstate that is not supported by the platform. Print IRTL only for states that are available in platform->supported_cstates. Delete has_c8910_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:27:26 +08:00
if (platform->supported_cstates & PC8) {
get_msr(base_cpu, MSR_PKGC8_IRTL, &msr);
fprintf(outf, "cpu%d: MSR_PKGC8_IRTL: 0x%08llx (", base_cpu, msr);
fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT",
(msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]);
}
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-06 17:15:55 -04:00
tools/power/turbostat: Use fine grained IRTL output It is pointless to dump the IRTL register for a package cstate that is not supported by the platform. Print IRTL only for states that are available in platform->supported_cstates. Delete has_c8910_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:27:26 +08:00
if (platform->supported_cstates & PC9) {
get_msr(base_cpu, MSR_PKGC9_IRTL, &msr);
fprintf(outf, "cpu%d: MSR_PKGC9_IRTL: 0x%08llx (", base_cpu, msr);
fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT",
(msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]);
}
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-06 17:15:55 -04:00
tools/power/turbostat: Use fine grained IRTL output It is pointless to dump the IRTL register for a package cstate that is not supported by the platform. Print IRTL only for states that are available in platform->supported_cstates. Delete has_c8910_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:27:26 +08:00
if (platform->supported_cstates & PC10) {
get_msr(base_cpu, MSR_PKGC10_IRTL, &msr);
fprintf(outf, "cpu%d: MSR_PKGC10_IRTL: 0x%08llx (", base_cpu, msr);
fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT",
(msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]);
}
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-06 17:15:55 -04:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
void free_fd_percpu(void)
{
int i;
tools/power turbostat: Add proper re-initialization for perf file descriptors Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-02-15 12:50:19 +01:00
if (!fd_percpu)
return;
tools/power turbostat: Allocate correct amount of fd and irq entries The tool uses topo.max_cpu_num to determine number of entries needed for fd_percpu[] and irqs_per_cpu[]. For example on a system with 4 CPUs topo.max_cpu_num is 3 so we get too small array for holding per-CPU items. Fix this to use right number of entries, which is topo.max_cpu_num + 1. Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 11:13:23 +03:00
for (i = 0; i < topo.max_cpu_num + 1; ++i) {
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
if (fd_percpu[i] != 0)
close(fd_percpu[i]);
}
free(fd_percpu);
tools/power turbostat: Add proper re-initialization for perf file descriptors Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-02-15 12:50:19 +01:00
fd_percpu = NULL;
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
}
tools/power turbostat: Add proper re-initialization for perf file descriptors Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-02-15 12:50:19 +01:00
void free_fd_instr_count_percpu(void)
{
if (!fd_instr_count_percpu)
return;
for (int i = 0; i < topo.max_cpu_num + 1; ++i) {
if (fd_instr_count_percpu[i] != 0)
close(fd_instr_count_percpu[i]);
}
free(fd_instr_count_percpu);
fd_instr_count_percpu = NULL;
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
}
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
void free_fd_cstate(void)
{
if (!ccstate_counter_info)
return;
const int counter_info_num = ccstate_counter_info_size;
for (int counter_id = 0; counter_id < counter_info_num; ++counter_id) {
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
if (ccstate_counter_info[counter_id].fd_perf_core != -1)
close(ccstate_counter_info[counter_id].fd_perf_core);
if (ccstate_counter_info[counter_id].fd_perf_pkg != -1)
close(ccstate_counter_info[counter_id].fd_perf_pkg);
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
}
free(ccstate_counter_info);
ccstate_counter_info = NULL;
ccstate_counter_info_size = 0;
}
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
void free_fd_msr(void)
{
if (!msr_counter_info)
return;
for (int cpu = 0; cpu < topo.max_cpu_num; ++cpu) {
if (msr_counter_info[cpu].fd_perf != -1)
close(msr_counter_info[cpu].fd_perf);
}
free(msr_counter_info);
msr_counter_info = NULL;
msr_counter_info_size = 0;
}
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
void free_fd_rapl_percpu(void)
{
if (!rapl_counter_info_perdomain)
return;
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs Save the highest core and package id when parsing topology to allocate enough memory when get_rapl_counters() is called with a core or a package id as a domain. Note that RAPL domains are per-package on Intel, but per-core on AMD. Thus, the RAPL code effectively runs in different modes on those two product lines. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-06 15:39:08 +02:00
const int num_domains = rapl_counter_info_perdomain_size;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
for (int domain_id = 0; domain_id < num_domains; ++domain_id) {
if (rapl_counter_info_perdomain[domain_id].fd_perf != -1)
close(rapl_counter_info_perdomain[domain_id].fd_perf);
}
free(rapl_counter_info_perdomain);
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs Save the highest core and package id when parsing topology to allocate enough memory when get_rapl_counters() is called with a core or a package id as a domain. Note that RAPL domains are per-package on Intel, but per-core on AMD. Thus, the RAPL code effectively runs in different modes on those two product lines. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-06 15:39:08 +02:00
rapl_counter_info_perdomain = NULL;
rapl_counter_info_perdomain_size = 0;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
void free_fd_added_perf_counters_(struct perf_counter_info *pp)
{
if (!pp)
return;
if (!pp->fd_perf_per_domain)
return;
while (pp) {
for (size_t domain = 0; domain < pp->num_domains; ++domain) {
if (pp->fd_perf_per_domain[domain] != -1) {
close(pp->fd_perf_per_domain[domain]);
pp->fd_perf_per_domain[domain] = -1;
}
}
free(pp->fd_perf_per_domain);
pp->fd_perf_per_domain = NULL;
pp = pp->next;
}
}
void free_fd_added_perf_counters(void)
{
free_fd_added_perf_counters_(sys.perf_tp);
free_fd_added_perf_counters_(sys.perf_cp);
free_fd_added_perf_counters_(sys.perf_pp);
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
void free_all_buffers(void)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
int i;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
CPU_FREE(cpu_present_set);
cpu_present_set = NULL;
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
cpu_present_setsize = 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power/turbostat: Handle cgroup v2 cpu limitation CPUs can be isolated via cgroup settings and turbostat should avoid migrating to these CPUs, just like it does for the '-c' cpus. Introduce cpu_effective_set to save the cgroup cpu limitation info from /sys/fs/cgroup/cpuset.cpus.effective. And use cpu_allowed_set as the intersection of cpu_present_set, cpu_effective_set and cpu_subset. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:39:22 +08:00
CPU_FREE(cpu_effective_set);
cpu_effective_set = NULL;
cpu_effective_setsize = 0;
tools/power/turbostat: Introduce cpu_allowed_set Turbostat supports "-c" parameter which limits output to system summary plus the specified cpu-set. But some code still uses cpu_present_set to read and dump the counters. Introduce cpu_allowed_set for code that should obey the specified cpu-set. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 11:01:17 +08:00
CPU_FREE(cpu_allowed_set);
cpu_allowed_set = NULL;
cpu_allowed_setsize = 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
CPU_FREE(cpu_affinity_set);
cpu_affinity_set = NULL;
cpu_affinity_setsize = 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
free(thread_even);
free(core_even);
free(package_even);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
thread_even = NULL;
core_even = NULL;
package_even = NULL;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
free(thread_odd);
free(core_odd);
free(package_odd);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
thread_odd = NULL;
core_odd = NULL;
package_odd = NULL;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
free(output_buffer);
output_buffer = NULL;
outp = NULL;
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
free_fd_percpu();
tools/power turbostat: Add proper re-initialization for perf file descriptors Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-02-15 12:50:19 +01:00
free_fd_instr_count_percpu();
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
free_fd_msr();
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
free_fd_rapl_percpu();
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
free_fd_cstate();
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
free_fd_added_perf_counters();
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
free(irq_column_2_cpu);
free(irqs_per_cpu);
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
for (i = 0; i <= topo.max_cpu_num; ++i) {
if (cpus[i].put_ids)
CPU_FREE(cpus[i].put_ids);
}
free(cpus);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/*
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:16 -07:00
* Parse a file containing a single int.
tools/power turbostat: Add Die column If the system has more than one software visible die per package, print a Die column. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-14 19:17:40 -08:00
* Return 0 if file can not be opened
* Exit if file can be opened, but can not be parsed
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
*/
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:16 -07:00
int parse_int_file(const char *fmt, ...)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:16 -07:00
va_list args;
char path[PATH_MAX];
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
FILE *filep;
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:16 -07:00
int value;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:16 -07:00
va_start(args, fmt);
vsnprintf(path, sizeof(path), fmt, args);
va_end(args);
tools/power turbostat: Add Die column If the system has more than one software visible die per package, print a Die column. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-14 19:17:40 -08:00
filep = fopen(path, "r");
if (!filep)
return 0;
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:18 -07:00
if (fscanf(filep, "%d", &value) != 1)
err(1, "%s: failed to parse number from file", path);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
fclose(filep);
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:16 -07:00
return value;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/*
* cpu_is_first_core_in_package(cpu)
* return 1 if given CPU is 1st core in package
*/
int cpu_is_first_core_in_package(int cpu)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:16 -07:00
return cpu == parse_int_file("/sys/devices/system/cpu/cpu%d/topology/core_siblings_list", cpu);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
int get_physical_package_id(int cpu)
{
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:16 -07:00
return parse_int_file("/sys/devices/system/cpu/cpu%d/topology/physical_package_id", cpu);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: Add Die column If the system has more than one software visible die per package, print a Die column. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-14 19:17:40 -08:00
int get_die_id(int cpu)
{
return parse_int_file("/sys/devices/system/cpu/cpu%d/topology/die_id", cpu);
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
int get_core_id(int cpu)
{
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:16 -07:00
return parse_int_file("/sys/devices/system/cpu/cpu%d/topology/core_id", cpu);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: Calculate additional node information for a package The code currently assumes each package has exactly one node. This is not the case for AMD systems and Intel systems with COD. AMD systems also may re-enumerate each node's core IDs starting at 0 (for example, an AMD processor may have two nodes, each with core IDs from 0 to 7). In order to properly enumerate the cores we need to track both the physical and logical node IDs. Add physical_node_id to track the node ID assigned by the kernel, and logical_node_id used by turbostat to track the nodes per package ie) a 0-based count within the package. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:30 -04:00
void set_node_data(void)
{
tools/power turbostat: Fix logical node enumeration to allow for non-sequential physical nodes turbostat fails on some multi-package topologies because the logical node enumeration assumes that the nodes are sequentially numbered, which causes the logical numa nodes to not be enumerated, or enumerated incorrectly. Use a more robust enumeration algorithm which allows for non-seqential physical nodes. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-07-26 09:08:54 -04:00
int pkg, node, lnode, cpu, cpux;
int cpu_count;
/* initialize logical_node_id */
for (cpu = 0; cpu <= topo.max_cpu_num; ++cpu)
cpus[cpu].logical_node_id = -1;
cpu_count = 0;
for (pkg = 0; pkg < topo.num_packages; pkg++) {
lnode = 0;
for (cpu = 0; cpu <= topo.max_cpu_num; ++cpu) {
if (cpus[cpu].physical_package_id != pkg)
continue;
/* find a cpu with an unset logical_node_id */
if (cpus[cpu].logical_node_id != -1)
continue;
cpus[cpu].logical_node_id = lnode;
node = cpus[cpu].physical_node_id;
cpu_count++;
/*
* find all matching cpus on this pkg and set
* the logical_node_id
*/
for (cpux = cpu; cpux <= topo.max_cpu_num; cpux++) {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
if ((cpus[cpux].physical_package_id == pkg) && (cpus[cpux].physical_node_id == node)) {
tools/power turbostat: Fix logical node enumeration to allow for non-sequential physical nodes turbostat fails on some multi-package topologies because the logical node enumeration assumes that the nodes are sequentially numbered, which causes the logical numa nodes to not be enumerated, or enumerated incorrectly. Use a more robust enumeration algorithm which allows for non-seqential physical nodes. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-07-26 09:08:54 -04:00
cpus[cpux].logical_node_id = lnode;
cpu_count++;
}
}
lnode++;
if (lnode > topo.nodes_per_pkg)
topo.nodes_per_pkg = lnode;
}
if (cpu_count >= topo.max_cpu_num)
break;
tools/power turbostat: Calculate additional node information for a package The code currently assumes each package has exactly one node. This is not the case for AMD systems and Intel systems with COD. AMD systems also may re-enumerate each node's core IDs starting at 0 (for example, an AMD processor may have two nodes, each with core IDs from 0 to 7). In order to properly enumerate the cores we need to track both the physical and logical node IDs. Add physical_node_id to track the node ID assigned by the kernel, and logical_node_id used by turbostat to track the nodes per package ie) a 0-based count within the package. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:30 -04:00
}
}
int get_physical_node_id(struct cpu_topology *thiscpu)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
{
char path[80];
FILE *filep;
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
int i;
int cpu = thiscpu->logical_cpu_id;
tools/power turbostat: correctly display more than 2 threads/core Without this update, turbostat displays only 2 threads per core. Some processors, such as Xeon Phi, have more. Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-15 10:09:50 -07:00
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
for (i = 0; i <= topo.max_cpu_num; i++) {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
sprintf(path, "/sys/devices/system/cpu/cpu%d/node%i/cpulist", cpu, i);
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
filep = fopen(path, "r");
if (!filep)
continue;
fclose(filep);
return i;
tools/power turbostat: correctly display more than 2 threads/core Without this update, turbostat displays only 2 threads per core. Some processors, such as Xeon Phi, have more. Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-15 10:09:50 -07:00
}
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
return -1;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power/turbostat: Abstrct function for parsing cpu string Abstract parse_cpu_str() which can update any specified cpu_set by a given cpu string. This can be used to handle further CPU limitations from other sources like cgroup. The cpu string parsing code is also enhanced to handle the strings that have an extra '\n' before string terminator. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:45:21 +08:00
static int parse_cpu_str(char *cpu_str, cpu_set_t *cpu_set, int cpu_set_size)
{
unsigned int start, end;
char *next = cpu_str;
while (next && *next) {
if (*next == '-') /* no negative cpu numbers */
return 1;
tools/power turbostat: Fix trailing '\n' parsing parse_cpu_string() parses the string input either from command line or from /sys/fs/cgroup/cpuset.cpus.effective to get a list of CPUs that turbostat can run with. The cpu string returned by /sys/fs/cgroup/cpuset.cpus.effective contains a trailing '\n', but strtoul() fails to treat this as an error. That says, for the code below val = ("\n", NULL, 10); val returns 0, and errno is also not set. As a result, CPU0 is erroneously considered as allowed CPU and this causes failures when turbostat tries to run on CPU0. get_counters: Could not migrate to CPU 0 ... turbostat: re-initialized with num_cpus 8, allowed_cpus 5 get_counters: Could not migrate to CPU 0 Add a check to return immediately if '\n' or '\0' is detected. Fixes: 8c3dd2c9e542 ("tools/power/turbostat: Abstrct function for parsing cpu string") Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-27 13:07:51 +08:00
if (*next == '\0' || *next == '\n')
break;
tools/power/turbostat: Abstrct function for parsing cpu string Abstract parse_cpu_str() which can update any specified cpu_set by a given cpu string. This can be used to handle further CPU limitations from other sources like cgroup. The cpu string parsing code is also enhanced to handle the strings that have an extra '\n' before string terminator. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:45:21 +08:00
start = strtoul(next, &next, 10);
if (start >= CPU_SUBSET_MAXCPUS)
return 1;
CPU_SET_S(start, cpu_set_size, cpu_set);
if (*next == '\0' || *next == '\n')
break;
if (*next == ',') {
next += 1;
continue;
}
if (*next == '-') {
next += 1; /* start range */
} else if (*next == '.') {
next += 1;
if (*next == '.')
next += 1; /* start range */
else
return 1;
}
end = strtoul(next, &next, 10);
if (end <= start)
return 1;
while (++start <= end) {
if (start >= CPU_SUBSET_MAXCPUS)
return 1;
CPU_SET_S(start, cpu_set_size, cpu_set);
}
if (*next == ',')
next += 1;
else if (*next != '\0' && *next != '\n')
return 1;
}
return 0;
}
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
int get_thread_siblings(struct cpu_topology *thiscpu)
{
char path[80], character;
FILE *filep;
unsigned long map;
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
int so, shift, sib_core;
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
int cpu = thiscpu->logical_cpu_id;
int offset = topo.max_cpu_num + 1;
size_t size;
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
int thread_id = 0;
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
thiscpu->put_ids = CPU_ALLOC((topo.max_cpu_num + 1));
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
if (thiscpu->thread_id < 0)
thiscpu->thread_id = thread_id++;
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
if (!thiscpu->put_ids)
return -1;
size = CPU_ALLOC_SIZE((topo.max_cpu_num + 1));
CPU_ZERO_S(size, thiscpu->put_ids);
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
sprintf(path, "/sys/devices/system/cpu/cpu%d/topology/thread_siblings", cpu);
tools/power turbostat: harden against cpu hotplug turbostat tends to get confused when CPUs are added and removed while it is running. There are races, such as checking the current cpu, and then reading a sysfs file that depends on that cpu number. Close the two issues that seem to come up the most. First, there is an infinite reset loop detector -- change that to allow more resets before giving up. Secondly, one of those file reads didn't really need to exit the program on failure... Signed-off-by: Len Brown <len.brown@intel.com>
2020-09-30 20:58:15 -04:00
filep = fopen(path, "r");
if (!filep) {
warnx("%s: open failed", path);
return -1;
}
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
do {
offset -= BITMASK_SIZE;
tools/power turbostat: Add checks for failure of fgets() and fscanf() Most calls to fgets() and fscanf() are followed by error checks. Add an exit-on-error in the remaining cases. Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-20 23:01:03 -04:00
if (fscanf(filep, "%lx%c", &map, &character) != 2)
err(1, "%s: failed to parse file", path);
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
for (shift = 0; shift < BITMASK_SIZE; shift++) {
if ((map >> shift) & 0x1) {
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
so = shift + offset;
sib_core = get_core_id(so);
if (sib_core == thiscpu->physical_core_id) {
CPU_SET_S(so, size, thiscpu->put_ids);
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
if ((so != cpu) && (cpus[so].thread_id < 0))
cpus[so].thread_id = thread_id++;
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
}
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
}
}
tools/power turbostat: replace strncmp with single character compare Using strncmp for a single character comparison is overly complicated, just use a simpler single character comparison instead. Also stops static analyzers (such as cppcheck) from complaining about strncmp on non-null terminated strings. Signed-off-by: Colin Ian King <colin.i.king@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-26 14:10:24 +01:00
} while (character == ',');
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
fclose(filep);
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
return CPU_COUNT_S(size, thiscpu->put_ids);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
/*
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
* run func(thread, core, package) in topology order
* skip non-present cpus
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
*/
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int for_all_cpus_2(int (func) (struct thread_data *, struct core_data *,
struct pkg_data *, struct thread_data *, struct core_data *,
struct pkg_data *), struct thread_data *thread_base,
struct core_data *core_base, struct pkg_data *pkg_base,
struct thread_data *thread_base2, struct core_data *core_base2, struct pkg_data *pkg_base2)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
{
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
int retval, pkg_no, node_no, core_no, thread_no;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
for (pkg_no = 0; pkg_no < topo.num_packages; ++pkg_no) {
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
for (node_no = 0; node_no < topo.nodes_per_pkg; ++node_no) {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
for (core_no = 0; core_no < topo.cores_per_node; ++core_no) {
for (thread_no = 0; thread_no < topo.threads_per_core; ++thread_no) {
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
struct thread_data *t, *t2;
struct core_data *c, *c2;
struct pkg_data *p, *p2;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
t = GET_THREAD(thread_base, thread_no, core_no, node_no, pkg_no);
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
tools/power/turbostat: Obey allowed CPUs when accessing CPU counters for_all_cpus/for_all_cpus_2 are used for accessing the per CPU counters, and they should follow the cpu_allowed_set instead of cpu_present_set. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 13:52:02 +08:00
if (cpu_is_not_allowed(t->cpu_id))
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
continue;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
t2 = GET_THREAD(thread_base2, thread_no, core_no, node_no, pkg_no);
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
c = GET_CORE(core_base, core_no, node_no, pkg_no);
c2 = GET_CORE(core_base2, core_no, node_no, pkg_no);
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
p = GET_PKG(pkg_base, pkg_no);
p2 = GET_PKG(pkg_base2, pkg_no);
retval = func(t, c, p, t2, c2, p2);
if (retval)
return retval;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
}
}
return 0;
}
/*
* run func(cpu) on every cpu in /proc/stat
* return max_cpu number
*/
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int for_all_proc_cpus(int (func) (int))
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
FILE *fp;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
int cpu_num;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
int retval;
turbostat: Factor out common function to open file and exit on failure Several different functions in turbostat contain the same pattern of opening a file and exiting on failure. Factor out a common fopen_or_die function for that. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:17 -07:00
fp = fopen_or_die(proc_stat, "r");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
retval = fscanf(fp, "cpu %*d %*d %*d %*d %*d %*d %*d %*d %*d %*d\n");
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:18 -07:00
if (retval != 0)
err(1, "%s: failed to parse format", proc_stat);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
while (1) {
retval = fscanf(fp, "cpu%u %*d %*d %*d %*d %*d %*d %*d %*d %*d %*d\n", &cpu_num);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
if (retval != 1)
break;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
retval = func(cpu_num);
if (retval) {
fclose(fp);
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
return (retval);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
fclose(fp);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power/turbostat: Handle cgroup v2 cpu limitation CPUs can be isolated via cgroup settings and turbostat should avoid migrating to these CPUs, just like it does for the '-c' cpus. Introduce cpu_effective_set to save the cgroup cpu limitation info from /sys/fs/cgroup/cpuset.cpus.effective. And use cpu_allowed_set as the intersection of cpu_present_set, cpu_effective_set and cpu_subset. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:39:22 +08:00
#define PATH_EFFECTIVE_CPUS "/sys/fs/cgroup/cpuset.cpus.effective"
static char cpu_effective_str[1024];
static int update_effective_str(bool startup)
{
FILE *fp;
char *pos;
char buf[1024];
int ret;
if (cpu_effective_str[0] == '\0' && !startup)
return 0;
fp = fopen(PATH_EFFECTIVE_CPUS, "r");
if (!fp)
return 0;
pos = fgets(buf, 1024, fp);
if (!pos)
err(1, "%s: file read failed\n", PATH_EFFECTIVE_CPUS);
fclose(fp);
ret = strncmp(cpu_effective_str, buf, 1024);
if (!ret)
return 0;
strncpy(cpu_effective_str, buf, 1024);
return 1;
}
static void update_effective_set(bool startup)
{
update_effective_str(startup);
if (parse_cpu_str(cpu_effective_str, cpu_effective_set, cpu_effective_setsize))
err(1, "%s: cpu str malformat %s\n", PATH_EFFECTIVE_CPUS, cpu_effective_str);
}
tools/power turbostat: Add proper re-initialization for perf file descriptors Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-02-15 12:50:19 +01:00
void linux_perf_init(void);
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
void msr_perf_init(void);
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
void rapl_perf_init(void);
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
void cstate_perf_init(void);
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
void added_perf_counters_init(void);
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
void pmt_init(void);
tools/power turbostat: Add proper re-initialization for perf file descriptors Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-02-15 12:50:19 +01:00
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
void re_initialize(void)
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
free_all_buffers();
tools/power/turbostat: Handle offlined CPUs in cpu_subset It is possible that the cpu_subset contains offlined CPUs. If this happens during start, exit immediately because this is likely an operator error that is best fixed by re-invoking. If this happens at runtime, give a warning only because turbostat should do its best effort to continue running. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 14:20:47 +08:00
setup_all_buffers(false);
tools/power turbostat: Add proper re-initialization for perf file descriptors Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-02-15 12:50:19 +01:00
linux_perf_init();
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
msr_perf_init();
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rapl_perf_init();
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
cstate_perf_init();
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
added_perf_counters_init();
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
pmt_init();
tools/power turbostat: Do not print negative LPI residency turbostat prints the abnormal SYS%LPI across suspend-to-idle: SYS%LPI = 114479815993277.50 This is reproduced by: Run a freeze cycle, e.g. "sleepgraph -m freeze -rtcwake 15". Then do a reboot. After boot up, launch the suspend-idle-idle and check the SYS%LPI field. The slp_so residence counter is in LPIT table, and BIOS does not clears this register across reset. The PMC expects the OS to calculate the LPI residency based on the delta. However, there is an firmware issue that the LPIT gets cleared to 0 during the second suspend to idle after the reboot, which brings negative delta value. [lenb: updated to print "neg" upon this BIOS failure] Reported-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2023-10-22 13:52:21 +08:00
fprintf(outf, "turbostat: re-initialized with num_cpus %d, allowed_cpus %d\n", topo.num_cpus,
topo.allowed_cpus);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: set max_num_cpus equal to the cpumask length Future fixes will use sysfs files that contain cpumask output. The code needs to know the length of the cpumask in order to determine which cpus are set in a cpumask. Currently topo.max_cpu_num is the maximum cpu number. It can be increased the the maximum value of cpus represented in cpumasks. Set max_num_cpus to the length of a cpumask. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:28 -04:00
void set_max_cpu_num(void)
{
FILE *filep;
tools/power turbostat: Use sched_getcpu() instead of hardcoded cpu 0 Disabling cpu 0 results in an error turbostat: /sys/devices/system/cpu/cpu0/topology/thread_siblings: open failed: No such file or directory Use sched_getcpu() instead of a hardcoded cpu 0 to get the max cpu number. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-06-29 15:26:57 -04:00
int base_cpu;
tools/power turbostat: set max_num_cpus equal to the cpumask length Future fixes will use sysfs files that contain cpumask output. The code needs to know the length of the cpumask in order to determine which cpus are set in a cpumask. Currently topo.max_cpu_num is the maximum cpu number. It can be increased the the maximum value of cpus represented in cpumasks. Set max_num_cpus to the length of a cpumask. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:28 -04:00
unsigned long dummy;
tools/power turbostat: Use sched_getcpu() instead of hardcoded cpu 0 Disabling cpu 0 results in an error turbostat: /sys/devices/system/cpu/cpu0/topology/thread_siblings: open failed: No such file or directory Use sched_getcpu() instead of a hardcoded cpu 0 to get the max cpu number. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-06-29 15:26:57 -04:00
char pathname[64];
tools/power turbostat: set max_num_cpus equal to the cpumask length Future fixes will use sysfs files that contain cpumask output. The code needs to know the length of the cpumask in order to determine which cpus are set in a cpumask. Currently topo.max_cpu_num is the maximum cpu number. It can be increased the the maximum value of cpus represented in cpumasks. Set max_num_cpus to the length of a cpumask. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:28 -04:00
tools/power turbostat: Use sched_getcpu() instead of hardcoded cpu 0 Disabling cpu 0 results in an error turbostat: /sys/devices/system/cpu/cpu0/topology/thread_siblings: open failed: No such file or directory Use sched_getcpu() instead of a hardcoded cpu 0 to get the max cpu number. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-06-29 15:26:57 -04:00
base_cpu = sched_getcpu();
if (base_cpu < 0)
err(1, "cannot find calling cpu ID");
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
sprintf(pathname, "/sys/devices/system/cpu/cpu%d/topology/thread_siblings", base_cpu);
tools/power turbostat: Use sched_getcpu() instead of hardcoded cpu 0 Disabling cpu 0 results in an error turbostat: /sys/devices/system/cpu/cpu0/topology/thread_siblings: open failed: No such file or directory Use sched_getcpu() instead of a hardcoded cpu 0 to get the max cpu number. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-06-29 15:26:57 -04:00
filep = fopen_or_die(pathname, "r");
tools/power turbostat: set max_num_cpus equal to the cpumask length Future fixes will use sysfs files that contain cpumask output. The code needs to know the length of the cpumask in order to determine which cpus are set in a cpumask. Currently topo.max_cpu_num is the maximum cpu number. It can be increased the the maximum value of cpus represented in cpumasks. Set max_num_cpus to the length of a cpumask. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:28 -04:00
topo.max_cpu_num = 0;
while (fscanf(filep, "%lx,", &dummy) == 1)
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
topo.max_cpu_num += BITMASK_SIZE;
tools/power turbostat: set max_num_cpus equal to the cpumask length Future fixes will use sysfs files that contain cpumask output. The code needs to know the length of the cpumask in order to determine which cpus are set in a cpumask. Currently topo.max_cpu_num is the maximum cpu number. It can be increased the the maximum value of cpus represented in cpumasks. Set max_num_cpus to the length of a cpumask. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:28 -04:00
fclose(filep);
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
topo.max_cpu_num--; /* 0 based */
tools/power turbostat: set max_num_cpus equal to the cpumask length Future fixes will use sysfs files that contain cpumask output. The code needs to know the length of the cpumask in order to determine which cpus are set in a cpumask. Currently topo.max_cpu_num is the maximum cpu number. It can be increased the the maximum value of cpus represented in cpumasks. Set max_num_cpus to the length of a cpumask. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:28 -04:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
/*
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
* count_cpus()
* remember the last one seen, it will be the max
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
*/
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
int count_cpus(int cpu)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
tools/power turbostat: No build warnings with -Wextra Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 23:45:18 -04:00
UNUSED(cpu);
tools/power turbostat: set max_num_cpus equal to the cpumask length Future fixes will use sysfs files that contain cpumask output. The code needs to know the length of the cpumask in order to determine which cpus are set in a cpumask. Currently topo.max_cpu_num is the maximum cpu number. It can be increased the the maximum value of cpus represented in cpumasks. Set max_num_cpus to the length of a cpumask. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:28 -04:00
topo.num_cpus++;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return 0;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
int mark_cpu_present(int cpu)
{
CPU_SET_S(cpu, cpu_present_setsize, cpu_present_set);
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 22:19:58 -04:00
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
int init_thread_id(int cpu)
{
cpus[cpu].thread_id = -1;
return 0;
}
tools/power turbostat: Allow using cpu device in perf counters on hybrid platforms Intel hybrid platforms expose different perf devices for P and E cores. Instead of one, "/sys/bus/event_source/devices/cpu" device, there are "/sys/bus/event_source/devices/{cpu_core,cpu_atom}". This, however makes it more complicated for the user, because most of the counters are available on both and had to be handled manually. This patch allows users to use "virtual" cpu device that is seemingly translated to cpu_core and cpu_atom perf devices, depending on the type of a CPU we are opening the counter for. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-20 18:47:59 +02:00
int set_my_cpu_type(void)
{
unsigned int eax, ebx, ecx, edx;
unsigned int max_level;
__cpuid(0, max_level, ebx, ecx, edx);
if (max_level < CPUID_LEAF_MODEL_ID)
return 0;
__cpuid(CPUID_LEAF_MODEL_ID, eax, ebx, ecx, edx);
return (eax >> CPUID_LEAF_MODEL_ID_CORE_TYPE_SHIFT);
}
int set_cpu_hybrid_type(int cpu)
{
if (cpu_migrate(cpu))
return -1;
int type = set_my_cpu_type();
cpus[cpu].type = type;
return 0;
}
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
/*
* snapshot_proc_interrupts()
*
* read and record summary of /proc/interrupts
*
* return 1 if config change requires a restart, else return 0
*/
int snapshot_proc_interrupts(void)
{
static FILE *fp;
int column, retval;
if (fp == NULL)
fp = fopen_or_die("/proc/interrupts", "r");
else
rewind(fp);
/* read 1st line of /proc/interrupts to get cpu* name for each column */
for (column = 0; column < topo.num_cpus; ++column) {
int cpu_number;
retval = fscanf(fp, " CPU%d", &cpu_number);
if (retval != 1)
break;
if (cpu_number > topo.max_cpu_num) {
warn("/proc/interrupts: cpu%d: > %d", cpu_number, topo.max_cpu_num);
return 1;
}
irq_column_2_cpu[column] = cpu_number;
irqs_per_cpu[cpu_number] = 0;
}
/* read /proc/interrupt count lines and sum up irqs per cpu */
while (1) {
int column;
char buf[64];
retval = fscanf(fp, " %s:", buf); /* flush irq# "N:" */
if (retval != 1)
break;
/* read the count per cpu */
for (column = 0; column < topo.num_cpus; ++column) {
int cpu_number, irq_count;
retval = fscanf(fp, " %d", &irq_count);
if (retval != 1)
break;
cpu_number = irq_column_2_cpu[column];
irqs_per_cpu[cpu_number] += irq_count;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
while (getc(fp) != '\n') ; /* flush interrupt description */
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
}
return 0;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 01:28:12 -05:00
/*
tools/power/turbostat: Unify graphics sysfs snapshots Graphics sysfs snapshots share similar logic. Combine them into one function to avoid code duplication. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-13 10:12:19 +08:00
* snapshot_graphics()
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 01:28:12 -05:00
*
tools/power/turbostat: Unify graphics sysfs snapshots Graphics sysfs snapshots share similar logic. Combine them into one function to avoid code duplication. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-13 10:12:19 +08:00
* record snapshot of specified graphics sysfs knob
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 01:28:12 -05:00
*
* return 1 if config change requires a restart, else return 0
*/
tools/power/turbostat: Unify graphics sysfs snapshots Graphics sysfs snapshots share similar logic. Combine them into one function to avoid code duplication. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-13 10:12:19 +08:00
int snapshot_graphics(int idx)
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 01:28:12 -05:00
{
int retval;
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
rewind(gfx_info[idx].fp);
tools/power turbostat: Consolidate graphics sysfs access Currently, there is an inconsistency in how graphics sysfs knobs are accessed: graphics residency sysfs knobs are opened and closed for each read, while graphics frequency sysfs knobs are opened once and remain open until turbostat exits. This inconsistency is confusing and adds unnecessary code complexity. Consolidate the access method by opening the sysfs files once and reusing the file pointers for subsequent accesses. This approach simplifies the code and ensures a consistent method for accessing graphics sysfs knobs. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:44 +08:00
tools/power/turbostat: Unify graphics sysfs snapshots Graphics sysfs snapshots share similar logic. Combine them into one function to avoid code duplication. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-13 10:12:19 +08:00
switch (idx) {
case GFX_rc6:
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
case SAM_mc6:
tools/power turbostat: Consolidate graphics sysfs access Currently, there is an inconsistency in how graphics sysfs knobs are accessed: graphics residency sysfs knobs are opened and closed for each read, while graphics frequency sysfs knobs are opened once and remain open until turbostat exits. This inconsistency is confusing and adds unnecessary code complexity. Consolidate the access method by opening the sysfs files once and reusing the file pointers for subsequent accesses. This approach simplifies the code and ensures a consistent method for accessing graphics sysfs knobs. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:44 +08:00
retval = fscanf(gfx_info[idx].fp, "%lld", &gfx_info[idx].val_ull);
tools/power/turbostat: Unify graphics sysfs snapshots Graphics sysfs snapshots share similar logic. Combine them into one function to avoid code duplication. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-13 10:12:19 +08:00
if (retval != 1)
err(1, "rc6");
return 0;
case GFX_MHz:
case GFX_ACTMHz:
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
case SAM_MHz:
case SAM_ACTMHz:
tools/power/turbostat: Unify graphics sysfs snapshots Graphics sysfs snapshots share similar logic. Combine them into one function to avoid code duplication. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-13 10:12:19 +08:00
retval = fscanf(gfx_info[idx].fp, "%d", &gfx_info[idx].val);
if (retval != 1)
err(1, "MHz");
return 0;
default:
return -EINVAL;
tools/power turbostat: Add a new GFXAMHz column that exposes gt_act_freq_mhz. The column already present called GFXMHz reads from gt_cur_freq_mhz, which represents the GT frequency that was requested, but power management might not be able to do that. So the new column will display what the actual frequency GT is running at. Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-22 15:02:07 -07:00
}
}
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
/*
* snapshot_cpu_lpi()
*
* record snapshot of
* /sys/devices/system/cpu/cpuidle/low_power_idle_cpu_residency_us
*/
int snapshot_cpu_lpi_us(void)
{
FILE *fp;
int retval;
fp = fopen_or_die("/sys/devices/system/cpu/cpuidle/low_power_idle_cpu_residency_us", "r");
retval = fscanf(fp, "%lld", &cpuidle_cur_cpu_lpi_us);
tools/power turbostat: Warn on bad ACPI LPIT data On some systems /sys/devices/system/cpu/cpuidle/low_power_idle_cpu_residency_us or /sys/devices/system/cpu/cpuidle/low_power_idle_system_residency_us return a file error because of bad ACPI LPIT data from a misconfigured BIOS. turbostat interprets this failure as a fatal error and outputs turbostat: CPU LPI: No data available If the ACPI LPIT sysfs files return an error output a warning instead of a fatal error, disable the ACPI LPIT evaluation code, and continue. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-09-25 08:59:26 -04:00
if (retval != 1) {
fprintf(stderr, "Disabling Low Power Idle CPU output\n");
BIC_NOT_PRESENT(BIC_CPU_LPI);
tools/power turbostat: fix file descriptor leaks Fix file descriptor leaks by closing fp before return. Addresses-Coverity-ID: 1444591 ("Resource leak") Addresses-Coverity-ID: 1444592 ("Resource leak") Fixes: 5ea7647b333f ("tools/power turbostat: Warn on bad ACPI LPIT data") Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com> Reviewed-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-04-08 11:12:40 -05:00
fclose(fp);
tools/power turbostat: Warn on bad ACPI LPIT data On some systems /sys/devices/system/cpu/cpuidle/low_power_idle_cpu_residency_us or /sys/devices/system/cpu/cpuidle/low_power_idle_system_residency_us return a file error because of bad ACPI LPIT data from a misconfigured BIOS. turbostat interprets this failure as a fatal error and outputs turbostat: CPU LPI: No data available If the ACPI LPIT sysfs files return an error output a warning instead of a fatal error, disable the ACPI LPIT evaluation code, and continue. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-09-25 08:59:26 -04:00
return -1;
}
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
fclose(fp);
return 0;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
/*
* snapshot_sys_lpi()
*
tools/power turbostat: Fix missing SYS_LPI counter on some Chromebooks Some Chromebook BIOS' do not export an ACPI LPIT, which is how Linux finds the residency counter for CPU and SYSTEM low power states, that is exports in /sys/devices/system/cpu/cpuidle/*residency_us When these sysfs attributes are missing, check the debugfs attrubte from the pmc_core driver, which accesses the same counter value. Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 18:26:05 -04:00
* record snapshot of sys_lpi_file
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
*/
int snapshot_sys_lpi_us(void)
{
FILE *fp;
int retval;
tools/power turbostat: Fix missing SYS_LPI counter on some Chromebooks Some Chromebook BIOS' do not export an ACPI LPIT, which is how Linux finds the residency counter for CPU and SYSTEM low power states, that is exports in /sys/devices/system/cpu/cpuidle/*residency_us When these sysfs attributes are missing, check the debugfs attrubte from the pmc_core driver, which accesses the same counter value. Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 18:26:05 -04:00
fp = fopen_or_die(sys_lpi_file, "r");
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
retval = fscanf(fp, "%lld", &cpuidle_cur_sys_lpi_us);
tools/power turbostat: Warn on bad ACPI LPIT data On some systems /sys/devices/system/cpu/cpuidle/low_power_idle_cpu_residency_us or /sys/devices/system/cpu/cpuidle/low_power_idle_system_residency_us return a file error because of bad ACPI LPIT data from a misconfigured BIOS. turbostat interprets this failure as a fatal error and outputs turbostat: CPU LPI: No data available If the ACPI LPIT sysfs files return an error output a warning instead of a fatal error, disable the ACPI LPIT evaluation code, and continue. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-09-25 08:59:26 -04:00
if (retval != 1) {
fprintf(stderr, "Disabling Low Power Idle System output\n");
BIC_NOT_PRESENT(BIC_SYS_LPI);
tools/power turbostat: fix leak of file descriptor on error return path Currently the error return path does not close the file fp and leaks a file descriptor. Fix this by closing the file. Fixes: 5ea7647b333f ("tools/power turbostat: Warn on bad ACPI LPIT data") Signed-off-by: Colin Ian King <colin.king@canonical.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-04-08 10:00:44 +01:00
fclose(fp);
tools/power turbostat: Warn on bad ACPI LPIT data On some systems /sys/devices/system/cpu/cpuidle/low_power_idle_cpu_residency_us or /sys/devices/system/cpu/cpuidle/low_power_idle_system_residency_us return a file error because of bad ACPI LPIT data from a misconfigured BIOS. turbostat interprets this failure as a fatal error and outputs turbostat: CPU LPI: No data available If the ACPI LPIT sysfs files return an error output a warning instead of a fatal error, disable the ACPI LPIT evaluation code, and continue. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-09-25 08:59:26 -04:00
return -1;
}
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
fclose(fp);
return 0;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
/*
* snapshot /proc and /sys files
*
* return 1 if configuration restart needed, else return 0
*/
int snapshot_proc_sysfs_files(void)
{
tools/power turbostat: fix zero IRQ count shown in one-shot command mode The IRQ column has been working for periodic mode, but not in one-shot command mode, it shows only 0. until now. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-14 22:07:52 -05:00
if (DO_BIC(BIC_IRQ))
if (snapshot_proc_interrupts())
return 1;
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_GFX_rc6))
tools/power/turbostat: Unify graphics sysfs snapshots Graphics sysfs snapshots share similar logic. Combine them into one function to avoid code duplication. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-13 10:12:19 +08:00
snapshot_graphics(GFX_rc6);
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 01:28:12 -05:00
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (DO_BIC(BIC_GFXMHz))
tools/power/turbostat: Unify graphics sysfs snapshots Graphics sysfs snapshots share similar logic. Combine them into one function to avoid code duplication. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-13 10:12:19 +08:00
snapshot_graphics(GFX_MHz);
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 00:37:54 -05:00
tools/power turbostat: Add a new GFXAMHz column that exposes gt_act_freq_mhz. The column already present called GFXMHz reads from gt_cur_freq_mhz, which represents the GT frequency that was requested, but power management might not be able to do that. So the new column will display what the actual frequency GT is running at. Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-22 15:02:07 -07:00
if (DO_BIC(BIC_GFXACTMHz))
tools/power/turbostat: Unify graphics sysfs snapshots Graphics sysfs snapshots share similar logic. Combine them into one function to avoid code duplication. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-13 10:12:19 +08:00
snapshot_graphics(GFX_ACTMHz);
tools/power turbostat: Add a new GFXAMHz column that exposes gt_act_freq_mhz. The column already present called GFXMHz reads from gt_cur_freq_mhz, which represents the GT frequency that was requested, but power management might not be able to do that. So the new column will display what the actual frequency GT is running at. Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-22 15:02:07 -07:00
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
if (DO_BIC(BIC_SAM_mc6))
snapshot_graphics(SAM_mc6);
if (DO_BIC(BIC_SAMMHz))
snapshot_graphics(SAM_MHz);
if (DO_BIC(BIC_SAMACTMHz))
snapshot_graphics(SAM_ACTMHz);
tools/power turbostat: Build-in "Low Power Idle" counters support Linux 4.15 exports the ACPI Low Power Idle Table's counters in /sys/devices/system/cpu/cpuidle/ low_power_idle_cpu_residency_us Show this in the "CPU%LPI" column. Today this reflects the "North Complex" residency in PC10, so expect it to closely follow "Pk%pc10". low_power_idle_system_residency_us Show this in the "SYS%LPI" column. Today, this reflects the North is in PC10, plus the PCH is sufficiently quiescent to save additional power via the "S0ix" system state, as measured by the PCH SLP_S0 counter. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:35:53 -04:00
if (DO_BIC(BIC_CPU_LPI))
snapshot_cpu_lpi_us();
if (DO_BIC(BIC_SYS_LPI))
snapshot_sys_lpi_us();
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
return 0;
}
tools/power turbostat: on SIGINT: sample, print and exit When running in interval-mode, catch interrupts and print a final data record before exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-15 14:57:37 -04:00
int exit_requested;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
static void signal_handler(int signal)
tools/power turbostat: on SIGINT: sample, print and exit When running in interval-mode, catch interrupts and print a final data record before exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-15 14:57:37 -04:00
{
switch (signal) {
case SIGINT:
exit_requested = 1;
if (debug)
fprintf(stderr, " SIGINT\n");
break;
tools/power turbostat: on SIGUSR1: sample, print and continue Interval-mode turbostat now catches and discards SIGUSR1. Thus, SIGUSR1 can be used to tell turbostat to cut short the current measurement interval. Turbostat will then start the next measurement interval using the regular interval length. This can be used to give turbostat variable intervals. Invoke turbostat with --interval LARGE_NUMBER_SEC and have a program that has permission to send it a SIGUSR1 always before LARGE_NUMBER_SEC expires. It may also be useful to use "--enable Time_Of_Day_Seconds" to observe the actual interval length. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-15 15:51:26 -04:00
case SIGUSR1:
if (debug > 1)
fprintf(stderr, "SIGUSR1\n");
break;
tools/power turbostat: on SIGINT: sample, print and exit When running in interval-mode, catch interrupts and print a final data record before exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-15 14:57:37 -04:00
}
}
void setup_signal_handler(void)
{
struct sigaction sa;
memset(&sa, 0, sizeof(sa));
sa.sa_handler = &signal_handler;
if (sigaction(SIGINT, &sa, NULL) < 0)
err(1, "sigaction SIGINT");
tools/power turbostat: on SIGUSR1: sample, print and continue Interval-mode turbostat now catches and discards SIGUSR1. Thus, SIGUSR1 can be used to tell turbostat to cut short the current measurement interval. Turbostat will then start the next measurement interval using the regular interval length. This can be used to give turbostat variable intervals. Invoke turbostat with --interval LARGE_NUMBER_SEC and have a program that has permission to send it a SIGUSR1 always before LARGE_NUMBER_SEC expires. It may also be useful to use "--enable Time_Of_Day_Seconds" to observe the actual interval length. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-15 15:51:26 -04:00
if (sigaction(SIGUSR1, &sa, NULL) < 0)
err(1, "sigaction SIGUSR1");
tools/power turbostat: on SIGINT: sample, print and exit When running in interval-mode, catch interrupts and print a final data record before exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-15 14:57:37 -04:00
}
tools/power turbostat: end current interval upon newline input In turbostat interval mode, a newline typed on standard input will now conclude the current interval. Data will immediately be collected and printed for that interval, and the next interval will be started. This is similar to the recently added SIGUSR1 feature. But that is for use by programs, while this is for interactive use. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-19 19:28:37 -04:00
tools/power turbostat: fix printing on input The recent patch that implements table printing on a keypress introduced a regression - turbostat prints the table almost continuously if it is run from a daemon program. The problem is also easy to reproduce like this: echo | turbostat The reason is that we cannot assume that stdin is always a TTY. It can be many things. This patch adds fixes the problem by limiting the new keypress functionality to TTYs only. If stdin is not a TTY, we just sleep for the full interval time. While on it, clean-up 'do_sleep()' to return no value, as callers do not expect that anyway. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-04 15:01:47 +03:00
void do_sleep(void)
tools/power turbostat: end current interval upon newline input In turbostat interval mode, a newline typed on standard input will now conclude the current interval. Data will immediately be collected and printed for that interval, and the next interval will be started. This is similar to the recently added SIGUSR1 feature. But that is for use by programs, while this is for interactive use. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-19 19:28:37 -04:00
{
tools/power turbostat: read from pipes too Commit '47936f944e78 tools/power turbostat: fix printing on input' make a valid fix, but it completely disabled piped stdin support, which is a valuable use-case. Indeed, if stdin is a pipe, turbostat won't read anything from it, so it becomes impossible to get turbostat output at user-defined moments, instead of the regular intervals. There is no reason why this should works for terminals, but not for pipes. This patch improves the situation. Instead of ignoring pipes, we read data from them but gracefully handle the EOF case. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-08-14 20:12:55 +03:00
struct timeval tout;
struct timespec rest;
tools/power turbostat: end current interval upon newline input In turbostat interval mode, a newline typed on standard input will now conclude the current interval. Data will immediately be collected and printed for that interval, and the next interval will be started. This is similar to the recently added SIGUSR1 feature. But that is for use by programs, while this is for interactive use. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-19 19:28:37 -04:00
fd_set readfds;
int retval;
FD_ZERO(&readfds);
FD_SET(0, &readfds);
tools/power turbostat: read from pipes too Commit '47936f944e78 tools/power turbostat: fix printing on input' make a valid fix, but it completely disabled piped stdin support, which is a valuable use-case. Indeed, if stdin is a pipe, turbostat won't read anything from it, so it becomes impossible to get turbostat output at user-defined moments, instead of the regular intervals. There is no reason why this should works for terminals, but not for pipes. This patch improves the situation. Instead of ignoring pipes, we read data from them but gracefully handle the EOF case. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-08-14 20:12:55 +03:00
if (ignore_stdin) {
tools/power turbostat: fix printing on input The recent patch that implements table printing on a keypress introduced a regression - turbostat prints the table almost continuously if it is run from a daemon program. The problem is also easy to reproduce like this: echo | turbostat The reason is that we cannot assume that stdin is always a TTY. It can be many things. This patch adds fixes the problem by limiting the new keypress functionality to TTYs only. If stdin is not a TTY, we just sleep for the full interval time. While on it, clean-up 'do_sleep()' to return no value, as callers do not expect that anyway. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-04 15:01:47 +03:00
nanosleep(&interval_ts, NULL);
return;
}
tools/power turbostat: end current interval upon newline input In turbostat interval mode, a newline typed on standard input will now conclude the current interval. Data will immediately be collected and printed for that interval, and the next interval will be started. This is similar to the recently added SIGUSR1 feature. But that is for use by programs, while this is for interactive use. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-19 19:28:37 -04:00
tools/power turbostat: read from pipes too Commit '47936f944e78 tools/power turbostat: fix printing on input' make a valid fix, but it completely disabled piped stdin support, which is a valuable use-case. Indeed, if stdin is a pipe, turbostat won't read anything from it, so it becomes impossible to get turbostat output at user-defined moments, instead of the regular intervals. There is no reason why this should works for terminals, but not for pipes. This patch improves the situation. Instead of ignoring pipes, we read data from them but gracefully handle the EOF case. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-08-14 20:12:55 +03:00
tout = interval_tv;
retval = select(1, &readfds, NULL, NULL, &tout);
tools/power turbostat: end current interval upon newline input In turbostat interval mode, a newline typed on standard input will now conclude the current interval. Data will immediately be collected and printed for that interval, and the next interval will be started. This is similar to the recently added SIGUSR1 feature. But that is for use by programs, while this is for interactive use. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-19 19:28:37 -04:00
if (retval == 1) {
switch (getc(stdin)) {
case 'q':
exit_requested = 1;
break;
tools/power turbostat: read from pipes too Commit '47936f944e78 tools/power turbostat: fix printing on input' make a valid fix, but it completely disabled piped stdin support, which is a valuable use-case. Indeed, if stdin is a pipe, turbostat won't read anything from it, so it becomes impossible to get turbostat output at user-defined moments, instead of the regular intervals. There is no reason why this should works for terminals, but not for pipes. This patch improves the situation. Instead of ignoring pipes, we read data from them but gracefully handle the EOF case. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-08-14 20:12:55 +03:00
case EOF:
/*
* 'stdin' is a pipe closed on the other end. There
* won't be any further input.
*/
ignore_stdin = 1;
/* Sleep the rest of the time */
rest.tv_sec = (tout.tv_sec + tout.tv_usec / 1000000);
rest.tv_nsec = (tout.tv_usec % 1000000) * 1000;
nanosleep(&rest, NULL);
tools/power turbostat: end current interval upon newline input In turbostat interval mode, a newline typed on standard input will now conclude the current interval. Data will immediately be collected and printed for that interval, and the next interval will be started. This is similar to the recently added SIGUSR1 feature. But that is for use by programs, while this is for interactive use. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-19 19:28:37 -04:00
}
}
}
tools/power turbostat: fix printing on input The recent patch that implements table printing on a keypress introduced a regression - turbostat prints the table almost continuously if it is run from a daemon program. The problem is also easy to reproduce like this: echo | turbostat The reason is that we cannot assume that stdin is always a TTY. It can be many things. This patch adds fixes the problem by limiting the new keypress functionality to TTYs only. If stdin is not a TTY, we just sleep for the full interval time. While on it, clean-up 'do_sleep()' to return no value, as callers do not expect that anyway. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-04 15:01:47 +03:00
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
int get_msr_sum(int cpu, off_t offset, unsigned long long *msr)
{
int ret, idx;
unsigned long long msr_cur, msr_last;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
assert(!no_msr);
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
if (!per_cpu_msr_sum)
return 1;
idx = offset_to_idx(offset);
if (idx < 0)
return idx;
/* get_msr_sum() = sum + (get_msr() - last) */
ret = get_msr(cpu, offset, &msr_cur);
if (ret)
return ret;
msr_last = per_cpu_msr_sum[cpu].entries[idx].last;
DELTA_WRAP32(msr_cur, msr_last);
*msr = msr_last + per_cpu_msr_sum[cpu].entries[idx].sum;
return 0;
}
timer_t timerid;
/* Timer callback, update the sum of MSRs periodically. */
static int update_msr_sum(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
int i, ret;
int cpu = t->cpu_id;
tools/power turbostat: No build warnings with -Wextra Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 23:45:18 -04:00
UNUSED(c);
UNUSED(p);
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
assert(!no_msr);
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
for (i = IDX_PKG_ENERGY; i < IDX_COUNT; i++) {
unsigned long long msr_cur, msr_last;
tools/power turbostat: Fix offset overflow issue in index converting The idx_to_offset() function returns type int (32-bit signed), but MSR_PKG_ENERGY_STAT is u32 and would be interpreted as a negative number. The end result is that it hits the if (offset < 0) check in update_msr_sum() which prevents the timer callback from updating the stat in the background when long durations are used. The similar issue exists in offset_to_idx() and update_msr_sum(). Fix this issue by converting the 'int' to 'off_t' accordingly. Fixes: 9972d5d84d76 ("tools/power turbostat: Enable accumulate RAPL display") Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-28 17:09:16 +08:00
off_t offset;
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
if (!idx_valid(i))
continue;
offset = idx_to_offset(i);
if (offset < 0)
continue;
ret = get_msr(cpu, offset, &msr_cur);
if (ret) {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "Can not update msr(0x%llx)\n", (unsigned long long)offset);
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
continue;
}
msr_last = per_cpu_msr_sum[cpu].entries[i].last;
per_cpu_msr_sum[cpu].entries[i].last = msr_cur & 0xffffffff;
DELTA_WRAP32(msr_cur, msr_last);
per_cpu_msr_sum[cpu].entries[i].sum += msr_last;
}
return 0;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
static void msr_record_handler(union sigval v)
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
{
tools/power turbostat: No build warnings with -Wextra Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 23:45:18 -04:00
UNUSED(v);
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
for_all_cpus(update_msr_sum, EVEN_COUNTERS);
}
void msr_sum_record(void)
{
struct itimerspec its;
struct sigevent sev;
per_cpu_msr_sum = calloc(topo.max_cpu_num + 1, sizeof(struct msr_sum_array));
if (!per_cpu_msr_sum) {
fprintf(outf, "Can not allocate memory for long time MSR.\n");
return;
}
/*
* Signal handler might be restricted, so use thread notifier instead.
*/
memset(&sev, 0, sizeof(struct sigevent));
sev.sigev_notify = SIGEV_THREAD;
sev.sigev_notify_function = msr_record_handler;
sev.sigev_value.sival_ptr = &timerid;
if (timer_create(CLOCK_REALTIME, &sev, &timerid) == -1) {
fprintf(outf, "Can not create timer.\n");
goto release_msr;
}
its.it_value.tv_sec = 0;
its.it_value.tv_nsec = 1;
/*
* A wraparound time has been calculated early.
* Some sources state that the peak power for a
* microprocessor is usually 1.5 times the TDP rating,
* use 2 * TDP for safety.
*/
its.it_interval.tv_sec = rapl_joule_counter_range / 2;
its.it_interval.tv_nsec = 0;
if (timer_settime(timerid, 0, &its, NULL) == -1) {
fprintf(outf, "Can not set timer.\n");
goto release_timer;
}
return;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
release_timer:
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
timer_delete(timerid);
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
release_msr:
tools/power turbostat: Introduce functions to accumulate RAPL consumption Since the RAPL Joule Counter is 32 bit, turbostat would only print a *star* instead of printing the actual energy consumed to indicate the overflow due to long duration. This does not meet the requirement from servers as the sampling time of turbostat is usually very long on servers. So maintain a set of MSR buffer, and update them periodically before the 32bit MSR register is wrapped round, so as to avoid the overflow. The idea is similar to the implementation of ktime_get(): Periodical MSR timer: total_rapl_sum += (current_rapl_msr - last_rapl_msr); Using get_msr_sum() to get the accumulated RAPL: return (current_rapl_msr - last_rapl_msr) + total_rapl_sum; The accumulated RAPL mechanism will be turned on in next patch. Originally-by: Aaron Lu <aaron.lwe@gmail.com> Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-04-18 16:31:57 +08:00
free(per_cpu_msr_sum);
}
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
/*
* set_my_sched_priority(pri)
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
* return previous priority on success
* return value < -20 on failure
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
*/
int set_my_sched_priority(int priority)
{
int retval;
int original_priority;
errno = 0;
original_priority = getpriority(PRIO_PROCESS, 0);
if (errno && (original_priority == -1))
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
return -21;
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
retval = setpriority(PRIO_PROCESS, 0, priority);
if (retval)
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
return -21;
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
errno = 0;
retval = getpriority(PRIO_PROCESS, 0);
if (retval != priority)
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
return -21;
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
return original_priority;
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
void turbostat_loop()
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
int retval;
tools/power turbostat: prevent infinite loop on migration error path Turbostat assumed if it can't migrate to a CPU, then the CPU must have gone off-line and turbostat should re-initialize with the new topology. But if turbostat can not migrate because it is restricted by a cpuset, then it will fail to migrate even after re-initialization, resulting in an infinite loop. Spit out a warning when we can't migrate and endure only 2 re-initialize cycles in a row before giving up and exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 22:38:05 -05:00
int restarted = 0;
tools/power turbostat: No build warnings with -Wextra Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 23:45:18 -04:00
unsigned int done_iters = 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: on SIGINT: sample, print and exit When running in interval-mode, catch interrupts and print a final data record before exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-15 14:57:37 -04:00
setup_signal_handler();
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
/*
* elevate own priority for interval mode
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
*
* ignore on error - we probably don't have permission to set it, but
* it's not a big deal
tools/power turbostat: elevate priority of interval mode This makes interval mode less likely to see delayed results on a heavily loaded system. Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:27:34 -04:00
*/
set_my_sched_priority(-20);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
restart:
tools/power turbostat: prevent infinite loop on migration error path Turbostat assumed if it can't migrate to a CPU, then the CPU must have gone off-line and turbostat should re-initialize with the new topology. But if turbostat can not migrate because it is restricted by a cpuset, then it will fail to migrate even after re-initialization, resulting in an infinite loop. Spit out a warning when we can't migrate and endure only 2 re-initialize cycles in a row before giving up and exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 22:38:05 -05:00
restarted++;
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
snapshot_proc_sysfs_files();
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
retval = for_all_cpus(get_counters, EVEN_COUNTERS);
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
first_counter_read = 0;
tools/power turbostat: graceful fail on garbage input When invald MSR's are specified on the command line, turbostat should simply print an error and exit. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-01 00:08:19 -04:00
if (retval < -1) {
exit(retval);
} else if (retval == -1) {
tools/power turbostat: harden against cpu hotplug turbostat tends to get confused when CPUs are added and removed while it is running. There are races, such as checking the current cpu, and then reading a sysfs file that depends on that cpu number. Close the two issues that seem to come up the most. First, there is an infinite reset loop detector -- change that to allow more resets before giving up. Secondly, one of those file reads didn't really need to exit the program on failure... Signed-off-by: Len Brown <len.brown@intel.com>
2020-09-30 20:58:15 -04:00
if (restarted > 10) {
tools/power turbostat: prevent infinite loop on migration error path Turbostat assumed if it can't migrate to a CPU, then the CPU must have gone off-line and turbostat should re-initialize with the new topology. But if turbostat can not migrate because it is restricted by a cpuset, then it will fail to migrate even after re-initialization, resulting in an infinite loop. Spit out a warning when we can't migrate and endure only 2 re-initialize cycles in a row before giving up and exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 22:38:05 -05:00
exit(retval);
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
re_initialize();
goto restart;
}
tools/power turbostat: prevent infinite loop on migration error path Turbostat assumed if it can't migrate to a CPU, then the CPU must have gone off-line and turbostat should re-initialize with the new topology. But if turbostat can not migrate because it is restricted by a cpuset, then it will fail to migrate even after re-initialization, resulting in an infinite loop. Spit out a warning when we can't migrate and endure only 2 re-initialize cycles in a row before giving up and exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 22:38:05 -05:00
restarted = 0;
tools/power turbostat: if --num_iterations, print for specific number of iterations There's a use case during test to only print specific round of iterations if --num_iterations is specified, for example, with this patch applied: turbostat -i 5 -n 4 will capture 4 samples with 5 seconds interval. [lenb: renamed to --num_iterations from --iterations] Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Reviewed-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-04-26 08:41:03 +08:00
done_iters = 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
gettimeofday(&tv_even, (struct timezone *)NULL);
while (1) {
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
if (for_all_proc_cpus(cpu_is_not_present)) {
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
re_initialize();
goto restart;
}
tools/power/turbostat: Handle cgroup v2 cpu limitation CPUs can be isolated via cgroup settings and turbostat should avoid migrating to these CPUs, just like it does for the '-c' cpus. Introduce cpu_effective_set to save the cgroup cpu limitation info from /sys/fs/cgroup/cpuset.cpus.effective. And use cpu_allowed_set as the intersection of cpu_present_set, cpu_effective_set and cpu_subset. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:39:22 +08:00
if (update_effective_str(false)) {
re_initialize();
goto restart;
}
tools/power turbostat: end current interval upon newline input In turbostat interval mode, a newline typed on standard input will now conclude the current interval. Data will immediately be collected and printed for that interval, and the next interval will be started. This is similar to the recently added SIGUSR1 feature. But that is for use by programs, while this is for interactive use. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-19 19:28:37 -04:00
do_sleep();
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
if (snapshot_proc_sysfs_files())
goto restart;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
retval = for_all_cpus(get_counters, ODD_COUNTERS);
tools/power turbostat: graceful fail on garbage input When invald MSR's are specified on the command line, turbostat should simply print an error and exit. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-01 00:08:19 -04:00
if (retval < -1) {
exit(retval);
} else if (retval == -1) {
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-29 22:19:58 -04:00
re_initialize();
goto restart;
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
gettimeofday(&tv_odd, (struct timezone *)NULL);
timersub(&tv_odd, &tv_even, &tv_delta);
tools/power turbostat: Gracefully handle ACPI S3 turbostat gives valid results across suspend to idle, aka freeze, whether invoked in interval mode, or in command mode. Indeed, this can be used to measure suspend to idle: turbostat echo freeze > /sys/power/state But this does not work across suspend to ACPI S3, because the processor counters, including the TSC, are reset on resume. Further, when turbostat detects a problem, it does't forgive the hardware, and interval mode will print *'s from there on out. Instead, upon detecting counters going backwards, simply reset and start over. Interval mode across ACPI S3: (observe TSC going backwards) root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 1 0.06 858 2294 0x0000000000000000 0 0 0.06 847 2294 0x0000002a254b98ac 1 1 0.06 878 2294 0x0000002a254efa3a 2 1 0.07 843 2294 0x0000002a2551df65 3 0 0.05 863 2294 0x0000002a2553fea2 turbostat: re-initialized with num_cpus 4 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 2 0.20 849 2294 0x0000000000000000 0 2 0.26 856 2294 0x0000000449abb60d 1 2 0.20 844 2294 0x0000000449b087ec 2 2 0.21 850 2294 0x0000000449b35d5d 3 1 0.12 839 2294 0x0000000449b5fd5a ^C Command mode across ACPI S3: root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 sleep 10 ./turbostat: Counter reset detected 14.196299 sec Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 20:31:46 -04:00
if (for_all_cpus_2(delta_cpu, ODD_COUNTERS, EVEN_COUNTERS)) {
re_initialize();
goto restart;
}
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
delta_platform(&platform_counters_odd, &platform_counters_even);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
compute_average(EVEN_COUNTERS);
format_all_counters(EVEN_COUNTERS);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
flush_output_stdout();
tools/power turbostat: on SIGINT: sample, print and exit When running in interval-mode, catch interrupts and print a final data record before exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-15 14:57:37 -04:00
if (exit_requested)
break;
tools/power turbostat: if --num_iterations, print for specific number of iterations There's a use case during test to only print specific round of iterations if --num_iterations is specified, for example, with this patch applied: turbostat -i 5 -n 4 will capture 4 samples with 5 seconds interval. [lenb: renamed to --num_iterations from --iterations] Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Reviewed-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-04-26 08:41:03 +08:00
if (num_iterations && ++done_iters >= num_iterations)
break;
tools/power turbostat: end current interval upon newline input In turbostat interval mode, a newline typed on standard input will now conclude the current interval. Data will immediately be collected and printed for that interval, and the next interval will be started. This is similar to the recently added SIGUSR1 feature. But that is for use by programs, while this is for interactive use. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-19 19:28:37 -04:00
do_sleep();
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
if (snapshot_proc_sysfs_files())
goto restart;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
retval = for_all_cpus(get_counters, EVEN_COUNTERS);
tools/power turbostat: graceful fail on garbage input When invald MSR's are specified on the command line, turbostat should simply print an error and exit. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-01 00:08:19 -04:00
if (retval < -1) {
exit(retval);
} else if (retval == -1) {
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
re_initialize();
goto restart;
}
gettimeofday(&tv_even, (struct timezone *)NULL);
timersub(&tv_even, &tv_odd, &tv_delta);
tools/power turbostat: Gracefully handle ACPI S3 turbostat gives valid results across suspend to idle, aka freeze, whether invoked in interval mode, or in command mode. Indeed, this can be used to measure suspend to idle: turbostat echo freeze > /sys/power/state But this does not work across suspend to ACPI S3, because the processor counters, including the TSC, are reset on resume. Further, when turbostat detects a problem, it does't forgive the hardware, and interval mode will print *'s from there on out. Instead, upon detecting counters going backwards, simply reset and start over. Interval mode across ACPI S3: (observe TSC going backwards) root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 1 0.06 858 2294 0x0000000000000000 0 0 0.06 847 2294 0x0000002a254b98ac 1 1 0.06 878 2294 0x0000002a254efa3a 2 1 0.07 843 2294 0x0000002a2551df65 3 0 0.05 863 2294 0x0000002a2553fea2 turbostat: re-initialized with num_cpus 4 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 2 0.20 849 2294 0x0000000000000000 0 2 0.26 856 2294 0x0000000449abb60d 1 2 0.20 844 2294 0x0000000449b087ec 2 2 0.21 850 2294 0x0000000449b35d5d 3 1 0.12 839 2294 0x0000000449b5fd5a ^C Command mode across ACPI S3: root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 sleep 10 ./turbostat: Counter reset detected 14.196299 sec Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 20:31:46 -04:00
if (for_all_cpus_2(delta_cpu, EVEN_COUNTERS, ODD_COUNTERS)) {
re_initialize();
goto restart;
}
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
delta_platform(&platform_counters_even, &platform_counters_odd);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
compute_average(ODD_COUNTERS);
format_all_counters(ODD_COUNTERS);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
flush_output_stdout();
tools/power turbostat: on SIGINT: sample, print and exit When running in interval-mode, catch interrupts and print a final data record before exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-15 14:57:37 -04:00
if (exit_requested)
break;
tools/power turbostat: if --num_iterations, print for specific number of iterations There's a use case during test to only print specific round of iterations if --num_iterations is specified, for example, with this patch applied: turbostat -i 5 -n 4 will capture 4 samples with 5 seconds interval. [lenb: renamed to --num_iterations from --iterations] Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Reviewed-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-04-26 08:41:03 +08:00
if (num_iterations && ++done_iters >= num_iterations)
break;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
}
void check_dev_msr()
{
struct stat sb;
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 08:34:28 -04:00
char pathname[32];
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
if (no_msr)
return;
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 08:34:28 -04:00
sprintf(pathname, "/dev/cpu/%d/msr", base_cpu);
if (stat(pathname, &sb))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
if (system("/sbin/modprobe msr > /dev/null 2>&1"))
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
no_msr = 1;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: Fix 32-bit capabilities warning warning: `turbostat' uses 32-bit capabilities (legacy support in use) Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 23:24:17 -04:00
/*
* check for CAP_SYS_RAWIO
* return 0 on success
* return 1 on fail
*/
int check_for_cap_sys_rawio(void)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
tools/power turbostat: Fix 32-bit capabilities warning warning: `turbostat' uses 32-bit capabilities (legacy support in use) Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 23:24:17 -04:00
cap_t caps;
cap_flag_value_t cap_flag_value;
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
int ret = 0;
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 00:36:50 -04:00
tools/power turbostat: Fix 32-bit capabilities warning warning: `turbostat' uses 32-bit capabilities (legacy support in use) Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 23:24:17 -04:00
caps = cap_get_proc();
if (caps == NULL)
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
return 1;
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 00:36:50 -04:00
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
if (cap_get_flag(caps, CAP_SYS_RAWIO, CAP_EFFECTIVE, &cap_flag_value)) {
ret = 1;
goto free_and_exit;
}
tools/power turbostat: Fix 32-bit capabilities warning warning: `turbostat' uses 32-bit capabilities (legacy support in use) Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 23:24:17 -04:00
if (cap_flag_value != CAP_SET) {
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
ret = 1;
goto free_and_exit;
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 00:36:50 -04:00
}
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
free_and_exit:
tools/power turbostat: Fix 32-bit capabilities warning warning: `turbostat' uses 32-bit capabilities (legacy support in use) Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 23:24:17 -04:00
if (cap_free(caps) == -1)
err(-6, "cap_free\n");
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
return ret;
tools/power turbostat: Fix 32-bit capabilities warning warning: `turbostat' uses 32-bit capabilities (legacy support in use) Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 23:24:17 -04:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
void check_msr_permission(void)
tools/power turbostat: Fix 32-bit capabilities warning warning: `turbostat' uses 32-bit capabilities (legacy support in use) Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 23:24:17 -04:00
{
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
int failed = 0;
tools/power turbostat: Fix 32-bit capabilities warning warning: `turbostat' uses 32-bit capabilities (legacy support in use) Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 23:24:17 -04:00
char pathname[32];
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
if (no_msr)
return;
tools/power turbostat: Fix 32-bit capabilities warning warning: `turbostat' uses 32-bit capabilities (legacy support in use) Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 23:24:17 -04:00
/* check for CAP_SYS_RAWIO */
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
failed += check_for_cap_sys_rawio();
tools/power turbostat: Fix 32-bit capabilities warning warning: `turbostat' uses 32-bit capabilities (legacy support in use) Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-19 23:24:17 -04:00
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 00:36:50 -04:00
/* test file permissions */
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 08:34:28 -04:00
sprintf(pathname, "/dev/cpu/%d/msr", base_cpu);
if (euidaccess(pathname, R_OK)) {
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
failed++;
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 00:36:50 -04:00
}
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
if (failed) {
warnx("Failed to access %s. Some of the counters may not be available\n"
"\tRun as root to enable them or use %s to disable the access explicitly", pathname, "--no-msr");
no_msr = 1;
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
void probe_bclk(void)
{
unsigned long long msr;
unsigned int base_ratio;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (!platform->has_nhm_msrs || no_msr)
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
return;
if (platform->bclk_freq == BCLK_100MHZ)
bclk = 100.00;
else if (platform->bclk_freq == BCLK_133MHZ)
bclk = 133.33;
else if (platform->bclk_freq == BCLK_SLV)
bclk = slm_bclk();
else
return;
get_msr(base_cpu, MSR_PLATFORM_INFO, &msr);
base_ratio = (msr >> 8) & 0xFF;
base_hz = base_ratio * bclk * 1000000;
has_base_hz = 1;
tools/power/turbostat: Abstract TSC tweak support On some models, the CPU base frequency is different from the TSC frequency, and the aperf/mperf counters are running at CPU base frequency instead of TSC frequency. Abstract support for TSC tweak. Given that tsc_tweak depends on base_hz, move the code to probe_bclk() after base_hz is available. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:37:23 +08:00
if (platform->enable_tsc_tweak)
tsc_tweak = base_hz / tsc_hz;
tools/power/turbostat: Abstract BCLK frequency support Abstract CPU base clock frequency support. Note that bclk is used by 1. calculate base_hz using MSR_PLATFORM_INFO, which is guarded by probe_nhm_msrs(). 2. dump MSR_PLATFORM_INFO and Turbo Ratio Limit MSRs, which are also guarded by probe_nhm_msrs(). Thus probe_bclk() works for probe_nhm_msrs() models only. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:22:48 +08:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
static void remove_underbar(char *s)
tools/power turbostat: Fix output formatting for ACPI CST enumeration turbostat formatting is broken with ACPI CST for enumeration. The problem is that the CX_ACPI% is eight characters long which does not work with tab formatting. One simple solution is to remove the underbar from the state name such that C1_ACPI will be displayed as C1ACPI. Signed-off-by: David Arcari <darcari@redhat.com> Cc: Len Brown <lenb@kernel.org> Cc: linux-kernel@vger.kernel.org Signed-off-by: Len Brown <len.brown@intel.com>
2020-08-10 10:43:30 -04:00
{
char *to = s;
while (*s) {
if (*s != '_')
*to++ = *s;
s++;
}
*to = 0;
}
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
static void dump_turbo_ratio_info(void)
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
{
tools/power turbostat: Do not dump TRL if turbo is not supported Do not dump turbo ratio limits if platform does not support turbo, because it is confusing and the TRL MSRs may even include misleading information. And they are not supposed to be relied on if turbo is not supported. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-09-16 14:59:50 +03:00
if (!has_turbo)
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
return;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (!platform->has_nhm_msrs || no_msr)
tools/power/turbostat: Relocate pstate probing code Introduce probe_pstates() and move all pstate probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 15:58:10 +08:00
return;
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
if (platform->trl_msrs & TRL_LIMIT2)
tools/power/turbostat: Rename some TRL functions Rename dump_hsw_turbo_ratio_limits() and dump_ivt_turbo_ratio_limits() to dump_turbo_ratio_limit2() and dump_turbo_ratio_limit1() because they dump MSR_TURBO_RATIO_LIMIT1/LIMIT2, and the MSRs' behavior is consistent when they are available. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-07-30 13:54:25 +08:00
dump_turbo_ratio_limit2();
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
if (platform->trl_msrs & TRL_LIMIT1)
tools/power/turbostat: Rename some TRL functions Rename dump_hsw_turbo_ratio_limits() and dump_ivt_turbo_ratio_limits() to dump_turbo_ratio_limit2() and dump_turbo_ratio_limit1() because they dump MSR_TURBO_RATIO_LIMIT1/LIMIT2, and the MSRs' behavior is consistent when they are available. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-07-30 13:54:25 +08:00
dump_turbo_ratio_limit1();
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
if (platform->trl_msrs & TRL_BASE) {
dump_turbo_ratio_limits(MSR_TURBO_RATIO_LIMIT);
tools/power turbostat: dump secondary Turbo-Ratio-Limit Intel Performance Hybrid processors have a 2nd MSR describing the turbo limits enforced on the Ecores. Note, TRL and Secondary-TRL are usually R/O information, but on overclock-capable parts, they can be written. Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-31 17:29:13 -10:00
if (is_hybrid)
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
dump_turbo_ratio_limits(MSR_SECONDARY_TURBO_RATIO_LIMIT);
tools/power turbostat: dump secondary Turbo-Ratio-Limit Intel Performance Hybrid processors have a 2nd MSR describing the turbo limits enforced on the Ecores. Note, TRL and Secondary-TRL are usually R/O information, but on overclock-capable parts, they can be written. Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-31 17:29:13 -10:00
}
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
if (platform->trl_msrs & TRL_ATOM)
tools/power turbostat: dump Atom P-states correctly Turbostat dumps MSR_TURBO_RATIO_LIMIT on Core Architecture. But Atom Architecture uses MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-12 23:49:18 -05:00
dump_atom_turbo_ratio_limits();
tools/power/turbostat: Abstract Turbo Ratio Limit MSRs support Abstract the support for MSR_TURBO_RATIO_LIMIT, MSR_TRUBO_RATIO_LIMIT1, MSR_TURBO_RATIO_LIMIT2, MSR_SECONDARY_TURBO_RATIO_LIMIT, MSR_ATOM_CORE_RATIOS and MSR_ATOM_CORE_TURBO_RATIOS. Delete has_turbo_ratio_group_limits(), has_turbo_ratio_limit(), has_atom_turbo_ratio_limit(), has_ivt_turbo_ratio_limit(), has_hsw_turbo_ratio_limit(), has_knl_turbo_ratio_limit() and has_glm_turbo_ratio_limit() CPU model checks. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-08 23:16:00 +08:00
if (platform->trl_msrs & TRL_KNL)
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-20 09:49:34 -07:00
dump_knl_turbo_ratio_limits();
tools/power/turbostat: Abstract Config TDP MSRs support Abstract the support for MSR_CONFIG_TDP_NOMINAL/LEVEL_1/LEVEL_2/CONTROL and MSR_TURBO_ACTIVATION_RATIO. Delete has_config_tdp() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 14:19:46 +08:00
if (platform->has_config_tdp)
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
dump_config_tdp();
tools/power turbostat: Do not dump TRL if turbo is not supported Do not dump turbo ratio limits if platform does not support turbo, because it is confusing and the TRL MSRs may even include misleading information. And they are not supposed to be relied on if turbo is not supported. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-09-16 14:59:50 +03:00
}
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-17 16:23:45 -04:00
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
static int read_sysfs_int(char *path)
{
FILE *input;
int retval = -1;
input = fopen(path, "r");
if (input == NULL) {
if (debug)
fprintf(outf, "NSFOD %s\n", path);
return (-1);
}
if (fscanf(input, "%d", &retval) != 1)
err(1, "%s: failed to read int from file", path);
fclose(input);
return (retval);
}
tools/power turbostat: Print cpuidle information Print cpuidle driver and governor. Originally-by: Antti Laakso <antti.laakso@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-21 00:47:47 -04:00
static void dump_sysfs_file(char *path)
{
FILE *input;
char cpuidle_buf[64];
input = fopen(path, "r");
if (input == NULL) {
if (debug)
fprintf(outf, "NSFOD %s\n", path);
return;
}
if (!fgets(cpuidle_buf, sizeof(cpuidle_buf), input))
err(1, "%s: failed to read file", path);
fclose(input);
fprintf(outf, "%s: %s", strrchr(path, '/') + 1, cpuidle_buf);
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: Harden probe_intel_uncore_frequency() If sysfs directory "intel_uncore_frequency/cluster00/" exists, then use uncore cluster code (now its own routine). The previous check for "intel_uncore_frequency/package_00_die_00/current_freq_khz", could be unreliable in the face of sparse die id's. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 14:45:10 -04:00
static void probe_intel_uncore_frequency_legacy(void)
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
{
int i, j;
tools/power turbostat: Expand probe_intel_uncore_frequency() Print current frequency along with the current (and initial) limits Probe and print uncore config also for machines using the new cluster API Signed-off-by: Len Brown <len.brown@intel.com>
2023-12-20 13:11:05 -05:00
char path[256];
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
for (i = 0; i < topo.num_packages; ++i) {
tools/power turbostat: Survive sparse die_id Turbostat assumed that every package had a die_id = 0. When this assumption was violated, it exited when looking for the package uncore frequency: turbostat: /sys/.../intel_uncore_frequency/package_01_die_00/current_freq_khz: open failed: No such file or directory Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 12:02:24 -04:00
for (j = 0; j <= topo.max_die_id; ++j) {
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
int k, l;
tools/power turbostat: Expand probe_intel_uncore_frequency() Print current frequency along with the current (and initial) limits Probe and print uncore config also for machines using the new cluster API Signed-off-by: Len Brown <len.brown@intel.com>
2023-12-20 13:11:05 -05:00
char path_base[128];
sprintf(path_base, "/sys/devices/system/cpu/intel_uncore_frequency/package_%02d_die_%02d", i,
j);
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
tools/power turbostat: Survive sparse die_id Turbostat assumed that every package had a die_id = 0. When this assumption was violated, it exited when looking for the package uncore frequency: turbostat: /sys/.../intel_uncore_frequency/package_01_die_00/current_freq_khz: open failed: No such file or directory Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 12:02:24 -04:00
if (access(path_base, R_OK))
continue;
BIC_PRESENT(BIC_UNCORE_MHZ);
if (quiet)
return;
tools/power turbostat: Expand probe_intel_uncore_frequency() Print current frequency along with the current (and initial) limits Probe and print uncore config also for machines using the new cluster API Signed-off-by: Len Brown <len.brown@intel.com>
2023-12-20 13:11:05 -05:00
sprintf(path, "%s/min_freq_khz", path_base);
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
k = read_sysfs_int(path);
tools/power turbostat: Expand probe_intel_uncore_frequency() Print current frequency along with the current (and initial) limits Probe and print uncore config also for machines using the new cluster API Signed-off-by: Len Brown <len.brown@intel.com>
2023-12-20 13:11:05 -05:00
sprintf(path, "%s/max_freq_khz", path_base);
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
l = read_sysfs_int(path);
tools/power turbostat: Expand probe_intel_uncore_frequency() Print current frequency along with the current (and initial) limits Probe and print uncore config also for machines using the new cluster API Signed-off-by: Len Brown <len.brown@intel.com>
2023-12-20 13:11:05 -05:00
fprintf(outf, "Uncore Frequency package%d die%d: %d - %d MHz ", i, j, k / 1000, l / 1000);
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
tools/power turbostat: Expand probe_intel_uncore_frequency() Print current frequency along with the current (and initial) limits Probe and print uncore config also for machines using the new cluster API Signed-off-by: Len Brown <len.brown@intel.com>
2023-12-20 13:11:05 -05:00
sprintf(path, "%s/initial_min_freq_khz", path_base);
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
k = read_sysfs_int(path);
tools/power turbostat: Expand probe_intel_uncore_frequency() Print current frequency along with the current (and initial) limits Probe and print uncore config also for machines using the new cluster API Signed-off-by: Len Brown <len.brown@intel.com>
2023-12-20 13:11:05 -05:00
sprintf(path, "%s/initial_max_freq_khz", path_base);
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
l = read_sysfs_int(path);
tools/power turbostat: Expand probe_intel_uncore_frequency() Print current frequency along with the current (and initial) limits Probe and print uncore config also for machines using the new cluster API Signed-off-by: Len Brown <len.brown@intel.com>
2023-12-20 13:11:05 -05:00
fprintf(outf, "(%d - %d MHz)", k / 1000, l / 1000);
sprintf(path, "%s/current_freq_khz", path_base);
k = read_sysfs_int(path);
fprintf(outf, " %d MHz\n", k / 1000);
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
}
}
tools/power turbostat: Harden probe_intel_uncore_frequency() If sysfs directory "intel_uncore_frequency/cluster00/" exists, then use uncore cluster code (now its own routine). The previous check for "intel_uncore_frequency/package_00_die_00/current_freq_khz", could be unreliable in the face of sparse die id's. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 14:45:10 -04:00
}
static void probe_intel_uncore_frequency_cluster(void)
{
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
int i, uncore_max_id;
tools/power turbostat: Harden probe_intel_uncore_frequency() If sysfs directory "intel_uncore_frequency/cluster00/" exists, then use uncore cluster code (now its own routine). The previous check for "intel_uncore_frequency/package_00_die_00/current_freq_khz", could be unreliable in the face of sparse die id's. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 14:45:10 -04:00
char path[256];
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
char path_base[128];
tools/power turbostat: Expand probe_intel_uncore_frequency() Print current frequency along with the current (and initial) limits Probe and print uncore config also for machines using the new cluster API Signed-off-by: Len Brown <len.brown@intel.com>
2023-12-20 13:11:05 -05:00
if (access("/sys/devices/system/cpu/intel_uncore_frequency/uncore00/current_freq_khz", R_OK))
return;
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
for (uncore_max_id = 0;; ++uncore_max_id) {
sprintf(path_base, "/sys/devices/system/cpu/intel_uncore_frequency/uncore%02d", uncore_max_id);
/* uncore## start at 00 and skips no numbers, so stop upon first missing */
if (access(path_base, R_OK)) {
uncore_max_id -= 1;
break;
}
}
for (i = uncore_max_id; i >= 0; --i) {
tools/power turbostat: Expand probe_intel_uncore_frequency() Print current frequency along with the current (and initial) limits Probe and print uncore config also for machines using the new cluster API Signed-off-by: Len Brown <len.brown@intel.com>
2023-12-20 13:11:05 -05:00
int k, l;
int package_id, domain_id, cluster_id;
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
char name_buf[16];
tools/power turbostat: Expand probe_intel_uncore_frequency() Print current frequency along with the current (and initial) limits Probe and print uncore config also for machines using the new cluster API Signed-off-by: Len Brown <len.brown@intel.com>
2023-12-20 13:11:05 -05:00
sprintf(path_base, "/sys/devices/system/cpu/intel_uncore_frequency/uncore%02d", i);
if (access(path_base, R_OK))
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
err(1, "%s: %s\n", __func__, path_base);
tools/power turbostat: Expand probe_intel_uncore_frequency() Print current frequency along with the current (and initial) limits Probe and print uncore config also for machines using the new cluster API Signed-off-by: Len Brown <len.brown@intel.com>
2023-12-20 13:11:05 -05:00
sprintf(path, "%s/package_id", path_base);
package_id = read_sysfs_int(path);
sprintf(path, "%s/domain_id", path_base);
domain_id = read_sysfs_int(path);
sprintf(path, "%s/fabric_cluster_id", path_base);
cluster_id = read_sysfs_int(path);
tools/power turbostat: Fix unc freq columns not showing with '-q' or '-l' Commit 78464d7681f7 ("tools/power turbostat: Add columns for clustered uncore frequency") introduced 'probe_intel_uncore_frequency_cluster()' in a way which prevents printing uncore frequency columns if either of the '-q' or '-l' options are used. Systems which do not have multiple uncore frequencies per package are unaffected by this regression. Fix the function so that uncore frequency columns are shown when either the '-l' or '-q' option is used by checking if 'quiet' is true after adding counters for the uncore frequency columns. Fixes: 78464d7681f7 ("tools/power turbostat: Add columns for clustered uncore frequency") Signed-off-by: Adam Hawley <adam.james.hawley@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-22 16:27:21 +03:00
sprintf(path, "%s/current_freq_khz", path_base);
sprintf(name_buf, "UMHz%d.%d", domain_id, cluster_id);
add_counter(0, path, name_buf, 0, SCOPE_PACKAGE, COUNTER_K2M, FORMAT_AVERAGE, 0, package_id);
if (quiet)
continue;
tools/power turbostat: Expand probe_intel_uncore_frequency() Print current frequency along with the current (and initial) limits Probe and print uncore config also for machines using the new cluster API Signed-off-by: Len Brown <len.brown@intel.com>
2023-12-20 13:11:05 -05:00
sprintf(path, "%s/min_freq_khz", path_base);
k = read_sysfs_int(path);
sprintf(path, "%s/max_freq_khz", path_base);
l = read_sysfs_int(path);
fprintf(outf, "Uncore Frequency package%d domain%d cluster%d: %d - %d MHz ", package_id, domain_id,
cluster_id, k / 1000, l / 1000);
sprintf(path, "%s/initial_min_freq_khz", path_base);
k = read_sysfs_int(path);
sprintf(path, "%s/initial_max_freq_khz", path_base);
l = read_sysfs_int(path);
fprintf(outf, "(%d - %d MHz)", k / 1000, l / 1000);
sprintf(path, "%s/current_freq_khz", path_base);
k = read_sysfs_int(path);
fprintf(outf, " %d MHz\n", k / 1000);
}
tools/power turbostat: Show uncore frequency When CONFIG_INTEL_UNCORE_FREQ_CONTROL is effective, (Linux 5.9 and later), print the current (and default) min and max uncore frequency limits. When that driver provides the current uncore frequency (Linux 5.18 and later), print a UncMHz column reflecting the current uncore frequency. Note that UncMHz is an instantaneous sample, not an average. eg. $ sudo ./turbostat -S --show frequency ... Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz) ... Avg_MHz Busy% Bzy_MHz TSC_MHz UncMHz 28 0.70 4049 3095 3900 Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-12 21:35:39 -10:00
}
tools/power turbostat: Harden probe_intel_uncore_frequency() If sysfs directory "intel_uncore_frequency/cluster00/" exists, then use uncore cluster code (now its own routine). The previous check for "intel_uncore_frequency/package_00_die_00/current_freq_khz", could be unreliable in the face of sparse die id's. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 14:45:10 -04:00
static void probe_intel_uncore_frequency(void)
{
if (!genuine_intel)
return;
if (access("/sys/devices/system/cpu/intel_uncore_frequency/uncore00", R_OK) == 0)
probe_intel_uncore_frequency_cluster();
else
probe_intel_uncore_frequency_legacy();
}
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
static void set_graphics_fp(char *path, int idx)
{
if (!access(path, R_OK))
gfx_info[idx].fp = fopen_or_die(path, "r");
}
tools/power turbostat: Add support for /sys/class/drm/card1 On some machines, the graphics device is enumerated as /sys/class/drm/card1 instead of /sys/class/drm/card0. The current implementation does not handle this scenario, resulting in the loss of graphics C6 residency and frequency information. Add support for /sys/class/drm/card1, ensuring that turbostat can retrieve and display the graphics columns for these platforms. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:46 +08:00
/* Enlarge this if there are /sys/class/drm/card2 ... */
#define GFX_MAX_CARDS 2
tools/power/turbostat: Relocate graphics probing code Introduce probe_graphics(), and move all graphics probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 16:01:12 +08:00
static void probe_graphics(void)
{
tools/power turbostat: Add support for /sys/class/drm/card1 On some machines, the graphics device is enumerated as /sys/class/drm/card1 instead of /sys/class/drm/card0. The current implementation does not handle this scenario, resulting in the loss of graphics C6 residency and frequency information. Add support for /sys/class/drm/card1, ensuring that turbostat can retrieve and display the graphics columns for these platforms. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:46 +08:00
char path[PATH_MAX];
int i;
tools/power/turbostat: Add support for Xe sysfs knobs Xe graphics driver uses different graphics sysfs knobs including /sys/class/drm/card0/device/tile0/gt0/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt0/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt0/freq0/act_freq /sys/class/drm/card0/device/tile0/gt1/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt1/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt1/freq0/act_freq Plus that, /sys/class/drm/card0/device/tile0/gt<n>/gtidle/name returns either gt<n>-rc or gt<n>-mc. rc is for GFX and mc is SA Media. Enhance turbostat to prefer the Xe sysfs knobs when they are available. Export gt<n>-rc via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt<n>-mc via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-12 23:56:02 +08:00
/* Xe graphics sysfs knobs */
if (!access("/sys/class/drm/card0/device/tile0/gt0/gtidle/idle_residency_ms", R_OK)) {
FILE *fp;
char buf[8];
bool gt0_is_gt;
fp = fopen("/sys/class/drm/card0/device/tile0/gt0/gtidle/name", "r");
if (!fp)
goto next;
if (!fread(buf, sizeof(char), 7, fp)) {
fclose(fp);
goto next;
}
fclose(fp);
if (!strncmp(buf, "gt0-rc", strlen("gt0-rc")))
gt0_is_gt = true;
else if (!strncmp(buf, "gt0-mc", strlen("gt0-mc")))
gt0_is_gt = false;
else
goto next;
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
set_graphics_fp("/sys/class/drm/card0/device/tile0/gt0/gtidle/idle_residency_ms", gt0_is_gt ? GFX_rc6 : SAM_mc6);
tools/power/turbostat: Add support for Xe sysfs knobs Xe graphics driver uses different graphics sysfs knobs including /sys/class/drm/card0/device/tile0/gt0/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt0/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt0/freq0/act_freq /sys/class/drm/card0/device/tile0/gt1/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt1/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt1/freq0/act_freq Plus that, /sys/class/drm/card0/device/tile0/gt<n>/gtidle/name returns either gt<n>-rc or gt<n>-mc. rc is for GFX and mc is SA Media. Enhance turbostat to prefer the Xe sysfs knobs when they are available. Export gt<n>-rc via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt<n>-mc via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-12 23:56:02 +08:00
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
set_graphics_fp("/sys/class/drm/card0/device/tile0/gt0/freq0/cur_freq", gt0_is_gt ? GFX_MHz : SAM_MHz);
tools/power/turbostat: Add support for Xe sysfs knobs Xe graphics driver uses different graphics sysfs knobs including /sys/class/drm/card0/device/tile0/gt0/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt0/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt0/freq0/act_freq /sys/class/drm/card0/device/tile0/gt1/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt1/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt1/freq0/act_freq Plus that, /sys/class/drm/card0/device/tile0/gt<n>/gtidle/name returns either gt<n>-rc or gt<n>-mc. rc is for GFX and mc is SA Media. Enhance turbostat to prefer the Xe sysfs knobs when they are available. Export gt<n>-rc via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt<n>-mc via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-12 23:56:02 +08:00
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
set_graphics_fp("/sys/class/drm/card0/device/tile0/gt0/freq0/act_freq", gt0_is_gt ? GFX_ACTMHz : SAM_ACTMHz);
tools/power/turbostat: Add support for Xe sysfs knobs Xe graphics driver uses different graphics sysfs knobs including /sys/class/drm/card0/device/tile0/gt0/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt0/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt0/freq0/act_freq /sys/class/drm/card0/device/tile0/gt1/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt1/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt1/freq0/act_freq Plus that, /sys/class/drm/card0/device/tile0/gt<n>/gtidle/name returns either gt<n>-rc or gt<n>-mc. rc is for GFX and mc is SA Media. Enhance turbostat to prefer the Xe sysfs knobs when they are available. Export gt<n>-rc via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt<n>-mc via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-12 23:56:02 +08:00
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
set_graphics_fp("/sys/class/drm/card0/device/tile0/gt1/gtidle/idle_residency_ms", gt0_is_gt ? SAM_mc6 : GFX_rc6);
tools/power/turbostat: Add support for Xe sysfs knobs Xe graphics driver uses different graphics sysfs knobs including /sys/class/drm/card0/device/tile0/gt0/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt0/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt0/freq0/act_freq /sys/class/drm/card0/device/tile0/gt1/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt1/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt1/freq0/act_freq Plus that, /sys/class/drm/card0/device/tile0/gt<n>/gtidle/name returns either gt<n>-rc or gt<n>-mc. rc is for GFX and mc is SA Media. Enhance turbostat to prefer the Xe sysfs knobs when they are available. Export gt<n>-rc via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt<n>-mc via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-12 23:56:02 +08:00
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
set_graphics_fp("/sys/class/drm/card0/device/tile0/gt1/freq0/cur_freq", gt0_is_gt ? SAM_MHz : GFX_MHz);
tools/power/turbostat: Add support for Xe sysfs knobs Xe graphics driver uses different graphics sysfs knobs including /sys/class/drm/card0/device/tile0/gt0/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt0/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt0/freq0/act_freq /sys/class/drm/card0/device/tile0/gt1/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt1/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt1/freq0/act_freq Plus that, /sys/class/drm/card0/device/tile0/gt<n>/gtidle/name returns either gt<n>-rc or gt<n>-mc. rc is for GFX and mc is SA Media. Enhance turbostat to prefer the Xe sysfs knobs when they are available. Export gt<n>-rc via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt<n>-mc via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-12 23:56:02 +08:00
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
set_graphics_fp("/sys/class/drm/card0/device/tile0/gt1/freq0/act_freq", gt0_is_gt ? SAM_ACTMHz : GFX_ACTMHz);
tools/power/turbostat: Add support for Xe sysfs knobs Xe graphics driver uses different graphics sysfs knobs including /sys/class/drm/card0/device/tile0/gt0/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt0/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt0/freq0/act_freq /sys/class/drm/card0/device/tile0/gt1/gtidle/idle_residency_ms /sys/class/drm/card0/device/tile0/gt1/freq0/cur_freq /sys/class/drm/card0/device/tile0/gt1/freq0/act_freq Plus that, /sys/class/drm/card0/device/tile0/gt<n>/gtidle/name returns either gt<n>-rc or gt<n>-mc. rc is for GFX and mc is SA Media. Enhance turbostat to prefer the Xe sysfs knobs when they are available. Export gt<n>-rc via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt<n>-mc via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-12 23:56:02 +08:00
goto end;
}
next:
tools/power/turbostat: Add support for new i915 sysfs knobs On Meteorlake platform, i915 driver supports the traditional graphics sysfs knobs including /sys/class/drm/card0/power/rc6_residency_ms /sys/class/drm/card0/gt_cur_freq_mhz /sys/class/drm/card0/gt_act_freq_mhz At the same time, it also supports /sys/class/drm/card0/gt/gt0/rc6_residency_ms /sys/class/drm/card0/gt/gt0/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt0/rps_act_freq_mhz /sys/class/drm/card0/gt/gt1/rc6_residency_ms /sys/class/drm/card0/gt/gt1/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt1/rps_act_freq_mhz gt0 is for GFX and gt1 is for SA Media. Enhance turbostat to prefer the i915 new sysfs knobs. Export gt0 via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt1 via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-22 09:52:24 +08:00
/* New i915 graphics sysfs knobs */
tools/power turbostat: Add support for /sys/class/drm/card1 On some machines, the graphics device is enumerated as /sys/class/drm/card1 instead of /sys/class/drm/card0. The current implementation does not handle this scenario, resulting in the loss of graphics C6 residency and frequency information. Add support for /sys/class/drm/card1, ensuring that turbostat can retrieve and display the graphics columns for these platforms. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:46 +08:00
for (i = 0; i < GFX_MAX_CARDS; i++) {
snprintf(path, PATH_MAX, "/sys/class/drm/card%d/gt/gt0/rc6_residency_ms", i);
if (!access(path, R_OK))
break;
}
tools/power/turbostat: Add support for new i915 sysfs knobs On Meteorlake platform, i915 driver supports the traditional graphics sysfs knobs including /sys/class/drm/card0/power/rc6_residency_ms /sys/class/drm/card0/gt_cur_freq_mhz /sys/class/drm/card0/gt_act_freq_mhz At the same time, it also supports /sys/class/drm/card0/gt/gt0/rc6_residency_ms /sys/class/drm/card0/gt/gt0/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt0/rps_act_freq_mhz /sys/class/drm/card0/gt/gt1/rc6_residency_ms /sys/class/drm/card0/gt/gt1/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt1/rps_act_freq_mhz gt0 is for GFX and gt1 is for SA Media. Enhance turbostat to prefer the i915 new sysfs knobs. Export gt0 via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt1 via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-22 09:52:24 +08:00
tools/power turbostat: Add support for /sys/class/drm/card1 On some machines, the graphics device is enumerated as /sys/class/drm/card1 instead of /sys/class/drm/card0. The current implementation does not handle this scenario, resulting in the loss of graphics C6 residency and frequency information. Add support for /sys/class/drm/card1, ensuring that turbostat can retrieve and display the graphics columns for these platforms. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:46 +08:00
if (i == GFX_MAX_CARDS)
goto legacy_i915;
tools/power/turbostat: Add support for new i915 sysfs knobs On Meteorlake platform, i915 driver supports the traditional graphics sysfs knobs including /sys/class/drm/card0/power/rc6_residency_ms /sys/class/drm/card0/gt_cur_freq_mhz /sys/class/drm/card0/gt_act_freq_mhz At the same time, it also supports /sys/class/drm/card0/gt/gt0/rc6_residency_ms /sys/class/drm/card0/gt/gt0/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt0/rps_act_freq_mhz /sys/class/drm/card0/gt/gt1/rc6_residency_ms /sys/class/drm/card0/gt/gt1/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt1/rps_act_freq_mhz gt0 is for GFX and gt1 is for SA Media. Enhance turbostat to prefer the i915 new sysfs knobs. Export gt0 via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt1 via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-22 09:52:24 +08:00
tools/power turbostat: Add support for /sys/class/drm/card1 On some machines, the graphics device is enumerated as /sys/class/drm/card1 instead of /sys/class/drm/card0. The current implementation does not handle this scenario, resulting in the loss of graphics C6 residency and frequency information. Add support for /sys/class/drm/card1, ensuring that turbostat can retrieve and display the graphics columns for these platforms. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:46 +08:00
snprintf(path, PATH_MAX, "/sys/class/drm/card%d/gt/gt0/rc6_residency_ms", i);
set_graphics_fp(path, GFX_rc6);
tools/power/turbostat: Add support for new i915 sysfs knobs On Meteorlake platform, i915 driver supports the traditional graphics sysfs knobs including /sys/class/drm/card0/power/rc6_residency_ms /sys/class/drm/card0/gt_cur_freq_mhz /sys/class/drm/card0/gt_act_freq_mhz At the same time, it also supports /sys/class/drm/card0/gt/gt0/rc6_residency_ms /sys/class/drm/card0/gt/gt0/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt0/rps_act_freq_mhz /sys/class/drm/card0/gt/gt1/rc6_residency_ms /sys/class/drm/card0/gt/gt1/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt1/rps_act_freq_mhz gt0 is for GFX and gt1 is for SA Media. Enhance turbostat to prefer the i915 new sysfs knobs. Export gt0 via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt1 via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-22 09:52:24 +08:00
tools/power turbostat: Add support for /sys/class/drm/card1 On some machines, the graphics device is enumerated as /sys/class/drm/card1 instead of /sys/class/drm/card0. The current implementation does not handle this scenario, resulting in the loss of graphics C6 residency and frequency information. Add support for /sys/class/drm/card1, ensuring that turbostat can retrieve and display the graphics columns for these platforms. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:46 +08:00
snprintf(path, PATH_MAX, "/sys/class/drm/card%d/gt/gt0/rps_cur_freq_mhz", i);
set_graphics_fp(path, GFX_MHz);
tools/power/turbostat: Add support for new i915 sysfs knobs On Meteorlake platform, i915 driver supports the traditional graphics sysfs knobs including /sys/class/drm/card0/power/rc6_residency_ms /sys/class/drm/card0/gt_cur_freq_mhz /sys/class/drm/card0/gt_act_freq_mhz At the same time, it also supports /sys/class/drm/card0/gt/gt0/rc6_residency_ms /sys/class/drm/card0/gt/gt0/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt0/rps_act_freq_mhz /sys/class/drm/card0/gt/gt1/rc6_residency_ms /sys/class/drm/card0/gt/gt1/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt1/rps_act_freq_mhz gt0 is for GFX and gt1 is for SA Media. Enhance turbostat to prefer the i915 new sysfs knobs. Export gt0 via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt1 via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-22 09:52:24 +08:00
tools/power turbostat: Add support for /sys/class/drm/card1 On some machines, the graphics device is enumerated as /sys/class/drm/card1 instead of /sys/class/drm/card0. The current implementation does not handle this scenario, resulting in the loss of graphics C6 residency and frequency information. Add support for /sys/class/drm/card1, ensuring that turbostat can retrieve and display the graphics columns for these platforms. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:46 +08:00
snprintf(path, PATH_MAX, "/sys/class/drm/card%d/gt/gt0/rps_act_freq_mhz", i);
set_graphics_fp(path, GFX_ACTMHz);
tools/power/turbostat: Add support for new i915 sysfs knobs On Meteorlake platform, i915 driver supports the traditional graphics sysfs knobs including /sys/class/drm/card0/power/rc6_residency_ms /sys/class/drm/card0/gt_cur_freq_mhz /sys/class/drm/card0/gt_act_freq_mhz At the same time, it also supports /sys/class/drm/card0/gt/gt0/rc6_residency_ms /sys/class/drm/card0/gt/gt0/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt0/rps_act_freq_mhz /sys/class/drm/card0/gt/gt1/rc6_residency_ms /sys/class/drm/card0/gt/gt1/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt1/rps_act_freq_mhz gt0 is for GFX and gt1 is for SA Media. Enhance turbostat to prefer the i915 new sysfs knobs. Export gt0 via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt1 via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-22 09:52:24 +08:00
tools/power turbostat: Add support for /sys/class/drm/card1 On some machines, the graphics device is enumerated as /sys/class/drm/card1 instead of /sys/class/drm/card0. The current implementation does not handle this scenario, resulting in the loss of graphics C6 residency and frequency information. Add support for /sys/class/drm/card1, ensuring that turbostat can retrieve and display the graphics columns for these platforms. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:46 +08:00
snprintf(path, PATH_MAX, "/sys/class/drm/card%d/gt/gt1/rc6_residency_ms", i);
set_graphics_fp(path, SAM_mc6);
tools/power/turbostat: Add support for new i915 sysfs knobs On Meteorlake platform, i915 driver supports the traditional graphics sysfs knobs including /sys/class/drm/card0/power/rc6_residency_ms /sys/class/drm/card0/gt_cur_freq_mhz /sys/class/drm/card0/gt_act_freq_mhz At the same time, it also supports /sys/class/drm/card0/gt/gt0/rc6_residency_ms /sys/class/drm/card0/gt/gt0/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt0/rps_act_freq_mhz /sys/class/drm/card0/gt/gt1/rc6_residency_ms /sys/class/drm/card0/gt/gt1/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt1/rps_act_freq_mhz gt0 is for GFX and gt1 is for SA Media. Enhance turbostat to prefer the i915 new sysfs knobs. Export gt0 via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt1 via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-22 09:52:24 +08:00
tools/power turbostat: Add support for /sys/class/drm/card1 On some machines, the graphics device is enumerated as /sys/class/drm/card1 instead of /sys/class/drm/card0. The current implementation does not handle this scenario, resulting in the loss of graphics C6 residency and frequency information. Add support for /sys/class/drm/card1, ensuring that turbostat can retrieve and display the graphics columns for these platforms. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:46 +08:00
snprintf(path, PATH_MAX, "/sys/class/drm/card%d/gt/gt1/rps_cur_freq_mhz", i);
set_graphics_fp(path, SAM_MHz);
snprintf(path, PATH_MAX, "/sys/class/drm/card%d/gt/gt1/rps_act_freq_mhz", i);
set_graphics_fp(path, SAM_ACTMHz);
goto end;
tools/power/turbostat: Add support for new i915 sysfs knobs On Meteorlake platform, i915 driver supports the traditional graphics sysfs knobs including /sys/class/drm/card0/power/rc6_residency_ms /sys/class/drm/card0/gt_cur_freq_mhz /sys/class/drm/card0/gt_act_freq_mhz At the same time, it also supports /sys/class/drm/card0/gt/gt0/rc6_residency_ms /sys/class/drm/card0/gt/gt0/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt0/rps_act_freq_mhz /sys/class/drm/card0/gt/gt1/rc6_residency_ms /sys/class/drm/card0/gt/gt1/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt1/rps_act_freq_mhz gt0 is for GFX and gt1 is for SA Media. Enhance turbostat to prefer the i915 new sysfs knobs. Export gt0 via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt1 via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-22 09:52:24 +08:00
tools/power turbostat: Add support for /sys/class/drm/card1 On some machines, the graphics device is enumerated as /sys/class/drm/card1 instead of /sys/class/drm/card0. The current implementation does not handle this scenario, resulting in the loss of graphics C6 residency and frequency information. Add support for /sys/class/drm/card1, ensuring that turbostat can retrieve and display the graphics columns for these platforms. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:46 +08:00
legacy_i915:
tools/power/turbostat: Add support for new i915 sysfs knobs On Meteorlake platform, i915 driver supports the traditional graphics sysfs knobs including /sys/class/drm/card0/power/rc6_residency_ms /sys/class/drm/card0/gt_cur_freq_mhz /sys/class/drm/card0/gt_act_freq_mhz At the same time, it also supports /sys/class/drm/card0/gt/gt0/rc6_residency_ms /sys/class/drm/card0/gt/gt0/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt0/rps_act_freq_mhz /sys/class/drm/card0/gt/gt1/rc6_residency_ms /sys/class/drm/card0/gt/gt1/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt1/rps_act_freq_mhz gt0 is for GFX and gt1 is for SA Media. Enhance turbostat to prefer the i915 new sysfs knobs. Export gt0 via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt1 via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-22 09:52:24 +08:00
/* Fall back to traditional i915 graphics sysfs knobs */
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
set_graphics_fp("/sys/class/drm/card0/power/rc6_residency_ms", GFX_rc6);
tools/power/turbostat: Relocate graphics probing code Introduce probe_graphics(), and move all graphics probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 16:01:12 +08:00
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
set_graphics_fp("/sys/class/drm/card0/gt_cur_freq_mhz", GFX_MHz);
if (!gfx_info[GFX_MHz].fp)
set_graphics_fp("/sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz", GFX_MHz);
tools/power/turbostat: Cache graphics sysfs path Graphics drivers (i915/Xe) have different sysfs knobs on different platforms, and it is possible that different sysfs knobs fit into the same turbostat columns. Instead of specifying different sysfs knobs every time, detect them once and cache the path for future use. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-12 14:23:37 +08:00
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
set_graphics_fp("/sys/class/drm/card0/gt_act_freq_mhz", GFX_ACTMHz);
if (!gfx_info[GFX_ACTMHz].fp)
set_graphics_fp("/sys/class/graphics/fb0/device/drm/card0/gt_act_freq_mhz", GFX_ACTMHz);
tools/power/turbostat: Relocate graphics probing code Introduce probe_graphics(), and move all graphics probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 16:01:12 +08:00
tools/power/turbostat: Add support for new i915 sysfs knobs On Meteorlake platform, i915 driver supports the traditional graphics sysfs knobs including /sys/class/drm/card0/power/rc6_residency_ms /sys/class/drm/card0/gt_cur_freq_mhz /sys/class/drm/card0/gt_act_freq_mhz At the same time, it also supports /sys/class/drm/card0/gt/gt0/rc6_residency_ms /sys/class/drm/card0/gt/gt0/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt0/rps_act_freq_mhz /sys/class/drm/card0/gt/gt1/rc6_residency_ms /sys/class/drm/card0/gt/gt1/rps_cur_freq_mhz /sys/class/drm/card0/gt/gt1/rps_act_freq_mhz gt0 is for GFX and gt1 is for SA Media. Enhance turbostat to prefer the i915 new sysfs knobs. Export gt0 via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz. Export gt1 via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-22 09:52:24 +08:00
end:
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
if (gfx_info[GFX_rc6].fp)
tools/power/turbostat: Cache graphics sysfs path Graphics drivers (i915/Xe) have different sysfs knobs on different platforms, and it is possible that different sysfs knobs fit into the same turbostat columns. Instead of specifying different sysfs knobs every time, detect them once and cache the path for future use. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-12 14:23:37 +08:00
BIC_PRESENT(BIC_GFX_rc6);
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
if (gfx_info[GFX_MHz].fp)
tools/power/turbostat: Cache graphics sysfs path Graphics drivers (i915/Xe) have different sysfs knobs on different platforms, and it is possible that different sysfs knobs fit into the same turbostat columns. Instead of specifying different sysfs knobs every time, detect them once and cache the path for future use. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-12 14:23:37 +08:00
BIC_PRESENT(BIC_GFXMHz);
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
if (gfx_info[GFX_ACTMHz].fp)
tools/power/turbostat: Relocate graphics probing code Introduce probe_graphics(), and move all graphics probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 16:01:12 +08:00
BIC_PRESENT(BIC_GFXACTMHz);
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
if (gfx_info[SAM_mc6].fp)
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
BIC_PRESENT(BIC_SAM_mc6);
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
if (gfx_info[SAM_MHz].fp)
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
BIC_PRESENT(BIC_SAMMHz);
tools/power turbostat: Cache graphics sysfs file descriptors during probe Snapshots of the graphics sysfs knobs are taken based on file descriptors. To optimize this process, open the files and cache the file descriptors during the graphics probe phase. As a result, the previously cached pathnames become redundant and are removed. This change aims to streamline the code without altering its functionality. No functional change intended. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-14 15:59:45 +08:00
if (gfx_info[SAM_ACTMHz].fp)
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media information via different sysfs knobs. Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for GFX. Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2024-03-13 10:30:04 +08:00
BIC_PRESENT(BIC_SAMACTMHz);
tools/power/turbostat: Relocate graphics probing code Introduce probe_graphics(), and move all graphics probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 16:01:12 +08:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
static void dump_sysfs_cstate_config(void)
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
{
char path[64];
char name_buf[16];
char desc[64];
FILE *input;
int state;
char *sp;
tools/power turbostat: Print cpuidle information Print cpuidle driver and governor. Originally-by: Antti Laakso <antti.laakso@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2020-03-21 00:47:47 -04:00
if (access("/sys/devices/system/cpu/cpuidle", R_OK)) {
fprintf(outf, "cpuidle not loaded\n");
return;
}
dump_sysfs_file("/sys/devices/system/cpu/cpuidle/current_driver");
dump_sysfs_file("/sys/devices/system/cpu/cpuidle/current_governor");
dump_sysfs_file("/sys/devices/system/cpu/cpuidle/current_governor_ro");
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
for (state = 0; state < 10; ++state) {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
sprintf(path, "/sys/devices/system/cpu/cpu%d/cpuidle/state%d/name", base_cpu, state);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
input = fopen(path, "r");
if (input == NULL)
continue;
tools/power turbostat: Add checks for failure of fgets() and fscanf() Most calls to fgets() and fscanf() are followed by error checks. Add an exit-on-error in the remaining cases. Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-20 23:01:03 -04:00
if (!fgets(name_buf, sizeof(name_buf), input))
err(1, "%s: failed to read file", path);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
/* truncate "C1-HSW\n" to "C1", or truncate "C1\n" to "C1" */
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
sp = strchr(name_buf, '-');
if (!sp)
sp = strchrnul(name_buf, '\n');
*sp = '\0';
fclose(input);
tools/power turbostat: Fix output formatting for ACPI CST enumeration turbostat formatting is broken with ACPI CST for enumeration. The problem is that the CX_ACPI% is eight characters long which does not work with tab formatting. One simple solution is to remove the underbar from the state name such that C1_ACPI will be displayed as C1ACPI. Signed-off-by: David Arcari <darcari@redhat.com> Cc: Len Brown <lenb@kernel.org> Cc: linux-kernel@vger.kernel.org Signed-off-by: Len Brown <len.brown@intel.com>
2020-08-10 10:43:30 -04:00
remove_underbar(name_buf);
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
sprintf(path, "/sys/devices/system/cpu/cpu%d/cpuidle/state%d/desc", base_cpu, state);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
input = fopen(path, "r");
if (input == NULL)
continue;
tools/power turbostat: Add checks for failure of fgets() and fscanf() Most calls to fgets() and fscanf() are followed by error checks. Add an exit-on-error in the remaining cases. Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-20 23:01:03 -04:00
if (!fgets(desc, sizeof(desc), input))
err(1, "%s: failed to read file", path);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
fprintf(outf, "cpu%d: %s: %s", base_cpu, name_buf, desc);
fclose(input);
}
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
static void dump_sysfs_pstate_config(void)
tools/power turbostat: dump p-state software config cpu1: cpufreq driver: acpi-cpufreq cpu1: cpufreq governor: ondemand cpufreq boost: 1 or cpu0: cpufreq driver: intel_pstate cpu0: cpufreq governor: powersave cpufreq intel_pstate no_turbo: 0 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-22 00:11:12 -05:00
{
char path[64];
char driver_buf[64];
char governor_buf[64];
FILE *input;
int turbo;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
sprintf(path, "/sys/devices/system/cpu/cpu%d/cpufreq/scaling_driver", base_cpu);
tools/power turbostat: dump p-state software config cpu1: cpufreq driver: acpi-cpufreq cpu1: cpufreq governor: ondemand cpufreq boost: 1 or cpu0: cpufreq driver: intel_pstate cpu0: cpufreq governor: powersave cpufreq intel_pstate no_turbo: 0 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-22 00:11:12 -05:00
input = fopen(path, "r");
if (input == NULL) {
tools/power turbostat: Do not display an error on systems without a cpufreq driver Running without a cpufreq driver is a valid case so warnings output in this case should not be to stderr. Use outf instead of stderr for these warnings. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-13 08:45:01 -04:00
fprintf(outf, "NSFOD %s\n", path);
tools/power turbostat: dump p-state software config cpu1: cpufreq driver: acpi-cpufreq cpu1: cpufreq governor: ondemand cpufreq boost: 1 or cpu0: cpufreq driver: intel_pstate cpu0: cpufreq governor: powersave cpufreq intel_pstate no_turbo: 0 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-22 00:11:12 -05:00
return;
}
tools/power turbostat: Add checks for failure of fgets() and fscanf() Most calls to fgets() and fscanf() are followed by error checks. Add an exit-on-error in the remaining cases. Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-20 23:01:03 -04:00
if (!fgets(driver_buf, sizeof(driver_buf), input))
err(1, "%s: failed to read file", path);
tools/power turbostat: dump p-state software config cpu1: cpufreq driver: acpi-cpufreq cpu1: cpufreq governor: ondemand cpufreq boost: 1 or cpu0: cpufreq driver: intel_pstate cpu0: cpufreq governor: powersave cpufreq intel_pstate no_turbo: 0 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-22 00:11:12 -05:00
fclose(input);
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
sprintf(path, "/sys/devices/system/cpu/cpu%d/cpufreq/scaling_governor", base_cpu);
tools/power turbostat: dump p-state software config cpu1: cpufreq driver: acpi-cpufreq cpu1: cpufreq governor: ondemand cpufreq boost: 1 or cpu0: cpufreq driver: intel_pstate cpu0: cpufreq governor: powersave cpufreq intel_pstate no_turbo: 0 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-22 00:11:12 -05:00
input = fopen(path, "r");
if (input == NULL) {
tools/power turbostat: Do not display an error on systems without a cpufreq driver Running without a cpufreq driver is a valid case so warnings output in this case should not be to stderr. Use outf instead of stderr for these warnings. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-13 08:45:01 -04:00
fprintf(outf, "NSFOD %s\n", path);
tools/power turbostat: dump p-state software config cpu1: cpufreq driver: acpi-cpufreq cpu1: cpufreq governor: ondemand cpufreq boost: 1 or cpu0: cpufreq driver: intel_pstate cpu0: cpufreq governor: powersave cpufreq intel_pstate no_turbo: 0 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-22 00:11:12 -05:00
return;
}
tools/power turbostat: Add checks for failure of fgets() and fscanf() Most calls to fgets() and fscanf() are followed by error checks. Add an exit-on-error in the remaining cases. Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-20 23:01:03 -04:00
if (!fgets(governor_buf, sizeof(governor_buf), input))
err(1, "%s: failed to read file", path);
tools/power turbostat: dump p-state software config cpu1: cpufreq driver: acpi-cpufreq cpu1: cpufreq governor: ondemand cpufreq boost: 1 or cpu0: cpufreq driver: intel_pstate cpu0: cpufreq governor: powersave cpufreq intel_pstate no_turbo: 0 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-22 00:11:12 -05:00
fclose(input);
fprintf(outf, "cpu%d: cpufreq driver: %s", base_cpu, driver_buf);
fprintf(outf, "cpu%d: cpufreq governor: %s", base_cpu, governor_buf);
sprintf(path, "/sys/devices/system/cpu/cpufreq/boost");
input = fopen(path, "r");
if (input != NULL) {
tools/power turbostat: Add checks for failure of fgets() and fscanf() Most calls to fgets() and fscanf() are followed by error checks. Add an exit-on-error in the remaining cases. Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-20 23:01:03 -04:00
if (fscanf(input, "%d", &turbo) != 1)
err(1, "%s: failed to parse number from file", path);
tools/power turbostat: dump p-state software config cpu1: cpufreq driver: acpi-cpufreq cpu1: cpufreq governor: ondemand cpufreq boost: 1 or cpu0: cpufreq driver: intel_pstate cpu0: cpufreq governor: powersave cpufreq intel_pstate no_turbo: 0 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-22 00:11:12 -05:00
fprintf(outf, "cpufreq boost: %d\n", turbo);
fclose(input);
}
sprintf(path, "/sys/devices/system/cpu/intel_pstate/no_turbo");
input = fopen(path, "r");
if (input != NULL) {
tools/power turbostat: Add checks for failure of fgets() and fscanf() Most calls to fgets() and fscanf() are followed by error checks. Add an exit-on-error in the remaining cases. Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-20 23:01:03 -04:00
if (fscanf(input, "%d", &turbo) != 1)
err(1, "%s: failed to parse number from file", path);
tools/power turbostat: dump p-state software config cpu1: cpufreq driver: acpi-cpufreq cpu1: cpufreq governor: ondemand cpufreq boost: 1 or cpu0: cpufreq driver: intel_pstate cpu0: cpufreq governor: powersave cpufreq intel_pstate no_turbo: 0 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-22 00:11:12 -05:00
fprintf(outf, "cpufreq intel_pstate no_turbo: %d\n", turbo);
fclose(input);
}
}
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
/*
* print_epb()
* Decode the ENERGY_PERF_BIAS MSR
*/
int print_epb(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
char *epb_string;
tools/power/turbostat: Read energy_perf_bias from sysfs ... instead of poking at the MSR directly. Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Len Brown <lenb@kernel.org> Cc: linux-pm@vger.kernel.org Link: https://lkml.kernel.org/r/20201029190259.3476-3-bp@alien8.de
2020-10-15 14:50:16 +02:00
int cpu, epb;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: No build warnings with -Wextra Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 23:45:18 -04:00
UNUSED(c);
UNUSED(p);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
if (!has_epb)
return 0;
cpu = t->cpu_id;
/* EPB is per-package */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (!is_cpu_first_thread_in_package(t, c, p))
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
return 0;
if (cpu_migrate(cpu)) {
tools/power turbostat: harden against cpu hotplug turbostat tends to get confused when CPUs are added and removed while it is running. There are races, such as checking the current cpu, and then reading a sysfs file that depends on that cpu number. Close the two issues that seem to come up the most. First, there is an infinite reset loop detector -- change that to allow more resets before giving up. Secondly, one of those file reads didn't really need to exit the program on failure... Signed-off-by: Len Brown <len.brown@intel.com>
2020-09-30 20:58:15 -04:00
fprintf(outf, "print_epb: Could not migrate to CPU %d\n", cpu);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
return -1;
}
tools/power/turbostat: Read energy_perf_bias from sysfs ... instead of poking at the MSR directly. Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Len Brown <lenb@kernel.org> Cc: linux-pm@vger.kernel.org Link: https://lkml.kernel.org/r/20201029190259.3476-3-bp@alien8.de
2020-10-15 14:50:16 +02:00
epb = get_epb(cpu);
if (epb < 0)
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
return 0;
tools/power/turbostat: Read energy_perf_bias from sysfs ... instead of poking at the MSR directly. Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Len Brown <lenb@kernel.org> Cc: linux-pm@vger.kernel.org Link: https://lkml.kernel.org/r/20201029190259.3476-3-bp@alien8.de
2020-10-15 14:50:16 +02:00
switch (epb) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
case ENERGY_PERF_BIAS_PERFORMANCE:
epb_string = "performance";
break;
case ENERGY_PERF_BIAS_NORMAL:
epb_string = "balanced";
break;
case ENERGY_PERF_BIAS_POWERSAVE:
epb_string = "powersave";
break;
default:
epb_string = "custom";
break;
}
tools/power/turbostat: Read energy_perf_bias from sysfs ... instead of poking at the MSR directly. Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Len Brown <lenb@kernel.org> Cc: linux-pm@vger.kernel.org Link: https://lkml.kernel.org/r/20201029190259.3476-3-bp@alien8.de
2020-10-15 14:50:16 +02:00
fprintf(outf, "cpu%d: EPB: %d (%s)\n", cpu, epb, epb_string);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
return 0;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
/*
* print_hwp()
* Decode the MSR_HWP_CAPABILITIES
*/
int print_hwp(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
unsigned long long msr;
int cpu;
tools/power turbostat: No build warnings with -Wextra Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 23:45:18 -04:00
UNUSED(c);
UNUSED(p);
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (no_msr)
return 0;
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
if (!has_hwp)
return 0;
cpu = t->cpu_id;
/* MSR_HWP_CAPABILITIES is per-package */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (!is_cpu_first_thread_in_package(t, c, p))
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
return 0;
if (cpu_migrate(cpu)) {
tools/power turbostat: harden against cpu hotplug turbostat tends to get confused when CPUs are added and removed while it is running. There are races, such as checking the current cpu, and then reading a sysfs file that depends on that cpu number. Close the two issues that seem to come up the most. First, there is an infinite reset loop detector -- change that to allow more resets before giving up. Secondly, one of those file reads didn't really need to exit the program on failure... Signed-off-by: Len Brown <len.brown@intel.com>
2020-09-30 20:58:15 -04:00
fprintf(outf, "print_hwp: Could not migrate to CPU %d\n", cpu);
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
return -1;
}
if (get_msr(cpu, MSR_PM_ENABLE, &msr))
return 0;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "cpu%d: MSR_PM_ENABLE: 0x%08llx (%sHWP)\n", cpu, msr, (msr & (1 << 0)) ? "" : "No-");
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
/* MSR_PM_ENABLE[1] == 1 if HWP is enabled and MSRs visible */
if ((msr & (1 << 0)) == 0)
return 0;
if (get_msr(cpu, MSR_HWP_CAPABILITIES, &msr))
return 0;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_HWP_CAPABILITIES: 0x%08llx "
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
"(high %d guar %d eff %d low %d)\n",
cpu, msr,
(unsigned int)HWP_HIGHEST_PERF(msr),
(unsigned int)HWP_GUARANTEED_PERF(msr),
(unsigned int)HWP_MOSTEFFICIENT_PERF(msr), (unsigned int)HWP_LOWEST_PERF(msr));
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
if (get_msr(cpu, MSR_HWP_REQUEST, &msr))
return 0;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_HWP_REQUEST: 0x%08llx "
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
"(min %d max %d des %d epp 0x%x window 0x%x pkg 0x%x)\n",
cpu, msr,
(unsigned int)(((msr) >> 0) & 0xff),
(unsigned int)(((msr) >> 8) & 0xff),
(unsigned int)(((msr) >> 16) & 0xff),
(unsigned int)(((msr) >> 24) & 0xff),
(unsigned int)(((msr) >> 32) & 0xff3), (unsigned int)(((msr) >> 42) & 0x1));
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
if (has_hwp_pkg) {
if (get_msr(cpu, MSR_HWP_REQUEST_PKG, &msr))
return 0;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_HWP_REQUEST_PKG: 0x%08llx "
tools/power turbostat: update HWP dump to decimal from hex Syntax only. The HWP CAPABILTIES and REQUEST ratios are more easily viewed in decimal -- just multiply by 100 and you get MHz... new: cpu0: MSR_HWP_CAPABILITIES: 0x010c1b23 (high 35 guar 27 eff 12 low 1) cpu0: MSR_HWP_REQUEST: 0x80002301 (min 1 max 35 des 0 epp 0x80 window 0x0 pkg 0x0) old: cpu0: MSR_HWP_CAPABILITIES: 0x010c1b23 (high 0x23 guar 0x1b eff 0xc low 0x1) cpu0: MSR_HWP_REQUEST: 0x80002301 (min 0x1 max 0x23 des 0x0 epp 0x80 window 0x0 pkg 0x0) Signed-off-by: Len Brown <len.brown@intel.com>
2017-03-04 18:18:28 -05:00
"(min %d max %d des %d epp 0x%x window 0x%x)\n",
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
cpu, msr,
(unsigned int)(((msr) >> 0) & 0xff),
(unsigned int)(((msr) >> 8) & 0xff),
(unsigned int)(((msr) >> 16) & 0xff),
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
(unsigned int)(((msr) >> 24) & 0xff), (unsigned int)(((msr) >> 32) & 0xff3));
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
}
if (has_hwp_notify) {
if (get_msr(cpu, MSR_HWP_INTERRUPT, &msr))
return 0;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_HWP_INTERRUPT: 0x%08llx "
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
"(%s_Guaranteed_Perf_Change, %s_Excursion_Min)\n",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
cpu, msr, ((msr) & 0x1) ? "EN" : "Dis", ((msr) & 0x2) ? "EN" : "Dis");
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
}
if (get_msr(cpu, MSR_HWP_STATUS, &msr))
return 0;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_HWP_STATUS: 0x%08llx "
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
"(%sGuaranteed_Perf_Change, %sExcursion_Min)\n",
tools/power turbostat: fix decoding of HWP_STATUS The "excursion to minimum" information is in bit2 in HWP_STATUS MSR. Fix the bitmask used for decoding the register. Signed-off-by: Antti Laakso <antti.laakso@intel.com> Reviewed-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2023-01-25 15:17:50 +02:00
cpu, msr, ((msr) & 0x1) ? "" : "No-", ((msr) & 0x4) ? "" : "No-");
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
return 0;
}
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
/*
* print_perf_limit()
*/
int print_perf_limit(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
unsigned long long msr;
int cpu;
tools/power turbostat: No build warnings with -Wextra Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 23:45:18 -04:00
UNUSED(c);
UNUSED(p);
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (no_msr)
return 0;
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
cpu = t->cpu_id;
/* per-package */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (!is_cpu_first_thread_in_package(t, c, p))
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
return 0;
if (cpu_migrate(cpu)) {
tools/power turbostat: harden against cpu hotplug turbostat tends to get confused when CPUs are added and removed while it is running. There are races, such as checking the current cpu, and then reading a sysfs file that depends on that cpu number. Close the two issues that seem to come up the most. First, there is an infinite reset loop detector -- change that to allow more resets before giving up. Secondly, one of those file reads didn't really need to exit the program on failure... Signed-off-by: Len Brown <len.brown@intel.com>
2020-09-30 20:58:15 -04:00
fprintf(outf, "print_perf_limit: Could not migrate to CPU %d\n", cpu);
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
return -1;
}
tools/power/turbostat: Abstract Perf Limit Reasons MSRs support Abstract the support for MSR_CORE/GFX/RING_PERF_LIMIT_REASONS MSRs. Delete perf_limit_reasons_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-07-30 14:25:17 +08:00
if (platform->plr_msrs & PLR_CORE) {
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
get_msr(cpu, MSR_CORE_PERF_LIMIT_REASONS, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_CORE_PERF_LIMIT_REASONS, 0x%08llx", cpu, msr);
fprintf(outf, " (Active: %s%s%s%s%s%s%s%s%s%s%s%s%s%s)",
tools/power turbostat: update PERF_LIMIT_REASONS decoding cosmetic only. order the decoding of MSR_PERF_LIMIT_REASONS bits from MSB to LSB -- which you notice when more than 1 bit is set and you are, say, comparing the output to the documentation... Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-13 16:30:57 -04:00
(msr & 1 << 15) ? "bit15, " : "",
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
(msr & 1 << 14) ? "bit14, " : "",
tools/power turbostat: update PERF_LIMIT_REASONS decoding cosmetic only. order the decoding of MSR_PERF_LIMIT_REASONS bits from MSB to LSB -- which you notice when more than 1 bit is set and you are, say, comparing the output to the documentation... Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-13 16:30:57 -04:00
(msr & 1 << 13) ? "Transitions, " : "",
(msr & 1 << 12) ? "MultiCoreTurbo, " : "",
(msr & 1 << 11) ? "PkgPwrL2, " : "",
(msr & 1 << 10) ? "PkgPwrL1, " : "",
(msr & 1 << 9) ? "CorePwr, " : "",
(msr & 1 << 8) ? "Amps, " : "",
(msr & 1 << 6) ? "VR-Therm, " : "",
(msr & 1 << 5) ? "Auto-HWP, " : "",
(msr & 1 << 4) ? "Graphics, " : "",
(msr & 1 << 2) ? "bit2, " : "",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
(msr & 1 << 1) ? "ThermStatus, " : "", (msr & 1 << 0) ? "PROCHOT, " : "");
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, " (Logged: %s%s%s%s%s%s%s%s%s%s%s%s%s%s)\n",
tools/power turbostat: update PERF_LIMIT_REASONS decoding cosmetic only. order the decoding of MSR_PERF_LIMIT_REASONS bits from MSB to LSB -- which you notice when more than 1 bit is set and you are, say, comparing the output to the documentation... Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-13 16:30:57 -04:00
(msr & 1 << 31) ? "bit31, " : "",
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
(msr & 1 << 30) ? "bit30, " : "",
tools/power turbostat: update PERF_LIMIT_REASONS decoding cosmetic only. order the decoding of MSR_PERF_LIMIT_REASONS bits from MSB to LSB -- which you notice when more than 1 bit is set and you are, say, comparing the output to the documentation... Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-13 16:30:57 -04:00
(msr & 1 << 29) ? "Transitions, " : "",
(msr & 1 << 28) ? "MultiCoreTurbo, " : "",
(msr & 1 << 27) ? "PkgPwrL2, " : "",
(msr & 1 << 26) ? "PkgPwrL1, " : "",
(msr & 1 << 25) ? "CorePwr, " : "",
(msr & 1 << 24) ? "Amps, " : "",
(msr & 1 << 22) ? "VR-Therm, " : "",
(msr & 1 << 21) ? "Auto-HWP, " : "",
(msr & 1 << 20) ? "Graphics, " : "",
(msr & 1 << 18) ? "bit18, " : "",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
(msr & 1 << 17) ? "ThermStatus, " : "", (msr & 1 << 16) ? "PROCHOT, " : "");
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
}
tools/power/turbostat: Abstract Perf Limit Reasons MSRs support Abstract the support for MSR_CORE/GFX/RING_PERF_LIMIT_REASONS MSRs. Delete perf_limit_reasons_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-07-30 14:25:17 +08:00
if (platform->plr_msrs & PLR_GFX) {
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
get_msr(cpu, MSR_GFX_PERF_LIMIT_REASONS, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_GFX_PERF_LIMIT_REASONS, 0x%08llx", cpu, msr);
fprintf(outf, " (Active: %s%s%s%s%s%s%s%s)",
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
(msr & 1 << 0) ? "PROCHOT, " : "",
(msr & 1 << 1) ? "ThermStatus, " : "",
(msr & 1 << 4) ? "Graphics, " : "",
(msr & 1 << 6) ? "VR-Therm, " : "",
(msr & 1 << 8) ? "Amps, " : "",
(msr & 1 << 9) ? "GFXPwr, " : "",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
(msr & 1 << 10) ? "PkgPwrL1, " : "", (msr & 1 << 11) ? "PkgPwrL2, " : "");
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, " (Logged: %s%s%s%s%s%s%s%s)\n",
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
(msr & 1 << 16) ? "PROCHOT, " : "",
(msr & 1 << 17) ? "ThermStatus, " : "",
(msr & 1 << 20) ? "Graphics, " : "",
(msr & 1 << 22) ? "VR-Therm, " : "",
(msr & 1 << 24) ? "Amps, " : "",
(msr & 1 << 25) ? "GFXPwr, " : "",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
(msr & 1 << 26) ? "PkgPwrL1, " : "", (msr & 1 << 27) ? "PkgPwrL2, " : "");
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
}
tools/power/turbostat: Abstract Perf Limit Reasons MSRs support Abstract the support for MSR_CORE/GFX/RING_PERF_LIMIT_REASONS MSRs. Delete perf_limit_reasons_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-07-30 14:25:17 +08:00
if (platform->plr_msrs & PLR_RING) {
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
get_msr(cpu, MSR_RING_PERF_LIMIT_REASONS, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_RING_PERF_LIMIT_REASONS, 0x%08llx", cpu, msr);
fprintf(outf, " (Active: %s%s%s%s%s%s)",
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
(msr & 1 << 0) ? "PROCHOT, " : "",
(msr & 1 << 1) ? "ThermStatus, " : "",
(msr & 1 << 6) ? "VR-Therm, " : "",
(msr & 1 << 8) ? "Amps, " : "",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
(msr & 1 << 10) ? "PkgPwrL1, " : "", (msr & 1 << 11) ? "PkgPwrL2, " : "");
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, " (Logged: %s%s%s%s%s%s)\n",
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
(msr & 1 << 16) ? "PROCHOT, " : "",
(msr & 1 << 17) ? "ThermStatus, " : "",
(msr & 1 << 22) ? "VR-Therm, " : "",
(msr & 1 << 24) ? "Amps, " : "",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
(msr & 1 << 26) ? "PkgPwrL1, " : "", (msr & 1 << 27) ? "PkgPwrL2, " : "");
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 02:39:52 -04:00
}
return 0;
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
#define RAPL_POWER_GRANULARITY 0x7FFF /* 15 bit power granularity */
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
#define RAPL_TIME_GRANULARITY 0x3F /* 6 bit time granularity */
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power/turbostat: Abstract hardcoded TDP value Different hardcoded TDP values are used when TDP can not be retrieved from the hardware. Abstract hardcoded TDP value. Delete CPU model checks in get_tdp_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:21:10 +08:00
double get_quirk_tdp(void)
{
if (platform->rapl_quirk_tdp)
return platform->rapl_quirk_tdp;
return 135.0;
}
tools/power/turbostat: Remove unused family/model parameters for RAPL functions RAPL probing can be done without family/model checking. Remove these parameters in rapl probe functions. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:28:04 +08:00
double get_tdp_intel(void)
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
{
unsigned long long msr;
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (platform->rapl_msrs & RAPL_PKG_POWER_INFO)
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 08:34:28 -04:00
if (!get_msr(base_cpu, MSR_PKG_POWER_INFO, &msr))
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
return ((msr >> 0) & RAPL_POWER_GRANULARITY) * rapl_power_units;
tools/power/turbostat: Abstract hardcoded TDP value Different hardcoded TDP values are used when TDP can not be retrieved from the hardware. Abstract hardcoded TDP value. Delete CPU model checks in get_tdp_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:21:10 +08:00
return get_quirk_tdp();
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
}
tools/power/turbostat: Remove unused family/model parameters for RAPL functions RAPL probing can be done without family/model checking. Remove these parameters in rapl probe functions. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:28:04 +08:00
double get_tdp_amd(void)
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
{
tools/power/turbostat: Abstract hardcoded TDP value Different hardcoded TDP values are used when TDP can not be retrieved from the hardware. Abstract hardcoded TDP value. Delete CPU model checks in get_tdp_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:21:10 +08:00
return get_quirk_tdp();
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
}
tools/power/turbostat: Remove unused family/model parameters for RAPL functions RAPL probing can be done without family/model checking. Remove these parameters in rapl probe functions. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:28:04 +08:00
void rapl_probe_intel(void)
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
{
unsigned long long msr;
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
unsigned int time_unit;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
double tdp;
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
const unsigned long long bic_watt_bits = BIC_SysWatt | BIC_PkgWatt | BIC_CorWatt | BIC_RAMWatt | BIC_GFXWatt;
const unsigned long long bic_joules_bits = BIC_Sys_J | BIC_Pkg_J | BIC_Cor_J | BIC_RAM_J | BIC_GFX_J;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
if (rapl_joules)
bic_enabled &= ~bic_watt_bits;
else
bic_enabled &= ~bic_joules_bits;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
if (!(platform->rapl_msrs & RAPL_PKG_PERF_STATUS))
bic_enabled &= ~BIC_PKG__;
if (!(platform->rapl_msrs & RAPL_DRAM_PERF_STATUS))
bic_enabled &= ~BIC_RAM__;
tools/power/turbostat: Simplify the logic for RAPL enumeration The support for each RAPL domains, as well as the support for the perf status of each RAPL domains, can be detected by checking the availabilities of the corresponding RAPL MSRs. Change the code accordingly and remove the hardcoded logic for each model. Note that this also fixes the INTEL_FAM6_ATOM_TREMONT model, which has RAPL_PKG_PERF_STATUS and MSR_DRAM_PERF_STATUS but doesn't have BIC_PKG__ and BIC_RAM__ set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-26 14:57:12 +08:00
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
/* units on package 0, verify later other packages match */
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 08:34:28 -04:00
if (get_msr(base_cpu, MSR_RAPL_POWER_UNIT, &msr))
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
return;
rapl_power_units = 1.0 / (1 << (msr & 0xF));
tools/power/turbostat: Abstract RAPL divisor support INTEL_FAM6_ATOM_SILVERMONT model needs a divisor to convert the raw Energy Units value from MSR_RAPL_POWER_UNIT. Abstract the support for RAPL divisor. Delete CPU model check in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 11:59:04 +08:00
if (platform->has_rapl_divisor)
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
rapl_energy_units = 1.0 * (1 << (msr >> 8 & 0x1F)) / 1000000;
else
rapl_energy_units = 1.0 / (1 << (msr >> 8 & 0x1F));
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power/turbostat: Abstract fixed DRAM Energy unit support Abstract the support for fixed Dram domain energy unit. Delete rapl_dram_energy_units_probe() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 12:21:58 +08:00
if (platform->has_fixed_rapl_unit)
rapl_dram_energy_units = (15.3 / 1000000);
else
rapl_dram_energy_units = rapl_energy_units;
tools/power turbostat: correct DRAM RAPL units on recent Xeon processors While not yet documented in the Software Developer's Manual, the data-sheet for modern Xeon states that DRAM RAPL ENERGY units are fixed at 15.3 uJ, rather than being discovered via MSR. Before this patch, DRAM energy on these products is over-stated by turbostat because the RAPL units are 4x larger. ref: "Xeon E5-2600 v3/E5-1600 v3 Datasheet Volume 2" http://www.intel.com/content/dam/www/public/us/en/documents/datasheets/xeon-e5-v3-datasheet-vol-2.pdf Signed-off-by: Andrey Semin <andrey.semin@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-12-05 00:07:00 -05:00
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
time_unit = msr >> 16 & 0xF;
if (time_unit == 0)
time_unit = 0xA;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
rapl_time_units = 1.0 / (1 << (time_unit));
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power/turbostat: Remove unused family/model parameters for RAPL functions RAPL probing can be done without family/model checking. Remove these parameters in rapl probe functions. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:28:04 +08:00
tdp = get_tdp_intel();
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
rapl_joule_counter_range = 0xFFFFFFFF * rapl_energy_units / tdp;
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
if (!quiet)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "RAPL: %.0f sec. Joule Counter Range, at %.0f Watts\n", rapl_joule_counter_range, tdp);
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power/turbostat: Remove unused family/model parameters for RAPL functions RAPL probing can be done without family/model checking. Remove these parameters in rapl probe functions. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:28:04 +08:00
void rapl_probe_amd(void)
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
{
unsigned long long msr;
double tdp;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
const unsigned long long bic_watt_bits = BIC_PkgWatt | BIC_CorWatt;
const unsigned long long bic_joules_bits = BIC_Pkg_J | BIC_Cor_J;
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
if (rapl_joules)
bic_enabled &= ~bic_watt_bits;
else
bic_enabled &= ~bic_joules_bits;
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
if (get_msr(base_cpu, MSR_RAPL_PWR_UNIT, &msr))
return;
rapl_time_units = ldexp(1.0, -(msr >> 16 & 0xf));
rapl_energy_units = ldexp(1.0, -(msr >> 8 & 0x1f));
rapl_power_units = ldexp(1.0, -(msr & 0xf));
tools/power/turbostat: Remove unused family/model parameters for RAPL functions RAPL probing can be done without family/model checking. Remove these parameters in rapl probe functions. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 10:28:04 +08:00
tdp = get_tdp_amd();
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
rapl_joule_counter_range = 0xFFFFFFFF * rapl_energy_units / tdp;
if (!quiet)
fprintf(outf, "RAPL: %.0f sec. Joule Counter Range, at %.0f Watts\n", rapl_joule_counter_range, tdp);
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
void print_power_limit_msr(int cpu, unsigned long long msr, char *label)
{
tools/power turbostat: print power values upto three decimal Print power values upto three decimal places in watts. Suggested-by: Tony Luck <tony.luck@intel.com> Signed-off-by: Sumeet Pawnikar <sumeet.r.pawnikar@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-09-30 18:40:18 +05:30
fprintf(outf, "cpu%d: %s: %sabled (%0.3f Watts, %f sec, clamp %sabled)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
cpu, label,
((msr >> 15) & 1) ? "EN" : "DIS",
((msr >> 0) & 0x7FFF) * rapl_power_units,
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
(1.0 + (((msr >> 22) & 0x3) / 4.0)) * (1 << ((msr >> 17) & 0x1F)) * rapl_time_units,
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
(((msr >> 16) & 1) ? "EN" : "DIS"));
return;
}
int print_rapl(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
unsigned long long msr;
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
const char *msr_name;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
int cpu;
tools/power turbostat: No build warnings with -Wextra Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 23:45:18 -04:00
UNUSED(c);
UNUSED(p);
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (!platform->rapl_msrs)
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
return 0;
/* RAPL counters are per package, so print only for 1st thread/package */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (!is_cpu_first_thread_in_package(t, c, p))
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
return 0;
cpu = t->cpu_id;
if (cpu_migrate(cpu)) {
tools/power turbostat: harden against cpu hotplug turbostat tends to get confused when CPUs are added and removed while it is running. There are races, such as checking the current cpu, and then reading a sysfs file that depends on that cpu number. Close the two issues that seem to come up the most. First, there is an infinite reset loop detector -- change that to allow more resets before giving up. Secondly, one of those file reads didn't really need to exit the program on failure... Signed-off-by: Len Brown <len.brown@intel.com>
2020-09-30 20:58:15 -04:00
fprintf(outf, "print_rapl: Could not migrate to CPU %d\n", cpu);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
return -1;
}
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (platform->rapl_msrs & RAPL_AMD_F17H) {
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
msr_name = "MSR_RAPL_PWR_UNIT";
if (get_msr(cpu, MSR_RAPL_PWR_UNIT, &msr))
return -1;
} else {
msr_name = "MSR_RAPL_POWER_UNIT";
if (get_msr(cpu, MSR_RAPL_POWER_UNIT, &msr))
return -1;
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-17 12:34:41 -04:00
fprintf(outf, "cpu%d: %s: 0x%08llx (%f Watts, %f Joules, %f sec.)\n", cpu, msr_name, msr,
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
rapl_power_units, rapl_energy_units, rapl_time_units);
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (platform->rapl_msrs & RAPL_PKG_POWER_INFO) {
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
if (get_msr(cpu, MSR_PKG_POWER_INFO, &msr))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
return -5;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_PKG_POWER_INFO: 0x%08llx (%.0f W TDP, RAPL %.0f - %.0f W, %f sec.)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
cpu, msr,
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
((msr >> 0) & RAPL_POWER_GRANULARITY) * rapl_power_units,
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
((msr >> 16) & RAPL_POWER_GRANULARITY) * rapl_power_units,
((msr >> 32) & RAPL_POWER_GRANULARITY) * rapl_power_units,
((msr >> 48) & RAPL_TIME_GRANULARITY) * rapl_time_units);
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
}
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (platform->rapl_msrs & RAPL_PKG) {
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
if (get_msr(cpu, MSR_PKG_POWER_LIMIT, &msr))
return -9;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_PKG_POWER_LIMIT: 0x%08llx (%slocked)\n",
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
cpu, msr, (msr >> 63) & 1 ? "" : "UN");
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
print_power_limit_msr(cpu, msr, "PKG Limit #1");
tools/power turbostat: print power values upto three decimal Print power values upto three decimal places in watts. Suggested-by: Tony Luck <tony.luck@intel.com> Signed-off-by: Sumeet Pawnikar <sumeet.r.pawnikar@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-09-30 18:40:18 +05:30
fprintf(outf, "cpu%d: PKG Limit #2: %sabled (%0.3f Watts, %f* sec, clamp %sabled)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
cpu,
((msr >> 47) & 1) ? "EN" : "DIS",
((msr >> 32) & 0x7FFF) * rapl_power_units,
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
(1.0 + (((msr >> 54) & 0x3) / 4.0)) * (1 << ((msr >> 49) & 0x1F)) * rapl_time_units,
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
((msr >> 48) & 1) ? "EN" : "DIS");
tools/power turbostat: Add Power Limit4 support Add Power Limit4 support. Signed-off-by: Sumeet Pawnikar <sumeet.r.pawnikar@intel.com> Acked-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-08-20 17:42:43 +05:30
if (get_msr(cpu, MSR_VR_CURRENT_CONFIG, &msr))
return -9;
fprintf(outf, "cpu%d: MSR_VR_CURRENT_CONFIG: 0x%08llx\n", cpu, msr);
fprintf(outf, "cpu%d: PKG Limit #4: %f Watts (%slocked)\n",
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
cpu, ((msr >> 0) & 0x1FFF) * rapl_power_units, (msr >> 31) & 1 ? "" : "UN");
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
}
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (platform->rapl_msrs & RAPL_DRAM_POWER_INFO) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
if (get_msr(cpu, MSR_DRAM_POWER_INFO, &msr))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
return -6;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_DRAM_POWER_INFO,: 0x%08llx (%.0f W TDP, RAPL %.0f - %.0f W, %f sec.)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
cpu, msr,
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
((msr >> 0) & RAPL_POWER_GRANULARITY) * rapl_power_units,
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
((msr >> 16) & RAPL_POWER_GRANULARITY) * rapl_power_units,
((msr >> 32) & RAPL_POWER_GRANULARITY) * rapl_power_units,
((msr >> 48) & RAPL_TIME_GRANULARITY) * rapl_time_units);
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 00:50:30 -04:00
}
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (platform->rapl_msrs & RAPL_DRAM) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
if (get_msr(cpu, MSR_DRAM_POWER_LIMIT, &msr))
return -9;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d: MSR_DRAM_POWER_LIMIT: 0x%08llx (%slocked)\n",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
cpu, msr, (msr >> 31) & 1 ? "" : "UN");
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
print_power_limit_msr(cpu, msr, "DRAM Limit");
}
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (platform->rapl_msrs & RAPL_CORE_POLICY) {
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
if (get_msr(cpu, MSR_PP0_POLICY, &msr))
return -7;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
fprintf(outf, "cpu%d: MSR_PP0_POLICY: %lld\n", cpu, msr & 0xF);
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 00:30:16 -05:00
}
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (platform->rapl_msrs & RAPL_CORE_POWER_LIMIT) {
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
if (get_msr(cpu, MSR_PP0_POWER_LIMIT, &msr))
return -9;
fprintf(outf, "cpu%d: MSR_PP0_POWER_LIMIT: 0x%08llx (%slocked)\n",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
cpu, msr, (msr >> 31) & 1 ? "" : "UN");
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
print_power_limit_msr(cpu, msr, "Cores Limit");
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
}
tools/power/turbostat: Abstract RAPL MSRs support Abstract the support for RAPL MSRs. Delete CPU model checks in rapl_probe_intel(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-28 15:09:40 +08:00
if (platform->rapl_msrs & RAPL_GFX) {
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
if (get_msr(cpu, MSR_PP1_POLICY, &msr))
return -8;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
fprintf(outf, "cpu%d: MSR_PP1_POLICY: %lld\n", cpu, msr & 0xF);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
if (get_msr(cpu, MSR_PP1_POWER_LIMIT, &msr))
return -9;
fprintf(outf, "cpu%d: MSR_PP1_POWER_LIMIT: 0x%08llx (%slocked)\n",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
cpu, msr, (msr >> 31) & 1 ? "" : "UN");
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
print_power_limit_msr(cpu, msr, "GFX Limit");
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
}
return 0;
}
tools/power/turbostat: Reorder some functions Reorder some functions to solve code depdency introduced by next patch. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-30 16:36:06 +08:00
/*
* probe_rapl()
*
* sets rapl_power_units, rapl_energy_units, rapl_time_units
*/
void probe_rapl(void)
tools/power turbostat: Fix Core C6 residency on Atom CPUs For Atom CPUs that have core cstate deeper than C6, MSR_CORE_C6_RESIDENCY actually returns the residency for both CC6 and deeper Core cstates. Thus, the real Core C6 residency should be the subtraction of MSR_CORE_C6_RESIDENCY return value and MSR_CORE_C6_RESIDENCY return value. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-21 22:22:47 +08:00
{
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (!platform->rapl_msrs || no_msr)
tools/power/turbostat: Reorder some functions Reorder some functions to solve code depdency introduced by next patch. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-30 16:36:06 +08:00
return;
tools/power turbostat: Fix Core C6 residency on Atom CPUs For Atom CPUs that have core cstate deeper than C6, MSR_CORE_C6_RESIDENCY actually returns the residency for both CC6 and deeper Core cstates. Thus, the real Core C6 residency should be the subtraction of MSR_CORE_C6_RESIDENCY return value and MSR_CORE_C6_RESIDENCY return value. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-21 22:22:47 +08:00
tools/power/turbostat: Reorder some functions Reorder some functions to solve code depdency introduced by next patch. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-30 16:36:06 +08:00
if (genuine_intel)
rapl_probe_intel();
if (authentic_amd || hygon_genuine)
rapl_probe_amd();
tools/power/turbostat: Relocate more probing related code Relocate more feature probing code outside of process_cpuids() into the corresponding probing functions. This improves the readability of code and the turbostat output. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-30 16:54:51 +08:00
if (quiet)
return;
for_all_cpus(print_rapl, ODD_COUNTERS);
tools/power turbostat: Fix Core C6 residency on Atom CPUs For Atom CPUs that have core cstate deeper than C6, MSR_CORE_C6_RESIDENCY actually returns the residency for both CC6 and deeper Core cstates. Thus, the real Core C6 residency should be the subtraction of MSR_CORE_C6_RESIDENCY return value and MSR_CORE_C6_RESIDENCY return value. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-21 22:22:47 +08:00
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
/*
* MSR_IA32_TEMPERATURE_TARGET indicates the temperature where
* the Thermal Control Circuit (TCC) activates.
* This is usually equal to tjMax.
*
* Older processors do not have this MSR, so there we guess,
* but also allow cmdline over-ride with -T.
*
* Several MSR temperature values are in units of degrees-C
* below this value, including the Digital Thermal Sensor (DTS),
* Package Thermal Management Sensor (PTM), and thermal event thresholds.
*/
Revert "tools/power turbostat: adjust for temperature offset" This reverts commit 6ff7cb371c4bea3dba03a56d774da925e78a5087. Apparently the TCC offset should not be used to adjust what temperature we show the user after all. (on most systems, TCC offset is 0, FWIW) Fixes: 6ff7cb371c4b Signed-off-by: Len Brown <len.brown@intel.com>
2021-03-11 18:36:35 -05:00
int set_temperature_target(struct thread_data *t, struct core_data *c, struct pkg_data *p)
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
{
unsigned long long msr;
tools/power turbostat: rename tcc variables There are two TCC activation temeprature. One is the default TCC activation temperature, also known as TJ_MAX. Another one is the effective TCC activation temperature, which is the subtraction of default TCC activation temperature and TCC offset. The name of variable tcc_activation_temp might be misleading here. Thus rename tcc_activation_temp to tj_max, and use tcc_default and tcc_offset to calculate the effective TCC activation temperature. No functional change in this patch. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-26 18:49:26 +08:00
unsigned int tcc_default, tcc_offset;
Revert "tools/power turbostat: adjust for temperature offset" This reverts commit 6ff7cb371c4bea3dba03a56d774da925e78a5087. Apparently the TCC offset should not be used to adjust what temperature we show the user after all. (on most systems, TCC offset is 0, FWIW) Fixes: 6ff7cb371c4b Signed-off-by: Len Brown <len.brown@intel.com>
2021-03-11 18:36:35 -05:00
int cpu;
tools/power turbostat: adjust for temperature offset cpu1: MSR_IA32_TEMPERATURE_TARGET: 0x05640000 (95 C) (100 default - 5 offset) Account for the new "offset" field in MSR_TEMPERATURE_TARGET. While this field is usually zero, ignoring it results in over-stating the current temperature, both per-core and per-package. Signed-off-by: Len Brown <len.brown@intel.com>
2020-09-29 17:28:42 -04:00
tools/power turbostat: No build warnings with -Wextra Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 23:45:18 -04:00
UNUSED(c);
UNUSED(p);
tools/power turbostat: rename tcc variables There are two TCC activation temeprature. One is the default TCC activation temperature, also known as TJ_MAX. Another one is the effective TCC activation temperature, which is the subtraction of default TCC activation temperature and TCC offset. The name of variable tcc_activation_temp might be misleading here. Thus rename tcc_activation_temp to tj_max, and use tcc_default and tcc_offset to calculate the effective TCC activation temperature. No functional change in this patch. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-26 18:49:26 +08:00
/* tj_max is used only for dts or ptm */
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
if (!(do_dts || do_ptm))
return 0;
/* this is a per-package concept */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (!is_cpu_first_thread_in_package(t, c, p))
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
return 0;
Revert "tools/power turbostat: adjust for temperature offset" This reverts commit 6ff7cb371c4bea3dba03a56d774da925e78a5087. Apparently the TCC offset should not be used to adjust what temperature we show the user after all. (on most systems, TCC offset is 0, FWIW) Fixes: 6ff7cb371c4b Signed-off-by: Len Brown <len.brown@intel.com>
2021-03-11 18:36:35 -05:00
cpu = t->cpu_id;
if (cpu_migrate(cpu)) {
fprintf(outf, "Could not migrate to CPU %d\n", cpu);
return -1;
}
tools/power turbostat: rename tcc variables There are two TCC activation temeprature. One is the default TCC activation temperature, also known as TJ_MAX. Another one is the effective TCC activation temperature, which is the subtraction of default TCC activation temperature and TCC offset. The name of variable tcc_activation_temp might be misleading here. Thus rename tcc_activation_temp to tj_max, and use tcc_default and tcc_offset to calculate the effective TCC activation temperature. No functional change in this patch. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-26 18:49:26 +08:00
if (tj_max_override != 0) {
tj_max = tj_max_override;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "cpu%d: Using cmdline TCC Target (%d C)\n", cpu, tj_max);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
return 0;
}
Revert "tools/power turbostat: adjust for temperature offset" This reverts commit 6ff7cb371c4bea3dba03a56d774da925e78a5087. Apparently the TCC offset should not be used to adjust what temperature we show the user after all. (on most systems, TCC offset is 0, FWIW) Fixes: 6ff7cb371c4b Signed-off-by: Len Brown <len.brown@intel.com>
2021-03-11 18:36:35 -05:00
/* Temperature Target MSR is Nehalem and newer only */
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (!platform->has_nhm_msrs || no_msr)
Revert "tools/power turbostat: adjust for temperature offset" This reverts commit 6ff7cb371c4bea3dba03a56d774da925e78a5087. Apparently the TCC offset should not be used to adjust what temperature we show the user after all. (on most systems, TCC offset is 0, FWIW) Fixes: 6ff7cb371c4b Signed-off-by: Len Brown <len.brown@intel.com>
2021-03-11 18:36:35 -05:00
goto guess;
if (get_msr(base_cpu, MSR_IA32_TEMPERATURE_TARGET, &msr))
goto guess;
tools/power turbostat: rename tcc variables There are two TCC activation temeprature. One is the default TCC activation temperature, also known as TJ_MAX. Another one is the effective TCC activation temperature, which is the subtraction of default TCC activation temperature and TCC offset. The name of variable tcc_activation_temp might be misleading here. Thus rename tcc_activation_temp to tj_max, and use tcc_default and tcc_offset to calculate the effective TCC activation temperature. No functional change in this patch. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-26 18:49:26 +08:00
tcc_default = (msr >> 16) & 0xFF;
Revert "tools/power turbostat: adjust for temperature offset" This reverts commit 6ff7cb371c4bea3dba03a56d774da925e78a5087. Apparently the TCC offset should not be used to adjust what temperature we show the user after all. (on most systems, TCC offset is 0, FWIW) Fixes: 6ff7cb371c4b Signed-off-by: Len Brown <len.brown@intel.com>
2021-03-11 18:36:35 -05:00
tools/power turbostat: add TCC Offset support The length of TCC Offset bits varies on different platforms. Decode TCC Offset bits only for the platforms that we have verified. For the others, only show default TCC activation temperature. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-21 23:22:14 +08:00
if (!quiet) {
tools/power/turbostat: Abstract TCC Offset bits support Abstract the support for different TCC Offset bits in MSR_IA32_TEMPERATURE_TARGET. Delete check_tcc_offset() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 23:04:44 +08:00
int bits = platform->tcc_offset_bits;
unsigned long long enabled = 0;
if (bits && !get_msr(base_cpu, MSR_PLATFORM_INFO, &enabled))
enabled = (enabled >> 30) & 1;
if (bits && enabled) {
tcc_offset = (msr >> 24) & GENMASK(bits - 1, 0);
tools/power turbostat: add TCC Offset support The length of TCC Offset bits varies on different platforms. Decode TCC Offset bits only for the platforms that we have verified. For the others, only show default TCC activation temperature. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-21 23:22:14 +08:00
fprintf(outf, "cpu%d: MSR_IA32_TEMPERATURE_TARGET: 0x%08llx (%d C) (%d default - %d offset)\n",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
cpu, msr, tcc_default - tcc_offset, tcc_default, tcc_offset);
tools/power/turbostat: Abstract TCC Offset bits support Abstract the support for different TCC Offset bits in MSR_IA32_TEMPERATURE_TARGET. Delete check_tcc_offset() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-25 23:04:44 +08:00
} else {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "cpu%d: MSR_IA32_TEMPERATURE_TARGET: 0x%08llx (%d C)\n", cpu, msr, tcc_default);
tools/power turbostat: add TCC Offset support The length of TCC Offset bits varies on different platforms. Decode TCC Offset bits only for the platforms that we have verified. For the others, only show default TCC activation temperature. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-21 23:22:14 +08:00
}
}
Revert "tools/power turbostat: adjust for temperature offset" This reverts commit 6ff7cb371c4bea3dba03a56d774da925e78a5087. Apparently the TCC offset should not be used to adjust what temperature we show the user after all. (on most systems, TCC offset is 0, FWIW) Fixes: 6ff7cb371c4b Signed-off-by: Len Brown <len.brown@intel.com>
2021-03-11 18:36:35 -05:00
tools/power turbostat: rename tcc variables There are two TCC activation temeprature. One is the default TCC activation temperature, also known as TJ_MAX. Another one is the effective TCC activation temperature, which is the subtraction of default TCC activation temperature and TCC offset. The name of variable tcc_activation_temp might be misleading here. Thus rename tcc_activation_temp to tj_max, and use tcc_default and tcc_offset to calculate the effective TCC activation temperature. No functional change in this patch. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-26 18:49:26 +08:00
if (!tcc_default)
Revert "tools/power turbostat: adjust for temperature offset" This reverts commit 6ff7cb371c4bea3dba03a56d774da925e78a5087. Apparently the TCC offset should not be used to adjust what temperature we show the user after all. (on most systems, TCC offset is 0, FWIW) Fixes: 6ff7cb371c4b Signed-off-by: Len Brown <len.brown@intel.com>
2021-03-11 18:36:35 -05:00
goto guess;
tools/power turbostat: rename tcc variables There are two TCC activation temeprature. One is the default TCC activation temperature, also known as TJ_MAX. Another one is the effective TCC activation temperature, which is the subtraction of default TCC activation temperature and TCC offset. The name of variable tcc_activation_temp might be misleading here. Thus rename tcc_activation_temp to tj_max, and use tcc_default and tcc_offset to calculate the effective TCC activation temperature. No functional change in this patch. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-26 18:49:26 +08:00
tj_max = tcc_default;
Revert "tools/power turbostat: adjust for temperature offset" This reverts commit 6ff7cb371c4bea3dba03a56d774da925e78a5087. Apparently the TCC offset should not be used to adjust what temperature we show the user after all. (on most systems, TCC offset is 0, FWIW) Fixes: 6ff7cb371c4b Signed-off-by: Len Brown <len.brown@intel.com>
2021-03-11 18:36:35 -05:00
return 0;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
Revert "tools/power turbostat: adjust for temperature offset" This reverts commit 6ff7cb371c4bea3dba03a56d774da925e78a5087. Apparently the TCC offset should not be used to adjust what temperature we show the user after all. (on most systems, TCC offset is 0, FWIW) Fixes: 6ff7cb371c4b Signed-off-by: Len Brown <len.brown@intel.com>
2021-03-11 18:36:35 -05:00
guess:
tools/power turbostat: rename tcc variables There are two TCC activation temeprature. One is the default TCC activation temperature, also known as TJ_MAX. Another one is the effective TCC activation temperature, which is the subtraction of default TCC activation temperature and TCC offset. The name of variable tcc_activation_temp might be misleading here. Thus rename tcc_activation_temp to tj_max, and use tcc_default and tcc_offset to calculate the effective TCC activation temperature. No functional change in this patch. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-26 18:49:26 +08:00
tj_max = TJMAX_DEFAULT;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "cpu%d: Guessing tjMax %d C, Please use -T to specify\n", cpu, tj_max);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
return 0;
}
tools/power turbostat: decode more CPUID fields for debugging, dump a few more fields: CPUID(1): SSE3 MONITOR EIST TM2 TSC MSR ACPI-TM TM cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MONITOR) Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-21 12:22:47 -05:00
tools/power/turbostat: Reorder some functions Reorder some functions to solve code depdency introduced by next patch. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-30 16:36:06 +08:00
int print_thermal(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
unsigned long long msr;
unsigned int dts, dts2;
int cpu;
UNUSED(c);
UNUSED(p);
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (no_msr)
return 0;
tools/power/turbostat: Reorder some functions Reorder some functions to solve code depdency introduced by next patch. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-30 16:36:06 +08:00
if (!(do_dts || do_ptm))
return 0;
cpu = t->cpu_id;
/* DTS is per-core, no need to print for each thread */
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (!is_cpu_first_thread_in_core(t, c, p))
tools/power/turbostat: Reorder some functions Reorder some functions to solve code depdency introduced by next patch. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-30 16:36:06 +08:00
return 0;
if (cpu_migrate(cpu)) {
fprintf(outf, "print_thermal: Could not migrate to CPU %d\n", cpu);
return -1;
}
tools/power/turbostat: Abstract several functions When detecting the primary thread/core in a core/package, current code doesn't handle the allowed CPUs. Abstract several functions for further fix of this issue. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 15:13:23 +08:00
if (do_ptm && is_cpu_first_core_in_package(t, c, p)) {
tools/power/turbostat: Reorder some functions Reorder some functions to solve code depdency introduced by next patch. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-30 16:36:06 +08:00
if (get_msr(cpu, MSR_IA32_PACKAGE_THERM_STATUS, &msr))
return 0;
dts = (msr >> 16) & 0x7F;
fprintf(outf, "cpu%d: MSR_IA32_PACKAGE_THERM_STATUS: 0x%08llx (%d C)\n", cpu, msr, tj_max - dts);
if (get_msr(cpu, MSR_IA32_PACKAGE_THERM_INTERRUPT, &msr))
return 0;
dts = (msr >> 16) & 0x7F;
dts2 = (msr >> 8) & 0x7F;
fprintf(outf, "cpu%d: MSR_IA32_PACKAGE_THERM_INTERRUPT: 0x%08llx (%d C, %d C)\n",
cpu, msr, tj_max - dts, tj_max - dts2);
}
if (do_dts && debug) {
unsigned int resolution;
if (get_msr(cpu, MSR_IA32_THERM_STATUS, &msr))
return 0;
dts = (msr >> 16) & 0x7F;
resolution = (msr >> 27) & 0xF;
fprintf(outf, "cpu%d: MSR_IA32_THERM_STATUS: 0x%08llx (%d C +/- %d)\n",
cpu, msr, tj_max - dts, resolution);
if (get_msr(cpu, MSR_IA32_THERM_INTERRUPT, &msr))
return 0;
dts = (msr >> 16) & 0x7F;
dts2 = (msr >> 8) & 0x7F;
fprintf(outf, "cpu%d: MSR_IA32_THERM_INTERRUPT: 0x%08llx (%d C, %d C)\n",
cpu, msr, tj_max - dts, tj_max - dts2);
}
return 0;
}
void probe_thermal(void)
{
if (!access("/sys/devices/system/cpu/cpu0/thermal_throttle/core_throttle_count", R_OK))
BIC_PRESENT(BIC_CORE_THROT_CNT);
else
BIC_NOT_PRESENT(BIC_CORE_THROT_CNT);
tools/power/turbostat: Relocate more probing related code Relocate more feature probing code outside of process_cpuids() into the corresponding probing functions. This improves the readability of code and the turbostat output. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-30 16:54:51 +08:00
for_all_cpus(set_temperature_target, ODD_COUNTERS);
if (quiet)
return;
for_all_cpus(print_thermal, ODD_COUNTERS);
tools/power/turbostat: Reorder some functions Reorder some functions to solve code depdency introduced by next patch. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-30 16:36:06 +08:00
}
int get_cpu_type(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
unsigned int eax, ebx, ecx, edx;
UNUSED(c);
UNUSED(p);
if (!genuine_intel)
return 0;
if (cpu_migrate(t->cpu_id)) {
fprintf(outf, "Could not migrate to CPU %d\n", t->cpu_id);
return -1;
}
if (max_level < 0x1a)
return 0;
__cpuid(0x1a, eax, ebx, ecx, edx);
eax = (eax >> 24) & 0xFF;
if (eax == 0x20)
t->is_atom = true;
return 0;
}
tools/power turbostat: indicate SMX and SGX support SGX presence is related to a SKL power workaround, so lets show when that is enabled. Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-11 13:26:03 -05:00
void decode_feature_control_msr(void)
{
unsigned long long msr;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (no_msr)
return;
tools/x86: Sync msr-index.h from kernel sources Sync msr-index.h to pull in recent renames of the IA32_FEATURE_CONTROL MSR definitions. Update KVM's VMX selftest and turbostat accordingly. Keep the full name in turbostat's output to avoid breaking someone's workflow, e.g. if a script is looking for the full name. While using the renamed defines is by no means necessary, do the sync now to avoid leaving a landmine that will get stepped on the next time msr-index.h needs to be refreshed for some other reason. No functional change intended. Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com> Signed-off-by: Borislav Petkov <bp@suse.de> Link: https://lkml.kernel.org/r/20191221044513.21680-4-sean.j.christopherson@intel.com
2019-12-20 20:44:57 -08:00
if (!get_msr(base_cpu, MSR_IA32_FEAT_CTL, &msr))
tools/power turbostat: indicate SMX and SGX support SGX presence is related to a SKL power workaround, so lets show when that is enabled. Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-11 13:26:03 -05:00
fprintf(outf, "cpu%d: MSR_IA32_FEATURE_CONTROL: 0x%08llx (%sLocked %s)\n",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
base_cpu, msr, msr & FEAT_CTL_LOCKED ? "" : "UN-", msr & (1 << 18) ? "SGX" : "");
tools/power turbostat: indicate SMX and SGX support SGX presence is related to a SKL power workaround, so lets show when that is enabled. Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-11 13:26:03 -05:00
}
tools/power turbostat: decode more CPUID fields for debugging, dump a few more fields: CPUID(1): SSE3 MONITOR EIST TM2 TSC MSR ACPI-TM TM cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MONITOR) Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-21 12:22:47 -05:00
void decode_misc_enable_msr(void)
{
unsigned long long msr;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (no_msr)
return;
tools/power turbostat: decode MSR_IA32_MISC_ENABLE only on Intel otherwise, turbostat bails on on AMD Opteron boxes: turbostat: cpu26: msr offset 0x1a0 read failed: Input/output error Reported-by: Kamil Kolakowski <kkolakow@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-06-23 20:45:54 -07:00
if (!genuine_intel)
return;
tools/power turbostat: decode more CPUID fields for debugging, dump a few more fields: CPUID(1): SSE3 MONITOR EIST TM2 TSC MSR ACPI-TM TM cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MONITOR) Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-21 12:22:47 -05:00
if (!get_msr(base_cpu, MSR_IA32_MISC_ENABLE, &msr))
tools/power turbostat: further decode MSR_IA32_MISC_ENABLE Decode MISC_ENABLE.NO_TURBO, also use the #defines in msr-index.h for decoding this register cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MWAIT TURBO) Although it is not architectural, decode also MSR_IA32_MISC_ENABLE.prefetch-disable (bit-9). documented to be present on: Core, P4, Intel-Xeon reserved on: Atom, Silvermont, Nehalem, SNB, PHI ec. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-11 23:17:24 -05:00
fprintf(outf, "cpu%d: MSR_IA32_MISC_ENABLE: 0x%08llx (%sTCC %sEIST %sMWAIT %sPREFETCH %sTURBO)\n",
tools/power turbostat: decode more CPUID fields for debugging, dump a few more fields: CPUID(1): SSE3 MONITOR EIST TM2 TSC MSR ACPI-TM TM cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MONITOR) Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-21 12:22:47 -05:00
base_cpu, msr,
tools/power turbostat: further decode MSR_IA32_MISC_ENABLE Decode MISC_ENABLE.NO_TURBO, also use the #defines in msr-index.h for decoding this register cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MWAIT TURBO) Although it is not architectural, decode also MSR_IA32_MISC_ENABLE.prefetch-disable (bit-9). documented to be present on: Core, P4, Intel-Xeon reserved on: Atom, Silvermont, Nehalem, SNB, PHI ec. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-11 23:17:24 -05:00
msr & MSR_IA32_MISC_ENABLE_TM1 ? "" : "No-",
msr & MSR_IA32_MISC_ENABLE_ENHANCED_SPEEDSTEP ? "" : "No-",
tools/power turbostat: fix MSR_IA32_MISC_ENABLE MWAIT printout MSR_IA32_MISC_ENABLE[18] is the MWAIT ENABLE bit, not DISABLE bit... so MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST No-MWAIT PREFETCH TURBO) should print as: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MWAIT PREFETCH TURBO) Signed-off-by: Len Brown <len.brown@intel.com>
2017-11-08 23:15:42 -05:00
msr & MSR_IA32_MISC_ENABLE_MWAIT ? "" : "No-",
tools/power turbostat: further decode MSR_IA32_MISC_ENABLE Decode MISC_ENABLE.NO_TURBO, also use the #defines in msr-index.h for decoding this register cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MWAIT TURBO) Although it is not architectural, decode also MSR_IA32_MISC_ENABLE.prefetch-disable (bit-9). documented to be present on: Core, P4, Intel-Xeon reserved on: Atom, Silvermont, Nehalem, SNB, PHI ec. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-11 23:17:24 -05:00
msr & MSR_IA32_MISC_ENABLE_PREFETCH_DISABLE ? "No-" : "",
msr & MSR_IA32_MISC_ENABLE_TURBO_DISABLE ? "No-" : "");
tools/power turbostat: decode more CPUID fields for debugging, dump a few more fields: CPUID(1): SSE3 MONITOR EIST TM2 TSC MSR ACPI-TM TM cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MONITOR) Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-21 12:22:47 -05:00
}
tools/power turbostat: decode MSR_MISC_FEATURE_CONTROL useful for observing if the BIOS disabled prefetch Not architectural, but docuemented as present on NHM, SNB and is present on others. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 01:26:16 -05:00
void decode_misc_feature_control(void)
{
unsigned long long msr;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (no_msr)
return;
tools/power/turbostat: Abstract MSR_MISC_FEATURE_CONTROL support Abstract MSR_MISC_FEATURE_CONTROL support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 15:26:32 +08:00
if (!platform->has_msr_misc_feature_control)
tools/power turbostat: decode MSR_MISC_FEATURE_CONTROL useful for observing if the BIOS disabled prefetch Not architectural, but docuemented as present on NHM, SNB and is present on others. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 01:26:16 -05:00
return;
if (!get_msr(base_cpu, MSR_MISC_FEATURE_CONTROL, &msr))
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf,
"cpu%d: MSR_MISC_FEATURE_CONTROL: 0x%08llx (%sL2-Prefetch %sL2-Prefetch-pair %sL1-Prefetch %sL1-IP-Prefetch)\n",
base_cpu, msr, msr & (0 << 0) ? "No-" : "", msr & (1 << 0) ? "No-" : "",
msr & (2 << 0) ? "No-" : "", msr & (3 << 0) ? "No-" : "");
tools/power turbostat: decode MSR_MISC_FEATURE_CONTROL useful for observing if the BIOS disabled prefetch Not architectural, but docuemented as present on NHM, SNB and is present on others. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 01:26:16 -05:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: Decode MSR_MISC_PWR_MGMT This MSR is helpful to show if P-state HW coordination is enabled or disabled. Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-03 01:35:36 -05:00
/*
* Decode MSR_MISC_PWR_MGMT
*
* Decode the bits according to the Nehalem documentation
* bit[0] seems to continue to have same meaning going forward
* bit[1] less so...
*/
void decode_misc_pwr_mgmt_msr(void)
{
unsigned long long msr;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (no_msr)
return;
tools/power/turbostat: Abstract MSR_MISC_PWR_MGMT support Abstract MSR_MISC_PWR_MGMT support. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-21 22:12:57 +08:00
if (!platform->has_msr_misc_pwr_mgmt)
tools/power turbostat: BYT does not have MSR_MISC_PWR_MGMT and so --debug fails with: turbostat: msr 1 offset 0x1aa read failed: Input/output error It seems that baytrail, and airmont do not have this MSR. It is included in subsequent Goldmont Atom. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-01 13:08:33 -05:00
return;
tools/power turbostat: Decode MSR_MISC_PWR_MGMT This MSR is helpful to show if P-state HW coordination is enabled or disabled. Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-03 01:35:36 -05:00
if (!get_msr(base_cpu, MSR_MISC_PWR_MGMT, &msr))
tools/power turbostat: Display HWP OOB status Display if the HWP is enabled in OOB (Out of band) mode. Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-11-11 14:29:48 -08:00
fprintf(outf, "cpu%d: MSR_MISC_PWR_MGMT: 0x%08llx (%sable-EIST_Coordination %sable-EPB %sable-OOB)\n",
tools/power turbostat: Decode MSR_MISC_PWR_MGMT This MSR is helpful to show if P-state HW coordination is enabled or disabled. Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-03 01:35:36 -05:00
base_cpu, msr,
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
msr & (1 << 0) ? "DIS" : "EN", msr & (1 << 1) ? "EN" : "DIS", msr & (1 << 8) ? "EN" : "DIS");
tools/power turbostat: Decode MSR_MISC_PWR_MGMT This MSR is helpful to show if P-state HW coordination is enabled or disabled. Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-03 01:35:36 -05:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: decode Baytrail CC6 and MC6 demotion configuration with --debug, see: cpu0: MSR_CC6_DEMOTION_POLICY_CONFIG: 0x00000000 (DISable-CC6-Demotion) cpu0: MSR_MC6_DEMOTION_POLICY_CONFIG: 0x00000000 (DISable-MC6-Demotion) Note that the hardware default is to enable demotion, and Linux started clearing these registers in 3.17. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-07 22:37:48 -05:00
/*
* Decode MSR_CC6_DEMOTION_POLICY_CONFIG, MSR_MC6_DEMOTION_POLICY_CONFIG
*
* This MSRs are present on Silvermont processors,
* Intel Atom processor E3000 series (Baytrail), and friends.
*/
void decode_c6_demotion_policy_msr(void)
{
unsigned long long msr;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (no_msr)
return;
tools/power/turbostat: Abstract MSR_CC6/MC6_DEMOTION_POLICY_CONFIG support Abstract the support for MSR_CC6/MC6_DEMOTION_POLICY_CONFIG. Delete has_slv_msrs() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 14:52:24 +08:00
if (!platform->has_msr_c6_demotion_policy_config)
return;
tools/power turbostat: decode Baytrail CC6 and MC6 demotion configuration with --debug, see: cpu0: MSR_CC6_DEMOTION_POLICY_CONFIG: 0x00000000 (DISable-CC6-Demotion) cpu0: MSR_MC6_DEMOTION_POLICY_CONFIG: 0x00000000 (DISable-MC6-Demotion) Note that the hardware default is to enable demotion, and Linux started clearing these registers in 3.17. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-07 22:37:48 -05:00
if (!get_msr(base_cpu, MSR_CC6_DEMOTION_POLICY_CONFIG, &msr))
fprintf(outf, "cpu%d: MSR_CC6_DEMOTION_POLICY_CONFIG: 0x%08llx (%sable-CC6-Demotion)\n",
base_cpu, msr, msr & (1 << 0) ? "EN" : "DIS");
if (!get_msr(base_cpu, MSR_MC6_DEMOTION_POLICY_CONFIG, &msr))
fprintf(outf, "cpu%d: MSR_MC6_DEMOTION_POLICY_CONFIG: 0x%08llx (%sable-MC6-Demotion)\n",
base_cpu, msr, msr & (1 << 0) ? "EN" : "DIS");
}
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
tools/power turbostat: Print /dev/cpu_dma_latency Users are puzzled when they use tuned performance and all their C-states vanish. Dump /dev/cpu_dma_latency and state whether the value is default, or constraining, to explain this situation. Signed-off-by: Len Brown <len.brown@intel.com>
2020-05-27 00:38:38 -04:00
void print_dev_latency(void)
{
char *path = "/dev/cpu_dma_latency";
int fd;
int value;
int retval;
fd = open(path, O_RDONLY);
if (fd < 0) {
tools/power turbostat: Fix warning upon failed /dev/cpu_dma_latency read Previously a failed read of /dev/cpu_dma_latency erroneously complained turbostat: capget(CAP_SYS_ADMIN) failed, try "# setcap cap_sys_admin=ep ./turbostat This went unnoticed because this file is typically visible to root, and turbostat was typically run as root. Going forward, when a non-root user can run turbostat... Complain about failed read access to this file only if --debug is used. Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-19 12:25:42 -06:00
if (debug)
warnx("Read %s failed", path);
tools/power turbostat: Print /dev/cpu_dma_latency Users are puzzled when they use tuned performance and all their C-states vanish. Dump /dev/cpu_dma_latency and state whether the value is default, or constraining, to explain this situation. Signed-off-by: Len Brown <len.brown@intel.com>
2020-05-27 00:38:38 -04:00
return;
}
retval = read(fd, (void *)&value, sizeof(int));
if (retval != sizeof(int)) {
tools/power turbostat: remove stray newlines from warn/warnx strings warn(3) terminates strings with newlines Signed-off-by: Len Brown <len.brown@intel.com>
2023-03-17 11:25:56 -04:00
warn("read failed %s", path);
tools/power turbostat: Print /dev/cpu_dma_latency Users are puzzled when they use tuned performance and all their C-states vanish. Dump /dev/cpu_dma_latency and state whether the value is default, or constraining, to explain this situation. Signed-off-by: Len Brown <len.brown@intel.com>
2020-05-27 00:38:38 -04:00
close(fd);
return;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "/dev/cpu_dma_latency: %d usec (%s)\n", value, value == 2000000000 ? "default" : "constrained");
tools/power turbostat: Print /dev/cpu_dma_latency Users are puzzled when they use tuned performance and all their C-states vanish. Dump /dev/cpu_dma_latency and state whether the value is default, or constraining, to explain this situation. Signed-off-by: Len Brown <len.brown@intel.com>
2020-05-27 00:38:38 -04:00
close(fd);
}
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
static int has_instr_count_access(void)
{
int fd;
int has_access;
if (no_perf)
return 0;
fd = open_perf_counter(base_cpu, PERF_TYPE_HARDWARE, PERF_COUNT_HW_INSTRUCTIONS, -1, 0);
has_access = fd != -1;
if (fd != -1)
close(fd);
if (!has_access)
warnx("Failed to access %s. Some of the counters may not be available\n"
"\tRun as root to enable them or use %s to disable the access explicitly",
"instructions retired perf counter", "--no-perf");
return has_access;
}
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
int add_rapl_perf_counter_(int cpu, struct rapl_counter_info_t *rci, const struct rapl_counter_arch_info *cai,
double *scale_, enum rapl_unit *unit_)
{
if (no_perf)
return -1;
tools/power turbostat: Include umask=%x in perf counter's config Some counters, like cpu/cache-misses/, expose and require umask=%x parameter alongside event=%x in the sysfs perf counter's event file. This change make sure we parse and use it when opening user added counters. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-24 13:17:30 +02:00
const double scale = read_perf_scale(cai->perf_subsys, cai->perf_name);
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
if (scale == 0.0)
return -1;
const enum rapl_unit unit = read_perf_rapl_unit(cai->perf_subsys, cai->perf_name);
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
if (unit == RAPL_UNIT_INVALID)
return -1;
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
const unsigned int rapl_type = read_perf_type(cai->perf_subsys);
tools/power turbostat: Include umask=%x in perf counter's config Some counters, like cpu/cache-misses/, expose and require umask=%x parameter alongside event=%x in the sysfs perf counter's event file. This change make sure we parse and use it when opening user added counters. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-24 13:17:30 +02:00
const unsigned int rapl_energy_pkg_config = read_perf_config(cai->perf_subsys, cai->perf_name);
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
const int fd_counter =
open_perf_counter(cpu, rapl_type, rapl_energy_pkg_config, rci->fd_perf, PERF_FORMAT_GROUP);
if (fd_counter == -1)
return -1;
/* If it's the first counter opened, make it a group descriptor */
if (rci->fd_perf == -1)
rci->fd_perf = fd_counter;
*scale_ = scale;
*unit_ = unit;
return fd_counter;
}
int add_rapl_perf_counter(int cpu, struct rapl_counter_info_t *rci, const struct rapl_counter_arch_info *cai,
double *scale, enum rapl_unit *unit)
{
int ret = add_rapl_perf_counter_(cpu, rci, cai, scale, unit);
tools/power turbostat: Move verbose counter messages to level 2 Printing information about the source and value during initialization and reading of the counter for each cpu, while useful when debugging, results in too verbose output. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 14:28:03 +02:00
if (debug >= 2)
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
fprintf(stderr, "%s: %d (cpu: %d)\n", __func__, ret, cpu);
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
return ret;
}
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
/*
tools/power turbosstat: fix comment remove duplicate "the" in comment Signed-off-by: Jiang Jian <jiangjian@cdjrlc.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-06-21 16:16:17 +08:00
* Linux-perf manages the HW instructions-retired counter
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
* by enabling when requested, and hiding rollover
*/
void linux_perf_init(void)
{
if (access("/proc/sys/kernel/perf_event_paranoid", F_OK))
return;
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
if (BIC_IS_ENABLED(BIC_IPC) && has_aperf) {
fd_instr_count_percpu = calloc(topo.max_cpu_num + 1, sizeof(int));
if (fd_instr_count_percpu == NULL)
err(-1, "calloc fd_instr_count_percpu");
}
}
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
void rapl_perf_init(void)
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
{
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs Save the highest core and package id when parsing topology to allocate enough memory when get_rapl_counters() is called with a core or a package id as a domain. Note that RAPL domains are per-package on Intel, but per-core on AMD. Thus, the RAPL code effectively runs in different modes on those two product lines. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-06 15:39:08 +02:00
const unsigned int num_domains = (platform->has_per_core_rapl ? topo.max_core_id : topo.max_package_id) + 1;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
bool *domain_visited = calloc(num_domains, sizeof(bool));
rapl_counter_info_perdomain = calloc(num_domains, sizeof(*rapl_counter_info_perdomain));
if (rapl_counter_info_perdomain == NULL)
err(-1, "calloc rapl_counter_info_percpu");
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs Save the highest core and package id when parsing topology to allocate enough memory when get_rapl_counters() is called with a core or a package id as a domain. Note that RAPL domains are per-package on Intel, but per-core on AMD. Thus, the RAPL code effectively runs in different modes on those two product lines. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-06 15:39:08 +02:00
rapl_counter_info_perdomain_size = num_domains;
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
/*
* Initialize rapl_counter_info_percpu
*/
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs Save the highest core and package id when parsing topology to allocate enough memory when get_rapl_counters() is called with a core or a package id as a domain. Note that RAPL domains are per-package on Intel, but per-core on AMD. Thus, the RAPL code effectively runs in different modes on those two product lines. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-06 15:39:08 +02:00
for (unsigned int domain_id = 0; domain_id < num_domains; ++domain_id) {
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
struct rapl_counter_info_t *rci = &rapl_counter_info_perdomain[domain_id];
tools/power turbostat: v2024.04.10 Much of turbostat can now run with perf, rather than using the MSR driver Some of turbostat can now run as a regular non-root user. Add some new output columns for some new GFX hardware. [This patch updates the version, but otherwise changes no function; it touches up some checkpatch issues from previous patches] Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-08 19:32:58 -04:00
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rci->fd_perf = -1;
for (size_t i = 0; i < NUM_RAPL_COUNTERS; ++i) {
rci->data[i] = 0;
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
rci->source[i] = COUNTER_SOURCE_NONE;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
}
}
/*
* Open/probe the counters
* If can't get it via perf, fallback to MSR
*/
for (size_t i = 0; i < ARRAY_SIZE(rapl_counter_arch_infos); ++i) {
const struct rapl_counter_arch_info *const cai = &rapl_counter_arch_infos[i];
bool has_counter = 0;
double scale;
enum rapl_unit unit;
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs Save the highest core and package id when parsing topology to allocate enough memory when get_rapl_counters() is called with a core or a package id as a domain. Note that RAPL domains are per-package on Intel, but per-core on AMD. Thus, the RAPL code effectively runs in different modes on those two product lines. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-06 15:39:08 +02:00
unsigned int next_domain;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
memset(domain_visited, 0, num_domains * sizeof(*domain_visited));
for (int cpu = 0; cpu < topo.max_cpu_num + 1; ++cpu) {
if (cpu_is_not_allowed(cpu))
continue;
/* Skip already seen and handled RAPL domains */
next_domain =
platform->has_per_core_rapl ? cpus[cpu].physical_core_id : cpus[cpu].physical_package_id;
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs Save the highest core and package id when parsing topology to allocate enough memory when get_rapl_counters() is called with a core or a package id as a domain. Note that RAPL domains are per-package on Intel, but per-core on AMD. Thus, the RAPL code effectively runs in different modes on those two product lines. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-06 15:39:08 +02:00
assert(next_domain < num_domains);
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
if (domain_visited[next_domain])
continue;
domain_visited[next_domain] = 1;
tools/power turbostat: Add RAPL psys as a built-in counter Introduce the counter as a part of global, platform counters structure. We open the counter for only one cpu, but otherwise treat it as an ordinary RAPL counter, allowing for grouped perf read. The counter is disabled by default, because it's interpretation may require additional, platform specific information, making it unsuitable for general use. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-02 15:05:15 +02:00
if ((cai->flags & RAPL_COUNTER_FLAG_PLATFORM_COUNTER) && (cpu != base_cpu))
continue;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
struct rapl_counter_info_t *rci = &rapl_counter_info_perdomain[next_domain];
/* Check if the counter is enabled and accessible */
if (BIC_IS_ENABLED(cai->bic) && (platform->rapl_msrs & cai->feature_mask)) {
/* Use perf API for this counter */
if (!no_perf && cai->perf_name
&& add_rapl_perf_counter(cpu, rci, cai, &scale, &unit) != -1) {
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
rci->source[cai->rci_index] = COUNTER_SOURCE_PERF;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rci->scale[cai->rci_index] = scale * cai->compat_scale;
rci->unit[cai->rci_index] = unit;
rci->flags[cai->rci_index] = cai->flags;
/* Use MSR for this counter */
} else if (!no_msr && cai->msr && probe_msr(cpu, cai->msr) == 0) {
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
rci->source[cai->rci_index] = COUNTER_SOURCE_MSR;
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rci->msr[cai->rci_index] = cai->msr;
rci->msr_mask[cai->rci_index] = cai->msr_mask;
rci->msr_shift[cai->rci_index] = cai->msr_shift;
rci->unit[cai->rci_index] = RAPL_UNIT_JOULES;
rci->scale[cai->rci_index] = *cai->platform_rapl_msr_scale * cai->compat_scale;
rci->flags[cai->rci_index] = cai->flags;
}
}
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
if (rci->source[cai->rci_index] != COUNTER_SOURCE_NONE)
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
has_counter = 1;
}
/* If any CPU has access to the counter, make it present */
if (has_counter)
BIC_PRESENT(cai->bic);
}
free(domain_visited);
}
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
/* Assumes msr_counter_info is populated */
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
static int has_amperf_access(void)
{
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
return msr_counter_arch_infos[MSR_ARCH_INFO_APERF_INDEX].present &&
msr_counter_arch_infos[MSR_ARCH_INFO_MPERF_INDEX].present;
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
}
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
int *get_cstate_perf_group_fd(struct cstate_counter_info_t *cci, const char *group_name)
{
if (strcmp(group_name, "cstate_core") == 0)
return &cci->fd_perf_core;
if (strcmp(group_name, "cstate_pkg") == 0)
return &cci->fd_perf_pkg;
return NULL;
}
int add_cstate_perf_counter_(int cpu, struct cstate_counter_info_t *cci, const struct cstate_counter_arch_info *cai)
tools/power/turbostat: Add skeleton support for cstate enumeration Add skeleton support for cstate enumeration. Note that the previous logic may override the cstate setting for multiple times for different reasons. The conversion to new cstate enumeration must be done step by step following the previous code order strictly. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:37 +08:00
{
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
if (no_perf)
return -1;
tools/power/turbostat: Add skeleton support for cstate enumeration Add skeleton support for cstate enumeration. Note that the previous logic may override the cstate setting for multiple times for different reasons. The conversion to new cstate enumeration must be done step by step following the previous code order strictly. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:37 +08:00
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
int *pfd_group = get_cstate_perf_group_fd(cci, cai->perf_subsys);
if (pfd_group == NULL)
return -1;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
const unsigned int type = read_perf_type(cai->perf_subsys);
tools/power turbostat: Include umask=%x in perf counter's config Some counters, like cpu/cache-misses/, expose and require umask=%x parameter alongside event=%x in the sysfs perf counter's event file. This change make sure we parse and use it when opening user added counters. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-24 13:17:30 +02:00
const unsigned int config = read_perf_config(cai->perf_subsys, cai->perf_name);
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
const int fd_counter = open_perf_counter(cpu, type, config, *pfd_group, PERF_FORMAT_GROUP);
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
if (fd_counter == -1)
return -1;
/* If it's the first counter opened, make it a group descriptor */
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
if (*pfd_group == -1)
*pfd_group = fd_counter;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
return fd_counter;
}
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
int add_cstate_perf_counter(int cpu, struct cstate_counter_info_t *cci, const struct cstate_counter_arch_info *cai)
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
{
int ret = add_cstate_perf_counter_(cpu, cci, cai);
tools/power turbostat: Move verbose counter messages to level 2 Printing information about the source and value during initialization and reading of the counter for each cpu, while useful when debugging, results in too verbose output. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 14:28:03 +02:00
if (debug >= 2)
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
fprintf(stderr, "%s: %d (cpu: %d)\n", __func__, ret, cpu);
return ret;
}
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
int add_msr_perf_counter_(int cpu, struct msr_counter_info_t *cci, const struct msr_counter_arch_info *cai)
{
if (no_perf)
return -1;
const unsigned int type = read_perf_type(cai->perf_subsys);
tools/power turbostat: Include umask=%x in perf counter's config Some counters, like cpu/cache-misses/, expose and require umask=%x parameter alongside event=%x in the sysfs perf counter's event file. This change make sure we parse and use it when opening user added counters. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-24 13:17:30 +02:00
const unsigned int config = read_perf_config(cai->perf_subsys, cai->perf_name);
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
const int fd_counter = open_perf_counter(cpu, type, config, cci->fd_perf, PERF_FORMAT_GROUP);
if (fd_counter == -1)
return -1;
/* If it's the first counter opened, make it a group descriptor */
if (cci->fd_perf == -1)
cci->fd_perf = fd_counter;
return fd_counter;
}
int add_msr_perf_counter(int cpu, struct msr_counter_info_t *cci, const struct msr_counter_arch_info *cai)
{
int ret = add_msr_perf_counter_(cpu, cci, cai);
if (debug)
fprintf(stderr, "%s: %s/%s: %d (cpu: %d)\n", __func__, cai->perf_subsys, cai->perf_name, ret, cpu);
return ret;
}
void msr_perf_init_(void)
{
const int mci_num = topo.max_cpu_num + 1;
msr_counter_info = calloc(mci_num, sizeof(*msr_counter_info));
if (!msr_counter_info)
err(1, "calloc msr_counter_info");
msr_counter_info_size = mci_num;
for (int cpu = 0; cpu < mci_num; ++cpu)
msr_counter_info[cpu].fd_perf = -1;
for (int cidx = 0; cidx < NUM_MSR_COUNTERS; ++cidx) {
struct msr_counter_arch_info *cai = &msr_counter_arch_infos[cidx];
cai->present = false;
for (int cpu = 0; cpu < mci_num; ++cpu) {
struct msr_counter_info_t *const cci = &msr_counter_info[cpu];
if (cpu_is_not_allowed(cpu))
continue;
if (cai->needed) {
/* Use perf API for this counter */
if (!no_perf && cai->perf_name && add_msr_perf_counter(cpu, cci, cai) != -1) {
cci->source[cai->rci_index] = COUNTER_SOURCE_PERF;
cai->present = true;
/* User MSR for this counter */
} else if (!no_msr && cai->msr && probe_msr(cpu, cai->msr) == 0) {
cci->source[cai->rci_index] = COUNTER_SOURCE_MSR;
cci->msr[cai->rci_index] = cai->msr;
cci->msr_mask[cai->rci_index] = cai->msr_mask;
cai->present = true;
}
}
}
}
}
/* Initialize data for reading perf counters from the MSR group. */
void msr_perf_init(void)
{
bool need_amperf = false, need_smi = false;
const bool need_soft_c1 = (!platform->has_msr_core_c1_res) && (platform->supported_cstates & CC1);
need_amperf = BIC_IS_ENABLED(BIC_Avg_MHz) || BIC_IS_ENABLED(BIC_Busy) || BIC_IS_ENABLED(BIC_Bzy_MHz)
|| BIC_IS_ENABLED(BIC_IPC) || need_soft_c1;
if (BIC_IS_ENABLED(BIC_SMI))
need_smi = true;
/* Enable needed counters */
msr_counter_arch_infos[MSR_ARCH_INFO_APERF_INDEX].needed = need_amperf;
msr_counter_arch_infos[MSR_ARCH_INFO_MPERF_INDEX].needed = need_amperf;
msr_counter_arch_infos[MSR_ARCH_INFO_SMI_INDEX].needed = need_smi;
msr_perf_init_();
const bool has_amperf = has_amperf_access();
const bool has_smi = msr_counter_arch_infos[MSR_ARCH_INFO_SMI_INDEX].present;
has_aperf_access = has_amperf;
if (has_amperf) {
BIC_PRESENT(BIC_Avg_MHz);
BIC_PRESENT(BIC_Busy);
BIC_PRESENT(BIC_Bzy_MHz);
BIC_PRESENT(BIC_SMI);
}
if (has_smi)
BIC_PRESENT(BIC_SMI);
}
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
void cstate_perf_init_(bool soft_c1)
{
bool has_counter;
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
bool *cores_visited = NULL, *pkg_visited = NULL;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
const int cores_visited_elems = topo.max_core_id + 1;
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
const int pkg_visited_elems = topo.max_package_id + 1;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
const int cci_num = topo.max_cpu_num + 1;
ccstate_counter_info = calloc(cci_num, sizeof(*ccstate_counter_info));
if (!ccstate_counter_info)
err(1, "calloc ccstate_counter_arch_info");
ccstate_counter_info_size = cci_num;
cores_visited = calloc(cores_visited_elems, sizeof(*cores_visited));
if (!cores_visited)
err(1, "calloc cores_visited");
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
pkg_visited = calloc(pkg_visited_elems, sizeof(*pkg_visited));
if (!pkg_visited)
err(1, "calloc pkg_visited");
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
/* Initialize cstate_counter_info_percpu */
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
for (int cpu = 0; cpu < cci_num; ++cpu) {
ccstate_counter_info[cpu].fd_perf_core = -1;
ccstate_counter_info[cpu].fd_perf_pkg = -1;
}
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
for (int cidx = 0; cidx < NUM_CSTATE_COUNTERS; ++cidx) {
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
has_counter = false;
memset(cores_visited, 0, cores_visited_elems * sizeof(*cores_visited));
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
memset(pkg_visited, 0, pkg_visited_elems * sizeof(*pkg_visited));
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
const struct cstate_counter_arch_info *cai = &ccstate_counter_arch_infos[cidx];
for (int cpu = 0; cpu < cci_num; ++cpu) {
tools/power/turbostat: Add skeleton support for cstate enumeration Add skeleton support for cstate enumeration. Note that the previous logic may override the cstate setting for multiple times for different reasons. The conversion to new cstate enumeration must be done step by step following the previous code order strictly. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:37 +08:00
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
struct cstate_counter_info_t *const cci = &ccstate_counter_info[cpu];
tools/power/turbostat: Add skeleton support for cstate enumeration Add skeleton support for cstate enumeration. Note that the previous logic may override the cstate setting for multiple times for different reasons. The conversion to new cstate enumeration must be done step by step following the previous code order strictly. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:37 +08:00
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
if (cpu_is_not_allowed(cpu))
continue;
const int core_id = cpus[cpu].physical_core_id;
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
const int pkg_id = cpus[cpu].physical_package_id;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
assert(core_id < cores_visited_elems);
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
assert(pkg_id < pkg_visited_elems);
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
const bool per_thread = cai->flags & CSTATE_COUNTER_FLAG_COLLECT_PER_THREAD;
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
const bool per_core = cai->flags & CSTATE_COUNTER_FLAG_COLLECT_PER_CORE;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
if (!per_thread && cores_visited[core_id])
continue;
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
if (!per_core && pkg_visited[pkg_id])
continue;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
const bool counter_needed = BIC_IS_ENABLED(cai->bic) ||
(soft_c1 && (cai->flags & CSTATE_COUNTER_FLAG_SOFT_C1_DEPENDENCY));
tools/power turbostat: Ignore pkg_cstate_limit when it is not available When running in no-msr mode, the pkg_cstate_limit is not populated, thus we use perf to determine if given pcstate counter is present on the platform. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-09 12:39:47 +02:00
const bool counter_supported = (platform->supported_cstates & cai->feature_mask);
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
if (counter_needed && counter_supported) {
/* Use perf API for this counter */
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
if (!no_perf && cai->perf_name && add_cstate_perf_counter(cpu, cci, cai) != -1) {
tools/power/turbostat: Add skeleton support for cstate enumeration Add skeleton support for cstate enumeration. Note that the previous logic may override the cstate setting for multiple times for different reasons. The conversion to new cstate enumeration must be done step by step following the previous code order strictly. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:37 +08:00
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
cci->source[cai->rci_index] = COUNTER_SOURCE_PERF;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
/* User MSR for this counter */
tools/power turbostat: Ignore pkg_cstate_limit when it is not available When running in no-msr mode, the pkg_cstate_limit is not populated, thus we use perf to determine if given pcstate counter is present on the platform. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-09 12:39:47 +02:00
} else if (!no_msr && cai->msr && pkg_cstate_limit >= cai->pkg_cstate_limit
&& probe_msr(cpu, cai->msr) == 0) {
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
cci->source[cai->rci_index] = COUNTER_SOURCE_MSR;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
cci->msr[cai->rci_index] = cai->msr;
}
}
tools/power turbostat: Replace enum rapl_source and cstate_source with counter_source Reuse the enum. It means the same thing in both cases. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 12:22:57 +02:00
if (cci->source[cai->rci_index] != COUNTER_SOURCE_NONE) {
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
has_counter = true;
cores_visited[core_id] = true;
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
pkg_visited[pkg_id] = true;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
}
}
/* If any CPU has access to the counter, make it present */
if (has_counter)
BIC_PRESENT(cai->bic);
}
free(cores_visited);
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
free(pkg_visited);
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
}
void cstate_perf_init(void)
{
/*
* If we don't have a C1 residency MSR, we calculate it "in software",
* but we need APERF, MPERF too.
*/
const bool soft_c1 = !platform->has_msr_core_c1_res && has_amperf_access()
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
&& platform->supported_cstates & CC1;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
if (soft_c1)
BIC_PRESENT(BIC_CPU_c1);
cstate_perf_init_(soft_c1);
}
void probe_cstates(void)
{
probe_cst_limit();
tools/power/turbostat: Add skeleton support for cstate enumeration Add skeleton support for cstate enumeration. Note that the previous logic may override the cstate setting for multiple times for different reasons. The conversion to new cstate enumeration must be done step by step following the previous code order strictly. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:37 +08:00
tools/power/turbostat: Relocate cstate probing code Move all cstate probing related code into probe_cstates(). Note that dump_platform_info() actually dumps both MSR_PLATFORM_INFO and MSR_IA32_POWER_CTL. MSR_PLATFORM_INFO is for pstate and MSR_IA32_POWER_CTL is for cstate. So split dump_platform_info() and dump MSR_IA32_POWER_CTL in probe_cstates(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-29 13:47:58 +08:00
if (platform->has_msr_module_c6_res_ms)
BIC_PRESENT(BIC_Mod_c6);
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (platform->has_ext_cst_msrs && !no_msr) {
tools/power/turbostat: Relocate cstate probing code Move all cstate probing related code into probe_cstates(). Note that dump_platform_info() actually dumps both MSR_PLATFORM_INFO and MSR_IA32_POWER_CTL. MSR_PLATFORM_INFO is for pstate and MSR_IA32_POWER_CTL is for cstate. So split dump_platform_info() and dump MSR_IA32_POWER_CTL in probe_cstates(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-29 13:47:58 +08:00
BIC_PRESENT(BIC_Totl_c0);
BIC_PRESENT(BIC_Any_c0);
BIC_PRESENT(BIC_GFX_c0);
BIC_PRESENT(BIC_CPUGFX);
}
if (quiet)
return;
dump_power_ctl();
dump_cst_cfg();
decode_c6_demotion_policy_msr();
print_dev_latency();
dump_sysfs_cstate_config();
tools/power/turbostat: Relocate more probing related code Relocate more feature probing code outside of process_cpuids() into the corresponding probing functions. This improves the readability of code and the turbostat output. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-30 16:54:51 +08:00
print_irtl();
tools/power/turbostat: Add skeleton support for cstate enumeration Add skeleton support for cstate enumeration. Note that the previous logic may override the cstate setting for multiple times for different reasons. The conversion to new cstate enumeration must be done step by step following the previous code order strictly. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:37 +08:00
}
tools/power/turbostat: Relocate lpi probing code Introduce probe_lpi(), and move all lpi probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 16:01:36 +08:00
void probe_lpi(void)
{
if (!access("/sys/devices/system/cpu/cpuidle/low_power_idle_cpu_residency_us", R_OK))
BIC_PRESENT(BIC_CPU_LPI);
else
BIC_NOT_PRESENT(BIC_CPU_LPI);
if (!access(sys_lpi_file_sysfs, R_OK)) {
sys_lpi_file = sys_lpi_file_sysfs;
BIC_PRESENT(BIC_SYS_LPI);
} else if (!access(sys_lpi_file_debugfs, R_OK)) {
sys_lpi_file = sys_lpi_file_debugfs;
BIC_PRESENT(BIC_SYS_LPI);
} else {
sys_lpi_file_sysfs = NULL;
BIC_NOT_PRESENT(BIC_SYS_LPI);
}
}
tools/power/turbostat: Relocate pstate probing code Introduce probe_pstates() and move all pstate probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 15:58:10 +08:00
void probe_pstates(void)
{
probe_bclk();
if (quiet)
return;
dump_platform_info();
dump_turbo_ratio_info();
dump_sysfs_pstate_config();
decode_misc_pwr_mgmt_msr();
tools/power/turbostat: Relocate more probing related code Relocate more feature probing code outside of process_cpuids() into the corresponding probing functions. This improves the readability of code and the turbostat output. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-30 16:54:51 +08:00
for_all_cpus(print_hwp, ODD_COUNTERS);
for_all_cpus(print_epb, ODD_COUNTERS);
for_all_cpus(print_perf_limit, ODD_COUNTERS);
tools/power/turbostat: Relocate pstate probing code Introduce probe_pstates() and move all pstate probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 15:58:10 +08:00
}
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
void process_cpuid()
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
{
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
unsigned int eax, ebx, ecx, edx;
unsigned int fms, family, model, stepping, ecx_flags, edx_flags;
tools/power turbostat: print microcode patch level (also available via "grep microcode /proc/cpuinfo") Signed-off-by: Len Brown <len.brown@intel.com>
2021-02-04 14:44:12 -05:00
unsigned long long ucode_patch = 0;
tools/power turbostat: Print ucode revision only if valid If the MSR read were to fail, turbostat would print "microcode 0x0" Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-12 13:48:14 +01:00
bool ucode_patch_valid = false;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
eax = ebx = ecx = edx = 0;
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 03:14:35 -04:00
__cpuid(0, max_level, ebx, ecx, edx);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
if (ebx == 0x756e6547 && ecx == 0x6c65746e && edx == 0x49656e69)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
genuine_intel = 1;
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
else if (ebx == 0x68747541 && ecx == 0x444d4163 && edx == 0x69746e65)
authentic_amd = 1;
tools/power turbostat: Add support for Hygon Fam 18h (Dhyana) RAPL Commit 9392bd98bba760be96ee ("tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL") and the commit 3316f99a9f1b68c578c5 ("tools/power turbostat: Also read package power on AMD F17h (Zen)") add AMD Fam 17h RAPL support. Hygon Family 18h(Dhyana) support RAPL in bit 14 of CPUID 0x80000007 EDX, and has MSRs RAPL_PWR_UNIT/CORE_ENERGY_STAT/PKG_ENERGY_STAT. So add Hygon Dhyana Family 18h support for RAPL. Already tested on Hygon multi-node systems and it shows correct per-core energy usage and the total package power. Signed-off-by: Pu Wen <puwen@hygon.cn> Reviewed-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2019-08-31 10:20:31 +08:00
else if (ebx == 0x6f677948 && ecx == 0x656e6975 && edx == 0x6e65476e)
hygon_genuine = 1;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
if (!quiet)
tools/power turbostat: print microcode patch level (also available via "grep microcode /proc/cpuinfo") Signed-off-by: Len Brown <len.brown@intel.com>
2021-02-04 14:44:12 -05:00
fprintf(outf, "CPUID(0): %.4s%.4s%.4s 0x%x CPUID levels\n",
(char *)&ebx, (char *)&edx, (char *)&ecx, max_level);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 03:14:35 -04:00
__cpuid(1, fms, ebx, ecx, edx);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
family = (fms >> 8) & 0xf;
model = (fms >> 4) & 0xf;
stepping = fms & 0xf;
tools/power turbostat: Read extended processor family from CPUID This fixes the reported family on modern AMD processors (e.g. Ryzen, which is family 0x17). Previously these processors all showed up as family 0xf. See the document https://support.amd.com/TechDocs/56255_OSRR.pdf section CPUID_Fn00000001_EAX for how to calculate the family from the BaseFamily and ExtFamily values. This matches the code in arch/x86/lib/cpu.c Signed-off-by: Calvin Walton <calvin.walton@kepstin.ca> Signed-off-by: Len Brown <len.brown@intel.com>
2018-07-27 07:50:53 -04:00
if (family == 0xf)
family += (fms >> 20) & 0xff;
if (family >= 6)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
model += ((fms >> 16) & 0xf) << 4;
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
ecx_flags = ecx;
edx_flags = edx;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (!no_msr) {
if (get_msr(sched_getcpu(), MSR_IA32_UCODE_REV, &ucode_patch))
warnx("get_msr(UCODE)");
else
ucode_patch_valid = true;
}
tools/power turbostat: print microcode patch level (also available via "grep microcode /proc/cpuinfo") Signed-off-by: Len Brown <len.brown@intel.com>
2021-02-04 14:44:12 -05:00
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
/*
* check max extended function levels of CPUID.
* This is needed to check for invariant TSC.
* This check is valid for both Intel and AMD.
*/
ebx = ecx = edx = 0;
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 03:14:35 -04:00
__cpuid(0x80000000, max_extended_level, ebx, ecx, edx);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
if (!quiet) {
tools/power turbostat: Print ucode revision only if valid If the MSR read were to fail, turbostat would print "microcode 0x0" Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-12 13:48:14 +01:00
fprintf(outf, "CPUID(1): family:model:stepping 0x%x:%x:%x (%d:%d:%d)",
family, model, stepping, family, model, stepping);
if (ucode_patch_valid)
fprintf(outf, " microcode 0x%x", (unsigned int)((ucode_patch >> 32) & 0xFFFFFFFF));
fputc('\n', outf);
tools/power turbostat: print microcode patch level (also available via "grep microcode /proc/cpuinfo") Signed-off-by: Len Brown <len.brown@intel.com>
2021-02-04 14:44:12 -05:00
fprintf(outf, "CPUID(0x80000000): max_extended_levels: 0x%x\n", max_extended_level);
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
fprintf(outf, "CPUID(1): %s %s %s %s %s %s %s %s %s %s\n",
ecx_flags & (1 << 0) ? "SSE3" : "-",
ecx_flags & (1 << 3) ? "MONITOR" : "-",
ecx_flags & (1 << 6) ? "SMX" : "-",
ecx_flags & (1 << 7) ? "EIST" : "-",
ecx_flags & (1 << 8) ? "TM2" : "-",
edx_flags & (1 << 4) ? "TSC" : "-",
edx_flags & (1 << 5) ? "MSR" : "-",
edx_flags & (1 << 22) ? "ACPI-TM" : "-",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
edx_flags & (1 << 28) ? "HT" : "-", edx_flags & (1 << 29) ? "TM" : "-");
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
}
tools/power/turbostat: Enable TCC Offset on more models All Models that duplicate INTEL_FAM6_CANNONLAKE_L support TCC Offset. Enable this feature on all these models. Delete obsolete model_orig. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-03-24 16:30:55 +08:00
tools/power/turbostat: Add skeleton support for table driven feature enumeration Turbostat supports a series of features that may diverge among different CPU models. Current code uses various of CPU model checks in different places to handle this, which makes the code hard to maintain. Add skeleton support for table driven feature enumeration to replace the current error-prone CPU model checks and global variables. Note: by comparing the CPU models with intel-family.h, it is found that turbostat support for below four Models are missing, including INTEL_FAM6_ICELAKE, INTEL_FAM6_ATOM_SILVERMONT_MID, INTEL_FAM6_ATOM_AIRMONT_MID and INTEL_FAM6_ATOM_AIRMONT_NP. Adding support for these models is a different work, thus it is not covered in this patch set. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 21:52:53 +08:00
probe_platform_features(family, model);
tools/power turbosat: fix AMD APIC-id output turbostat recently gained a feature adding APIC and X2APIC columns. While they are disabled by-default, they are enabled with --debug or when explicitly requested, eg. $ sudo turbostat --quiet --show Package,Node,Core,CPU,APIC,X2APIC date But these columns erroneously showed zeros on AMD hardware. This patch corrects the APIC and X2APIC [sic] columns on AMD. Signed-off-by: Len Brown <len.brown@intel.com>
2018-08-07 20:22:28 -04:00
if (!(edx_flags & (1 << 5)))
errx(1, "CPUID: no MSR");
tools/power turbostat: CPUID(0x16) leaf shows base, max, and bus frequency This CPUID leaf is available on Skylake: CPUID(0x16): base_mhz: 1500 max_mhz: 2200 bus_mhz: 100 Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-23 02:30:51 -05:00
if (max_extended_level >= 0x80000007) {
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 00:12:33 -05:00
/*
* Non-Stop TSC is advertised by CPUID.EAX=0x80000007: EDX.bit8
* this check is valid for both Intel and AMD
*/
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 03:14:35 -04:00
__cpuid(0x80000007, eax, ebx, ecx, edx);
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 00:12:33 -05:00
has_invariant_tsc = edx & (1 << 8);
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
/*
* APERF/MPERF is advertised by CPUID.EAX=0x6: ECX.bit0
* this check is valid for both Intel and AMD
*/
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 03:14:35 -04:00
__cpuid(0x6, eax, ebx, ecx, edx);
tools: turbostat: fix bitwise and operand bug could cause false positive on indicating presence of invarient TSC or APERF support. Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: Len Brown <len.brown@intel.com>
2011-01-21 15:11:19 +01:00
has_aperf = ecx & (1 << 0);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
do_dts = eax & (1 << 0);
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (do_dts)
BIC_PRESENT(BIC_CoreTmp);
tools/power turbostat: decode CPUID(6).TURBO show the CPUID feature for turbo to clarify the case when it may not be shown in MISC_ENABLE CPUID(6): APERF, TURBO, DTS, PTM, No-HWP, No-HWPnotify, No-HWPwindow, No-HWPepp, No-HWPpkg, EPB cpu4: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MWAIT TURBO) Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 00:50:08 -05:00
has_turbo = eax & (1 << 1);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
do_ptm = eax & (1 << 6);
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
if (do_ptm)
BIC_PRESENT(BIC_PkgTmp);
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
has_hwp = eax & (1 << 7);
has_hwp_notify = eax & (1 << 8);
has_hwp_activity_window = eax & (1 << 9);
has_hwp_epp = eax & (1 << 10);
has_hwp_pkg = eax & (1 << 11);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
has_epb = ecx & (1 << 3);
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
if (!quiet)
tools/power turbostat: decode CPUID(6).TURBO show the CPUID feature for turbo to clarify the case when it may not be shown in MISC_ENABLE CPUID(6): APERF, TURBO, DTS, PTM, No-HWP, No-HWPnotify, No-HWPwindow, No-HWPepp, No-HWPpkg, EPB cpu4: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MWAIT TURBO) Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 00:50:08 -05:00
fprintf(outf, "CPUID(6): %sAPERF, %sTURBO, %sDTS, %sPTM, %sHWP, "
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
"%sHWPnotify, %sHWPwindow, %sHWPepp, %sHWPpkg, %sEPB\n",
has_aperf ? "" : "No-",
tools/power turbostat: decode CPUID(6).TURBO show the CPUID feature for turbo to clarify the case when it may not be shown in MISC_ENABLE CPUID(6): APERF, TURBO, DTS, PTM, No-HWP, No-HWPnotify, No-HWPwindow, No-HWPepp, No-HWPpkg, EPB cpu4: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MWAIT TURBO) Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 00:50:08 -05:00
has_turbo ? "" : "No-",
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 01:36:39 -05:00
do_dts ? "" : "No-",
do_ptm ? "" : "No-",
has_hwp ? "" : "No-",
has_hwp_notify ? "" : "No-",
has_hwp_activity_window ? "" : "No-",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
has_hwp_epp ? "" : "No-", has_hwp_pkg ? "" : "No-", has_epb ? "" : "No-");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
if (!quiet)
tools/power turbostat: decode more CPUID fields for debugging, dump a few more fields: CPUID(1): SSE3 MONITOR EIST TM2 TSC MSR ACPI-TM TM cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MONITOR) Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-21 12:22:47 -05:00
decode_misc_enable_msr();
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
if (max_level >= 0x7 && !quiet) {
tools/power turbostat: indicate SMX and SGX support SGX presence is related to a SKL power workaround, so lets show when that is enabled. Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-11 13:26:03 -05:00
int has_sgx;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: indicate SMX and SGX support SGX presence is related to a SKL power workaround, so lets show when that is enabled. Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-11 13:26:03 -05:00
ecx = 0;
__cpuid_count(0x7, 0, eax, ebx, ecx, edx);
has_sgx = ebx & (1 << 2);
tools/power turbostat: dump CPUID.7.EDX.Hybrid CPUID leaf 7 EDX now tells us if the processor has hybrid CPUs Signed-off-by: Len Brown <len.brown@intel.com>
2022-05-31 17:08:25 -10:00
is_hybrid = edx & (1 << 15);
fprintf(outf, "CPUID(7): %sSGX %sHybrid\n", has_sgx ? "" : "No-", is_hybrid ? "" : "No-");
tools/power turbostat: indicate SMX and SGX support SGX presence is related to a SKL power workaround, so lets show when that is enabled. Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-11 13:26:03 -05:00
if (has_sgx)
decode_feature_control_msr();
}
tools/power turbostat: CPUID(0x16) leaf shows base, max, and bus frequency This CPUID leaf is available on Skylake: CPUID(0x16): base_mhz: 1500 max_mhz: 2200 bus_mhz: 100 Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-23 02:30:51 -05:00
if (max_level >= 0x15) {
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-01 21:02:57 -04:00
unsigned int eax_crystal;
unsigned int ebx_tsc;
/*
* CPUID 15H TSC/Crystal ratio, possibly Crystal Hz
*/
eax_crystal = ebx_tsc = crystal_hz = edx = 0;
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 03:14:35 -04:00
__cpuid(0x15, eax_crystal, ebx_tsc, crystal_hz, edx);
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-01 21:02:57 -04:00
if (ebx_tsc != 0) {
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
if (!quiet && (ebx != 0))
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "CPUID(0x15): eax_crystal: %d ebx_tsc: %d ecx_crystal_hz: %d\n",
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-01 21:02:57 -04:00
eax_crystal, ebx_tsc, crystal_hz);
if (crystal_hz == 0)
tools/power/turbostat: Abstract hardcoded Crystal Clock frequency Abstract the support for hardcoded Crystal Clock frequency, which is used when crystal clock is not available from CPUID.15. Delete CPU model checks in process_cpuid(). Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-04-22 11:39:49 +08:00
crystal_hz = platform->crystal_freq;
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-01 21:02:57 -04:00
if (crystal_hz) {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tsc_hz = (unsigned long long)crystal_hz *ebx_tsc / eax_crystal;
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
if (!quiet)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "TSC: %lld MHz (%d Hz * %d / %d / 1000000)\n",
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tsc_hz / 1000000, crystal_hz, ebx_tsc, eax_crystal);
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-01 21:02:57 -04:00
}
}
}
tools/power turbostat: CPUID(0x16) leaf shows base, max, and bus frequency This CPUID leaf is available on Skylake: CPUID(0x16): base_mhz: 1500 max_mhz: 2200 bus_mhz: 100 Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-23 02:30:51 -05:00
if (max_level >= 0x16) {
unsigned int base_mhz, max_mhz, bus_mhz, edx;
/*
* CPUID 16H Base MHz, Max MHz, Bus MHz
*/
base_mhz = max_mhz = bus_mhz = edx = 0;
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 03:14:35 -04:00
__cpuid(0x16, base_mhz, max_mhz, bus_mhz, edx);
tools/power turbostat: Read base_hz and bclk from CPUID.16H if available If MSRs cannot be read, values can be obtained from cpuid. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-22 15:22:35 +01:00
bclk = bus_mhz;
base_hz = base_mhz * 1000000;
has_base_hz = 1;
if (platform->enable_tsc_tweak)
tsc_tweak = base_hz / tsc_hz;
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
if (!quiet)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "CPUID(0x16): base_mhz: %d max_mhz: %d bus_mhz: %d\n",
tools/power turbostat: CPUID(0x16) leaf shows base, max, and bus frequency This CPUID leaf is available on Skylake: CPUID(0x16): base_mhz: 1500 max_mhz: 2200 bus_mhz: 100 Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-23 02:30:51 -05:00
base_mhz, max_mhz, bus_mhz);
}
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-01 21:02:57 -04:00
tools/power turbostat: KNL workaround for %Busy and Avg_MHz KNL increments APERF and MPERF every 1024 clocks. This is compliant with the architecture specification, which requires that only the ratio of APERF/MPERF need be valid. However, turbostat takes advantage of the fact that these two MSRs increment every un-halted clock at the actual and base frequency: AVG_MHz = APERF_delta/measurement_interval %Busy = MPERF_delta/TSC_delta This quirk is needed for these calculations to also work on KNL, which would otherwise show a value 1024x smaller than expected. Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-14 13:31:00 +02:00
if (has_aperf)
tools/power/turbostat: Abstract aperf/mperf multiplier support Abstract aperf/mperf multiplier support. Delete is_knl() CPU model check. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 15:49:10 +08:00
aperf_mperf_multiplier = platform->need_perf_multiplier ? 1024 : 1;
tools/power turbostat: KNL workaround for %Busy and Avg_MHz KNL increments APERF and MPERF every 1024 clocks. This is compliant with the architecture specification, which requires that only the ratio of APERF/MPERF need be valid. However, turbostat takes advantage of the fact that these two MSRs increment every un-halted clock at the actual and base frequency: AVG_MHz = APERF_delta/measurement_interval %Busy = MPERF_delta/TSC_delta This quirk is needed for these calculations to also work on KNL, which would otherwise show a value 1024x smaller than expected. Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-14 13:31:00 +02:00
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
BIC_PRESENT(BIC_IRQ);
BIC_PRESENT(BIC_TSC_MHz);
tools/power/turbostat: Introduce probe_pm_features() Feature probe has nothing to do with CPUID, thus it should not be in process_cpuids(). Introduce probe_pm_features() and move all feature probing functions into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-13 23:32:39 +08:00
}
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
static void counter_info_init(void)
{
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
for (int i = 0; i < NUM_CSTATE_COUNTERS; ++i) {
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
struct cstate_counter_arch_info *const cai = &ccstate_counter_arch_infos[i];
if (platform->has_msr_knl_core_c6_residency && cai->msr == MSR_CORE_C6_RESIDENCY)
cai->msr = MSR_KNL_CORE_C6_RESIDENCY;
if (!platform->has_msr_core_c1_res && cai->msr == MSR_CORE_C1_RES)
cai->msr = 0;
tools/power turbostat: Read Package-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-08 15:00:14 +02:00
if (platform->has_msr_atom_pkg_c6_residency && cai->msr == MSR_PKG_C6_RESIDENCY)
cai->msr = MSR_ATOM_PKG_C6_RESIDENCY;
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
}
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
for (int i = 0; i < NUM_MSR_COUNTERS; ++i) {
msr_counter_arch_infos[i].present = false;
msr_counter_arch_infos[i].needed = false;
}
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
}
tools/power/turbostat: Introduce probe_pm_features() Feature probe has nothing to do with CPUID, thus it should not be in process_cpuids(). Introduce probe_pm_features() and move all feature probing functions into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-13 23:32:39 +08:00
void probe_pm_features(void)
{
tools/power/turbostat: Relocate pstate probing code Introduce probe_pstates() and move all pstate probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 15:58:10 +08:00
probe_pstates();
tools/power/turbostat: Add skeleton support for cstate enumeration Add skeleton support for cstate enumeration. Note that the previous logic may override the cstate setting for multiple times for different reasons. The conversion to new cstate enumeration must be done step by step following the previous code order strictly. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-27 22:37:37 +08:00
probe_cstates();
tools/power/turbostat: Relocate lpi probing code Introduce probe_lpi(), and move all lpi probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 16:01:36 +08:00
probe_lpi();
tools/power/turbostat: Rename uncore probing function Rename intel_uncore_frequency_probe() to probe_intel_uncore_frequency() to be consistent with other probing function names. Probe uncore frequency right after probing cstates. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 16:00:15 +08:00
probe_intel_uncore_frequency();
tools/power/turbostat: Relocate graphics probing code Introduce probe_graphics(), and move all graphics probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 16:01:12 +08:00
probe_graphics();
tools/power/turbostat: Rename rapl probing function Rename rapl_probe() to probe_rapl() to be consistent with other probing function names. Probe rapl after probing uncore frequency. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 16:00:33 +08:00
probe_rapl();
tools/power/turbostat: Relocate thermal probing code Introduce probe_thermal(), and move all thermal probing related code into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-08-31 16:02:16 +08:00
probe_thermal();
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (platform->has_nhm_msrs && !no_msr)
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
BIC_PRESENT(BIC_SMI);
tools/power turbostat: Decode MSR_MISC_PWR_MGMT This MSR is helpful to show if P-state HW coordination is enabled or disabled. Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-03 01:35:36 -05:00
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
if (!quiet)
tools/power turbostat: decode MSR_MISC_FEATURE_CONTROL useful for observing if the BIOS disabled prefetch Not architectural, but docuemented as present on NHM, SNB and is present on others. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 01:26:16 -05:00
decode_misc_feature_control();
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
/*
* in /dev/cpu/ return success for names that are numbers
* ie. filter out ".", "..", "microcode".
*/
int dir_filter(const struct dirent *dirp)
{
if (isdigit(dirp->d_name[0]))
return 1;
else
return 0;
}
tools/power/turbostat: Handle offlined CPUs in cpu_subset It is possible that the cpu_subset contains offlined CPUs. If this happens during start, exit immediately because this is likely an operator error that is best fixed by re-invoking. If this happens at runtime, give a warning only because turbostat should do its best effort to continue running. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 14:20:47 +08:00
void topology_probe(bool startup)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
{
int i;
int max_core_id = 0;
int max_package_id = 0;
int max_siblings = 0;
/* Initialize num_cpus, max_cpu_num */
tools/power turbostat: set max_num_cpus equal to the cpumask length Future fixes will use sysfs files that contain cpumask output. The code needs to know the length of the cpumask in order to determine which cpus are set in a cpumask. Currently topo.max_cpu_num is the maximum cpu number. It can be increased the the maximum value of cpus represented in cpumasks. Set max_num_cpus to the length of a cpumask. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:28 -04:00
set_max_cpu_num();
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
topo.num_cpus = 0;
for_all_proc_cpus(count_cpus);
tools/power turbostat: Honor --show CPU, even when even when num_cpus=1 Honor --show CPU and --show Core when "topo.num_cpus == 1". Previously turbostat assumed that on a 1-CPU system, these columns should never appear. Honoring these flags makes it easier for several programs that parse turbostat output. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-09-17 22:33:26 +02:00
if (!summary_only)
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
BIC_PRESENT(BIC_CPU);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
if (debug > 1)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "num_cpus %d max_cpu_num %d\n", topo.num_cpus, topo.max_cpu_num);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
cpus = calloc(1, (topo.max_cpu_num + 1) * sizeof(struct cpu_topology));
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:18 -07:00
if (cpus == NULL)
err(1, "calloc cpus");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/*
* Allocate and initialize cpu_present_set
*/
cpu_present_set = CPU_ALLOC((topo.max_cpu_num + 1));
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:18 -07:00
if (cpu_present_set == NULL)
err(3, "CPU_ALLOC");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
cpu_present_setsize = CPU_ALLOC_SIZE((topo.max_cpu_num + 1));
CPU_ZERO_S(cpu_present_setsize, cpu_present_set);
for_all_proc_cpus(mark_cpu_present);
tools/power/turbostat: Handle cgroup v2 cpu limitation CPUs can be isolated via cgroup settings and turbostat should avoid migrating to these CPUs, just like it does for the '-c' cpus. Introduce cpu_effective_set to save the cgroup cpu limitation info from /sys/fs/cgroup/cpuset.cpus.effective. And use cpu_allowed_set as the intersection of cpu_present_set, cpu_effective_set and cpu_subset. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:39:22 +08:00
/*
* Allocate and initialize cpu_effective_set
*/
cpu_effective_set = CPU_ALLOC((topo.max_cpu_num + 1));
if (cpu_effective_set == NULL)
err(3, "CPU_ALLOC");
cpu_effective_setsize = CPU_ALLOC_SIZE((topo.max_cpu_num + 1));
CPU_ZERO_S(cpu_effective_setsize, cpu_effective_set);
update_effective_set(startup);
tools/power/turbostat: Introduce cpu_allowed_set Turbostat supports "-c" parameter which limits output to system summary plus the specified cpu-set. But some code still uses cpu_present_set to read and dump the counters. Introduce cpu_allowed_set for code that should obey the specified cpu-set. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 11:01:17 +08:00
/*
* Allocate and initialize cpu_allowed_set
*/
cpu_allowed_set = CPU_ALLOC((topo.max_cpu_num + 1));
if (cpu_allowed_set == NULL)
err(3, "CPU_ALLOC");
cpu_allowed_setsize = CPU_ALLOC_SIZE((topo.max_cpu_num + 1));
CPU_ZERO_S(cpu_allowed_setsize, cpu_allowed_set);
tools/power turbostat: add --cpu parameter With the --cpu parameter, turbostat prints only lines for the specified set of CPUs: sudo ./turbostat --quiet --show Core,CPU --cpu 0,1,3..5,6-7 Core CPU - - 0 0 0 4 1 1 1 5 2 6 3 3 3 7 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 23:54:15 -05:00
/*
tools/power/turbostat: Handle cgroup v2 cpu limitation CPUs can be isolated via cgroup settings and turbostat should avoid migrating to these CPUs, just like it does for the '-c' cpus. Introduce cpu_effective_set to save the cgroup cpu limitation info from /sys/fs/cgroup/cpuset.cpus.effective. And use cpu_allowed_set as the intersection of cpu_present_set, cpu_effective_set and cpu_subset. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:39:22 +08:00
* Validate and update cpu_allowed_set.
tools/power/turbostat: Handle offlined CPUs in cpu_subset It is possible that the cpu_subset contains offlined CPUs. If this happens during start, exit immediately because this is likely an operator error that is best fixed by re-invoking. If this happens at runtime, give a warning only because turbostat should do its best effort to continue running. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 14:20:47 +08:00
*
tools/power/turbostat: Handle cgroup v2 cpu limitation CPUs can be isolated via cgroup settings and turbostat should avoid migrating to these CPUs, just like it does for the '-c' cpus. Introduce cpu_effective_set to save the cgroup cpu limitation info from /sys/fs/cgroup/cpuset.cpus.effective. And use cpu_allowed_set as the intersection of cpu_present_set, cpu_effective_set and cpu_subset. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:39:22 +08:00
* Make sure all cpus in cpu_subset are also in cpu_present_set during startup.
* Give a warning when cpus in cpu_subset become unavailable at runtime.
* Give a warning when cpus are not effective because of cgroup setting.
tools/power/turbostat: Handle offlined CPUs in cpu_subset It is possible that the cpu_subset contains offlined CPUs. If this happens during start, exit immediately because this is likely an operator error that is best fixed by re-invoking. If this happens at runtime, give a warning only because turbostat should do its best effort to continue running. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 14:20:47 +08:00
*
tools/power/turbostat: Handle cgroup v2 cpu limitation CPUs can be isolated via cgroup settings and turbostat should avoid migrating to these CPUs, just like it does for the '-c' cpus. Introduce cpu_effective_set to save the cgroup cpu limitation info from /sys/fs/cgroup/cpuset.cpus.effective. And use cpu_allowed_set as the intersection of cpu_present_set, cpu_effective_set and cpu_subset. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:39:22 +08:00
* cpu_allowed_set is the intersection of cpu_present_set/cpu_effective_set/cpu_subset.
tools/power turbostat: add --cpu parameter With the --cpu parameter, turbostat prints only lines for the specified set of CPUs: sudo ./turbostat --quiet --show Core,CPU --cpu 0,1,3..5,6-7 Core CPU - - 0 0 0 4 1 1 1 5 2 6 3 3 3 7 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 23:54:15 -05:00
*/
for (i = 0; i < CPU_SUBSET_MAXCPUS; ++i) {
tools/power/turbostat: Handle cgroup v2 cpu limitation CPUs can be isolated via cgroup settings and turbostat should avoid migrating to these CPUs, just like it does for the '-c' cpus. Introduce cpu_effective_set to save the cgroup cpu limitation info from /sys/fs/cgroup/cpuset.cpus.effective. And use cpu_allowed_set as the intersection of cpu_present_set, cpu_effective_set and cpu_subset. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:39:22 +08:00
if (cpu_subset && !CPU_ISSET_S(i, cpu_subset_size, cpu_subset))
tools/power/turbostat: Introduce cpu_allowed_set Turbostat supports "-c" parameter which limits output to system summary plus the specified cpu-set. But some code still uses cpu_present_set to read and dump the counters. Introduce cpu_allowed_set for code that should obey the specified cpu-set. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 11:01:17 +08:00
continue;
tools/power/turbostat: Handle cgroup v2 cpu limitation CPUs can be isolated via cgroup settings and turbostat should avoid migrating to these CPUs, just like it does for the '-c' cpus. Introduce cpu_effective_set to save the cgroup cpu limitation info from /sys/fs/cgroup/cpuset.cpus.effective. And use cpu_allowed_set as the intersection of cpu_present_set, cpu_effective_set and cpu_subset. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:39:22 +08:00
if (!CPU_ISSET_S(i, cpu_present_setsize, cpu_present_set)) {
if (cpu_subset) {
/* cpus in cpu_subset must be in cpu_present_set during startup */
tools/power/turbostat: Handle offlined CPUs in cpu_subset It is possible that the cpu_subset contains offlined CPUs. If this happens during start, exit immediately because this is likely an operator error that is best fixed by re-invoking. If this happens at runtime, give a warning only because turbostat should do its best effort to continue running. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 14:20:47 +08:00
if (startup)
err(1, "cpu%d not present", i);
else
fprintf(stderr, "cpu%d not present\n", i);
}
tools/power/turbostat: Handle cgroup v2 cpu limitation CPUs can be isolated via cgroup settings and turbostat should avoid migrating to these CPUs, just like it does for the '-c' cpus. Introduce cpu_effective_set to save the cgroup cpu limitation info from /sys/fs/cgroup/cpuset.cpus.effective. And use cpu_allowed_set as the intersection of cpu_present_set, cpu_effective_set and cpu_subset. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:39:22 +08:00
continue;
tools/power/turbostat: Introduce cpu_allowed_set Turbostat supports "-c" parameter which limits output to system summary plus the specified cpu-set. But some code still uses cpu_present_set to read and dump the counters. Introduce cpu_allowed_set for code that should obey the specified cpu-set. No functional change. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 11:01:17 +08:00
}
tools/power/turbostat: Handle cgroup v2 cpu limitation CPUs can be isolated via cgroup settings and turbostat should avoid migrating to these CPUs, just like it does for the '-c' cpus. Introduce cpu_effective_set to save the cgroup cpu limitation info from /sys/fs/cgroup/cpuset.cpus.effective. And use cpu_allowed_set as the intersection of cpu_present_set, cpu_effective_set and cpu_subset. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:39:22 +08:00
if (CPU_COUNT_S(cpu_effective_setsize, cpu_effective_set)) {
if (!CPU_ISSET_S(i, cpu_effective_setsize, cpu_effective_set)) {
fprintf(stderr, "cpu%d not effective\n", i);
continue;
}
}
CPU_SET_S(i, cpu_allowed_setsize, cpu_allowed_set);
tools/power turbostat: add --cpu parameter With the --cpu parameter, turbostat prints only lines for the specified set of CPUs: sudo ./turbostat --quiet --show Core,CPU --cpu 0,1,3..5,6-7 Core CPU - - 0 0 0 4 1 1 1 5 2 6 3 3 3 7 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 23:54:15 -05:00
}
tools/power/turbostat: Obey allowed CPUs during startup Set turbostat CPU affinity to make sure turbostat is running on one of the allowed CPUs. Set base_cpu to the first allowed CPU so that some platform information is dumped using one of the allowed CPUs. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 14:25:25 +08:00
if (!CPU_COUNT_S(cpu_allowed_setsize, cpu_allowed_set))
err(-ENODEV, "No valid cpus found");
sched_setaffinity(0, cpu_allowed_setsize, cpu_allowed_set);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/*
* Allocate and initialize cpu_affinity_set
*/
cpu_affinity_set = CPU_ALLOC((topo.max_cpu_num + 1));
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:18 -07:00
if (cpu_affinity_set == NULL)
err(3, "CPU_ALLOC");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
cpu_affinity_setsize = CPU_ALLOC_SIZE((topo.max_cpu_num + 1));
CPU_ZERO_S(cpu_affinity_setsize, cpu_affinity_set);
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
for_all_proc_cpus(init_thread_id);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: Allow using cpu device in perf counters on hybrid platforms Intel hybrid platforms expose different perf devices for P and E cores. Instead of one, "/sys/bus/event_source/devices/cpu" device, there are "/sys/bus/event_source/devices/{cpu_core,cpu_atom}". This, however makes it more complicated for the user, because most of the counters are available on both and had to be handled manually. This patch allows users to use "virtual" cpu device that is seemingly translated to cpu_core and cpu_atom perf devices, depending on the type of a CPU we are opening the counter for. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-20 18:47:59 +02:00
for_all_proc_cpus(set_cpu_hybrid_type);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/*
* For online cpus
* find max_core_id, max_package_id
*/
for (i = 0; i <= topo.max_cpu_num; ++i) {
int siblings;
if (cpu_is_not_present(i)) {
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
if (debug > 1)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "cpu%d NOT PRESENT\n", i);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
continue;
}
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
cpus[i].logical_cpu_id = i;
/* get package information */
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
cpus[i].physical_package_id = get_physical_package_id(i);
if (cpus[i].physical_package_id > max_package_id)
max_package_id = cpus[i].physical_package_id;
tools/power turbostat: Add Die column If the system has more than one software visible die per package, print a Die column. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-14 19:17:40 -08:00
/* get die information */
cpus[i].die_id = get_die_id(i);
tools/power turbostat: Remember global max_die_id This is necessary to gracefully handle sparse die_id's. no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 11:56:48 -04:00
if (cpus[i].die_id > topo.max_die_id)
topo.max_die_id = cpus[i].die_id;
tools/power turbostat: Add Die column If the system has more than one software visible die per package, print a Die column. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-14 19:17:40 -08:00
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
/* get numa node information */
tools/power turbostat: Calculate additional node information for a package The code currently assumes each package has exactly one node. This is not the case for AMD systems and Intel systems with COD. AMD systems also may re-enumerate each node's core IDs starting at 0 (for example, an AMD processor may have two nodes, each with core IDs from 0 to 7). In order to properly enumerate the cores we need to track both the physical and logical node IDs. Add physical_node_id to track the node ID assigned by the kernel, and logical_node_id used by turbostat to track the nodes per package ie) a 0-based count within the package. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:30 -04:00
cpus[i].physical_node_id = get_physical_node_id(&cpus[i]);
if (cpus[i].physical_node_id > topo.max_node_num)
topo.max_node_num = cpus[i].physical_node_id;
tools/power turbostat: Fix node and siblings lookup data The turbostat code only looks at thread_siblings_list to determine if processing units/threads are on the same the core. This works well on Intel systems which have a shared L1 instruction and data cache. This does not work on AMD systems which have shared L1 instruction cache but separate L1 data caches. Other utilities also check sibling's core ID to determine if the processing unit shares the same core. Additionally, the cpu_topology *cpus list used in topology_probe() can be used elsewhere in the code to simplify things. Export *cpus to the entire turbostat code, and add Processing Unit/Thread IDs information to each cpu_topology struct. Confirm that the thread is on the same core as indicated by thread_siblings_list. [v2]: Fixup CPU_* usage that caused gcc malloc error. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 22:08:58 -04:00
/* get core information */
cpus[i].physical_core_id = get_core_id(i);
if (cpus[i].physical_core_id > max_core_id)
max_core_id = cpus[i].physical_core_id;
/* get thread information */
siblings = get_thread_siblings(&cpus[i]);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
if (siblings > max_siblings)
max_siblings = siblings;
tools/power turbostat: fix bogus summary values This patch fixes a regression introduced in commit 8cb48b32a5de ("tools/power turbostat: track thread ID in cpu_topology") Turbostat uses incorrect cores number ('topo.num_cores') - its value is count of logical CPUs, instead of count of physical cores. So it is twice as large as it should be on a typical Intel system. For example, on a 6 core Xeon system 'topo.num_cores' is 12, and on a 52 core Xeon system 'topo.num_cores' is 104. And interestingly, on a 68-core Knights Landing Intel system 'topo.num_cores' is 272, because this system has 4 logical CPUs per core. As a result, some of the turbostat calculations are incorrect. For example, on idle 52-core Xeon system when all cores are ~99% in Core C6 (CPU%c6), the summary (very first) line shows ~48% Core C6, while it should be ~99%. This patch fixes the problem by fixing 'topo.num_cores' calculation. Was: 1. Init 'thread_id' for all CPUs to -1 2. Run 'get_thread_siblings()' which sets it to 0 or 1 3. Increment 'topo.num_cores' when thread_id != -1 (bug!) Now: 1. Init 'thread_id' for all CPUs to -1 2. Run 'get_thread_siblings()' which sets it to 0 or 1 3. Increment 'topo.num_cores' when thread_id is not 0 I did not have a chance to test this on an AMD machine, and only tested on a couple of Intel Xeons (6 and 52 cores). Reported-by: Vladislav Govtva <vladislav.govtva@intel.com> Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-07-25 11:52:06 +03:00
if (cpus[i].thread_id == 0)
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
topo.num_cores++;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs Save the highest core and package id when parsing topology to allocate enough memory when get_rapl_counters() is called with a core or a package id as a domain. Note that RAPL domains are per-package on Intel, but per-core on AMD. Thus, the RAPL code effectively runs in different modes on those two product lines. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-06 15:39:08 +02:00
topo.max_core_id = max_core_id;
topo.max_package_id = max_package_id;
tools/power turbostat: Calculate additional node information for a package The code currently assumes each package has exactly one node. This is not the case for AMD systems and Intel systems with COD. AMD systems also may re-enumerate each node's core IDs starting at 0 (for example, an AMD processor may have two nodes, each with core IDs from 0 to 7). In order to properly enumerate the cores we need to track both the physical and logical node IDs. Add physical_node_id to track the node ID assigned by the kernel, and logical_node_id used by turbostat to track the nodes per package ie) a 0-based count within the package. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:30 -04:00
tools/power turbostat: remove num_ from cpu_topology struct Cleanup, remove num_ from num_nodes_per_pkg, num_cores_per_node, and num_threads_per_node. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:33 -04:00
topo.cores_per_node = max_core_id + 1;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
if (debug > 1)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "max_core_id %d, sizing for %d cores per package\n", max_core_id, topo.cores_per_node);
tools/power turbostat: Honor --show CPU, even when even when num_cpus=1 Honor --show CPU and --show Core when "topo.num_cpus == 1". Previously turbostat assumed that on a 1-CPU system, these columns should never appear. Honoring these flags makes it easier for several programs that parse turbostat output. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-09-17 22:33:26 +02:00
if (!summary_only)
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
BIC_PRESENT(BIC_Core);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: Remember global max_die_id This is necessary to gracefully handle sparse die_id's. no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 11:56:48 -04:00
topo.num_die = topo.max_die_id + 1;
tools/power turbostat: Add Die column If the system has more than one software visible die per package, print a Die column. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-14 19:17:40 -08:00
if (debug > 1)
tools/power turbostat: Remember global max_die_id This is necessary to gracefully handle sparse die_id's. no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-21 11:56:48 -04:00
fprintf(outf, "max_die_id %d, sizing for %d die\n", topo.max_die_id, topo.num_die);
tools/power turbostat: Add Die column If the system has more than one software visible die per package, print a Die column. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-14 19:17:40 -08:00
if (!summary_only && topo.num_die > 1)
BIC_PRESENT(BIC_Die);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
topo.num_packages = max_package_id + 1;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
if (debug > 1)
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "max_package_id %d, sizing for %d packages\n", max_package_id, topo.num_packages);
tools/power turbostat: show package number, even without --debug On multi-package systems, the "Package" column was being displayed only if --debug was used. Show it always. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:43:41 -05:00
if (!summary_only && topo.num_packages > 1)
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
BIC_PRESENT(BIC_Package);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: Calculate additional node information for a package The code currently assumes each package has exactly one node. This is not the case for AMD systems and Intel systems with COD. AMD systems also may re-enumerate each node's core IDs starting at 0 (for example, an AMD processor may have two nodes, each with core IDs from 0 to 7). In order to properly enumerate the cores we need to track both the physical and logical node IDs. Add physical_node_id to track the node ID assigned by the kernel, and logical_node_id used by turbostat to track the nodes per package ie) a 0-based count within the package. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:30 -04:00
set_node_data();
if (debug > 1)
tools/power turbostat: remove num_ from cpu_topology struct Cleanup, remove num_ from num_nodes_per_pkg, num_cores_per_node, and num_threads_per_node. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:33 -04:00
fprintf(outf, "nodes_per_pkg %d\n", topo.nodes_per_pkg);
tools/power turbostat: Add Node in output Output a Node column if there is more than one node/socket. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:35 -04:00
if (!summary_only && topo.nodes_per_pkg > 1)
BIC_PRESENT(BIC_Node);
tools/power turbostat: Calculate additional node information for a package The code currently assumes each package has exactly one node. This is not the case for AMD systems and Intel systems with COD. AMD systems also may re-enumerate each node's core IDs starting at 0 (for example, an AMD processor may have two nodes, each with core IDs from 0 to 7). In order to properly enumerate the cores we need to track both the physical and logical node IDs. Add physical_node_id to track the node ID assigned by the kernel, and logical_node_id used by turbostat to track the nodes per package ie) a 0-based count within the package. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:30 -04:00
tools/power turbostat: remove num_ from cpu_topology struct Cleanup, remove num_ from num_nodes_per_pkg, num_cores_per_node, and num_threads_per_node. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:33 -04:00
topo.threads_per_core = max_siblings;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
if (debug > 1)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
fprintf(outf, "max_siblings %d\n", max_siblings);
tools/power turbostat: Fix logical node enumeration to allow for non-sequential physical nodes turbostat fails on some multi-package topologies because the logical node enumeration assumes that the nodes are sequentially numbered, which causes the logical numa nodes to not be enumerated, or enumerated incorrectly. Use a more robust enumeration algorithm which allows for non-seqential physical nodes. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-07-26 09:08:54 -04:00
if (debug < 1)
return;
for (i = 0; i <= topo.max_cpu_num; ++i) {
tools/power turbostat: reduce debug output A recent turbostat release increased topo.max_cpu_num to make it convenient to handle sysfs bitmaps of 32-cpus. But users, who regularly make use of "--debug", then saw a bunch of output for cpus that were not present. Remove that extra output by checking a cpu is online before dumping its info. Signed-off-by: Len Brown <len.brown@intel.com> Cc: Prarit Bhargava <prarit@redhat.com>
2018-09-21 22:26:57 -04:00
if (cpu_is_not_present(i))
continue;
tools/power turbostat: Fix logical node enumeration to allow for non-sequential physical nodes turbostat fails on some multi-package topologies because the logical node enumeration assumes that the nodes are sequentially numbered, which causes the logical numa nodes to not be enumerated, or enumerated incorrectly. Use a more robust enumeration algorithm which allows for non-seqential physical nodes. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-07-26 09:08:54 -04:00
fprintf(outf,
tools/power turbostat: Add Die column If the system has more than one software visible die per package, print a Die column. Signed-off-by: Len Brown <len.brown@intel.com>
2019-02-14 19:17:40 -08:00
"cpu %d pkg %d die %d node %d lnode %d core %d thread %d\n",
i, cpus[i].physical_package_id, cpus[i].die_id,
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
cpus[i].physical_node_id, cpus[i].logical_node_id, cpus[i].physical_core_id, cpus[i].thread_id);
tools/power turbostat: Fix logical node enumeration to allow for non-sequential physical nodes turbostat fails on some multi-package topologies because the logical node enumeration assumes that the nodes are sequentially numbered, which causes the logical numa nodes to not be enumerated, or enumerated incorrectly. Use a more robust enumeration algorithm which allows for non-seqential physical nodes. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-07-26 09:08:54 -04:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
void allocate_counters(struct thread_data **t, struct core_data **c, struct pkg_data **p)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
{
int i;
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
int num_cores = topo.cores_per_node * topo.nodes_per_pkg * topo.num_packages;
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
int num_threads = topo.threads_per_core * num_cores;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
*t = calloc(num_threads, sizeof(struct thread_data));
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
if (*t == NULL)
goto error;
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
for (i = 0; i < num_threads; i++)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
(*t)[i].cpu_id = -1;
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
*c = calloc(num_cores, sizeof(struct core_data));
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
if (*c == NULL)
goto error;
tools/power/turbostat: Obey allowed CPUs for primary thread/core detection Thread_id doesn't tell if a CPU is allowed or not. Detect allowed CPUs only and use the first detected thread/core as the primary thread/core of a core/package. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 09:59:14 +08:00
for (i = 0; i < num_cores; i++) {
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
(*c)[i].core_id = -1;
tools/power/turbostat: Obey allowed CPUs for primary thread/core detection Thread_id doesn't tell if a CPU is allowed or not. Detect allowed CPUs only and use the first detected thread/core as the primary thread/core of a core/package. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 09:59:14 +08:00
(*c)[i].base_cpu = -1;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: fix bugs in --add option When --add was used more than once, overflowed buffers caused some counters to be stored on top of others, corrupting the results. Simplify the code by simply reserving space for up to 16 added counters per each cpu, core, package. Per-cpu added counters were being printed only per-core. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 22:22:13 -05:00
*p = calloc(topo.num_packages, sizeof(struct pkg_data));
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
if (*p == NULL)
goto error;
tools/power/turbostat: Obey allowed CPUs for primary thread/core detection Thread_id doesn't tell if a CPU is allowed or not. Detect allowed CPUs only and use the first detected thread/core as the primary thread/core of a core/package. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 09:59:14 +08:00
for (i = 0; i < topo.num_packages; i++) {
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
(*p)[i].package_id = i;
tools/power/turbostat: Obey allowed CPUs for primary thread/core detection Thread_id doesn't tell if a CPU is allowed or not. Detect allowed CPUs only and use the first detected thread/core as the primary thread/core of a core/package. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 09:59:14 +08:00
(*p)[i].base_cpu = -1;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return;
error:
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:18 -07:00
err(1, "calloc counters");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/*
* init_counter()
*
* set FIRST_THREAD_IN_CORE and FIRST_CORE_IN_PACKAGE
*/
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
void init_counter(struct thread_data *thread_base, struct core_data *core_base, struct pkg_data *pkg_base, int cpu_id)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
{
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
int pkg_id = cpus[cpu_id].physical_package_id;
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
int node_id = cpus[cpu_id].logical_node_id;
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
int core_id = cpus[cpu_id].physical_core_id;
int thread_id = cpus[cpu_id].thread_id;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
struct thread_data *t;
struct core_data *c;
struct pkg_data *p;
tools/power turbostat: fix segfault on 'no node' machines Running turbostat on machines that don't expose nodes in sysfs (no /sys/bus/node) causes a segfault or a -nan value diesplayed in the log. This is caused by physical_node_id being reported as -1 and logical_node_id being calculated as a negative number resulting in the new GET_THREAD/GET_CORE returning an incorrect address. Signed-off-by: Nathan Ciobanu <nathan.d.ciobanu@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-08 15:15:12 -07:00
/* Workaround for systems where physical_node_id==-1
* and logical_node_id==(-1 - topo.num_cpus)
*/
if (node_id < 0)
node_id = 0;
tools/power turbostat: add node information into turbostat calculations The previous patches have added node information to turbostat, but the counters code does not take it into account. Add node information from cpu_topology calculations to turbostat counters. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:34 -04:00
t = GET_THREAD(thread_base, thread_id, core_id, node_id, pkg_id);
c = GET_CORE(core_base, core_id, node_id, pkg_id);
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
p = GET_PKG(pkg_base, pkg_id);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
t->cpu_id = cpu_id;
tools/power/turbostat: Obey allowed CPUs for primary thread/core detection Thread_id doesn't tell if a CPU is allowed or not. Detect allowed CPUs only and use the first detected thread/core as the primary thread/core of a core/package. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 09:59:14 +08:00
if (!cpu_is_not_allowed(cpu_id)) {
if (c->base_cpu < 0)
c->base_cpu = t->cpu_id;
if (p->base_cpu < 0)
p->base_cpu = t->cpu_id;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
c->core_id = core_id;
p->package_id = pkg_id;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
int initialize_counters(int cpu_id)
{
tools/power turbostat: track thread ID in cpu_topology The code can be simplified if the cpu_topology *cpus tracks the thread IDs. This removes an additional file lookup and simplifies the counter initialization code. Add thread ID to cpu_topology information and cleanup the counter initialization code. v2: prevent thread_id from being overwritten Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 10:04:31 -04:00
init_counter(EVEN_COUNTERS, cpu_id);
init_counter(ODD_COUNTERS, cpu_id);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
return 0;
}
void allocate_output_buffer()
{
tools/power turbostat: fix buffer overrun turbostat could be terminated by general protection fault on some latest hardwares which (for example) support 9 levels of C-states and show 18 "tADDED" lines. That bloats the total output and finally causes buffer overrun. So let's extend the buffer to avoid this. Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Signed-off-by: Len Brown <len.brown@intel.com>
2019-04-03 16:02:14 +09:00
output_buffer = calloc(1, (1 + topo.num_cpus) * 2048);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
outp = output_buffer;
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:18 -07:00
if (outp == NULL)
err(-1, "calloc output buffer");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
void allocate_fd_percpu(void)
{
tools/power turbostat: Allocate correct amount of fd and irq entries The tool uses topo.max_cpu_num to determine number of entries needed for fd_percpu[] and irqs_per_cpu[]. For example on a system with 4 CPUs topo.max_cpu_num is 3 so we get too small array for holding per-CPU items. Fix this to use right number of entries, which is topo.max_cpu_num + 1. Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 11:13:23 +03:00
fd_percpu = calloc(topo.max_cpu_num + 1, sizeof(int));
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
if (fd_percpu == NULL)
err(-1, "calloc fd_percpu");
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
void allocate_irq_buffers(void)
{
irq_column_2_cpu = calloc(topo.num_cpus, sizeof(int));
if (irq_column_2_cpu == NULL)
err(-1, "calloc %d", topo.num_cpus);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
tools/power turbostat: Allocate correct amount of fd and irq entries The tool uses topo.max_cpu_num to determine number of entries needed for fd_percpu[] and irqs_per_cpu[]. For example on a system with 4 CPUs topo.max_cpu_num is 3 so we get too small array for holding per-CPU items. Fix this to use right number of entries, which is topo.max_cpu_num + 1. Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 11:13:23 +03:00
irqs_per_cpu = calloc(topo.max_cpu_num + 1, sizeof(int));
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
if (irqs_per_cpu == NULL)
tools/power turbostat: Allocate correct amount of fd and irq entries The tool uses topo.max_cpu_num to determine number of entries needed for fd_percpu[] and irqs_per_cpu[]. For example on a system with 4 CPUs topo.max_cpu_num is 3 so we get too small array for holding per-CPU items. Fix this to use right number of entries, which is topo.max_cpu_num + 1. Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 11:13:23 +03:00
err(-1, "calloc %d", topo.max_cpu_num + 1);
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
int update_topo(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
topo.allowed_cpus++;
if ((int)t->cpu_id == c->base_cpu)
topo.allowed_cores++;
if ((int)t->cpu_id == p->base_cpu)
topo.allowed_packages++;
return 0;
}
void topology_update(void)
{
topo.allowed_cpus = 0;
topo.allowed_cores = 0;
topo.allowed_packages = 0;
for_all_cpus(update_topo, ODD_COUNTERS);
}
tools/power turbostat: Do not print negative LPI residency turbostat prints the abnormal SYS%LPI across suspend-to-idle: SYS%LPI = 114479815993277.50 This is reproduced by: Run a freeze cycle, e.g. "sleepgraph -m freeze -rtcwake 15". Then do a reboot. After boot up, launch the suspend-idle-idle and check the SYS%LPI field. The slp_so residence counter is in LPIT table, and BIOS does not clears this register across reset. The PMC expects the OS to calculate the LPI residency based on the delta. However, there is an firmware issue that the LPIT gets cleared to 0 during the second suspend to idle after the reboot, which brings negative delta value. [lenb: updated to print "neg" upon this BIOS failure] Reported-by: Todd Brandt <todd.e.brandt@intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2023-10-22 13:52:21 +08:00
tools/power/turbostat: Handle offlined CPUs in cpu_subset It is possible that the cpu_subset contains offlined CPUs. If this happens during start, exit immediately because this is likely an operator error that is best fixed by re-invoking. If this happens at runtime, give a warning only because turbostat should do its best effort to continue running. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 14:20:47 +08:00
void setup_all_buffers(bool startup)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
{
tools/power/turbostat: Handle offlined CPUs in cpu_subset It is possible that the cpu_subset contains offlined CPUs. If this happens during start, exit immediately because this is likely an operator error that is best fixed by re-invoking. If this happens at runtime, give a warning only because turbostat should do its best effort to continue running. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 14:20:47 +08:00
topology_probe(startup);
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 23:48:05 -05:00
allocate_irq_buffers();
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-26 20:51:02 -05:00
allocate_fd_percpu();
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
allocate_counters(&thread_even, &core_even, &package_even);
allocate_counters(&thread_odd, &core_odd, &package_odd);
allocate_output_buffer();
for_all_proc_cpus(initialize_counters);
tools/power/turbostat: Obey allowed CPUs for system summary System summary should summarize the information for allowed CPUs instead of all the present CPUs. Introduce topology information for allowed CPUs, and use them to get system summary. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-06 18:35:26 +08:00
topology_update();
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
}
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 08:34:28 -04:00
void set_base_cpu(void)
{
tools/power/turbostat: Obey allowed CPUs during startup Set turbostat CPU affinity to make sure turbostat is running on one of the allowed CPUs. Set base_cpu to the first allowed CPU so that some platform information is dumped using one of the allowed CPUs. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 14:25:25 +08:00
int i;
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 08:34:28 -04:00
tools/power/turbostat: Obey allowed CPUs during startup Set turbostat CPU affinity to make sure turbostat is running on one of the allowed CPUs. Set base_cpu to the first allowed CPU so that some platform information is dumped using one of the allowed CPUs. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-04 14:25:25 +08:00
for (i = 0; i < topo.max_cpu_num + 1; ++i) {
if (cpu_is_not_allowed(i))
continue;
base_cpu = i;
if (debug > 1)
fprintf(outf, "base_cpu = %d\n", base_cpu);
return;
}
err(-ENODEV, "No valid cpus found");
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 08:34:28 -04:00
}
tools/power turbostat: Clear added counters when in no-msr mode If user request --no-msr or is not able to access the MSRs, turbostat should clear all the counters added with --add. Because MSR access permission checks are done after the cmdline is parsed, the decision has to be defered up until the transition into no-msr mode happen. Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-14 11:36:55 +01:00
bool has_added_counters(void)
{
/*
* It only makes sense to call this after the command line is parsed,
* otherwise sys structure is not populated.
*/
return sys.added_core_counters | sys.added_thread_counters | sys.added_package_counters;
}
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
void check_msr_access(void)
{
check_dev_msr();
check_msr_permission();
if (no_msr)
bic_disable_msr_access();
}
void check_perf_access(void)
{
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
if (no_perf || !BIC_IS_ENABLED(BIC_IPC) || !has_instr_count_access())
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
bic_enabled &= ~BIC_IPC;
}
tools/power turbostat: Allow using cpu device in perf counters on hybrid platforms Intel hybrid platforms expose different perf devices for P and E cores. Instead of one, "/sys/bus/event_source/devices/cpu" device, there are "/sys/bus/event_source/devices/{cpu_core,cpu_atom}". This, however makes it more complicated for the user, because most of the counters are available on both and had to be handled manually. This patch allows users to use "virtual" cpu device that is seemingly translated to cpu_core and cpu_atom perf devices, depending on the type of a CPU we are opening the counter for. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-20 18:47:59 +02:00
bool perf_has_hybrid_devices(void)
{
/*
* 0: unknown
* 1: has separate perf device for p and e core
* -1: doesn't have separate perf device for p and e core
*/
static int cached;
if (cached > 0)
return true;
if (cached < 0)
return false;
if (access("/sys/bus/event_source/devices/cpu_core", F_OK)) {
cached = -1;
return false;
}
if (access("/sys/bus/event_source/devices/cpu_atom", F_OK)) {
cached = -1;
return false;
}
cached = 1;
return true;
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
int added_perf_counters_init_(struct perf_counter_info *pinfo)
{
size_t num_domains = 0;
unsigned int next_domain;
bool *domain_visited;
unsigned int perf_type, perf_config;
double perf_scale;
int fd_perf;
if (!pinfo)
return 0;
const size_t max_num_domains = MAX(topo.max_cpu_num + 1, MAX(topo.max_core_id + 1, topo.max_package_id + 1));
domain_visited = calloc(max_num_domains, sizeof(*domain_visited));
while (pinfo) {
switch (pinfo->scope) {
case SCOPE_CPU:
num_domains = topo.max_cpu_num + 1;
break;
case SCOPE_CORE:
num_domains = topo.max_core_id + 1;
break;
case SCOPE_PACKAGE:
num_domains = topo.max_package_id + 1;
break;
}
/* Allocate buffer for file descriptor for each domain. */
pinfo->fd_perf_per_domain = calloc(num_domains, sizeof(*pinfo->fd_perf_per_domain));
if (!pinfo->fd_perf_per_domain)
errx(1, "%s: alloc %s", __func__, "fd_perf_per_domain");
for (size_t i = 0; i < num_domains; ++i)
pinfo->fd_perf_per_domain[i] = -1;
pinfo->num_domains = num_domains;
pinfo->scale = 1.0;
memset(domain_visited, 0, max_num_domains * sizeof(*domain_visited));
for (int cpu = 0; cpu < topo.max_cpu_num + 1; ++cpu) {
next_domain = cpu_to_domain(pinfo, cpu);
assert(next_domain < num_domains);
if (cpu_is_not_allowed(cpu))
continue;
if (domain_visited[next_domain])
continue;
tools/power turbostat: Allow using cpu device in perf counters on hybrid platforms Intel hybrid platforms expose different perf devices for P and E cores. Instead of one, "/sys/bus/event_source/devices/cpu" device, there are "/sys/bus/event_source/devices/{cpu_core,cpu_atom}". This, however makes it more complicated for the user, because most of the counters are available on both and had to be handled manually. This patch allows users to use "virtual" cpu device that is seemingly translated to cpu_core and cpu_atom perf devices, depending on the type of a CPU we are opening the counter for. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-20 18:47:59 +02:00
/*
* Intel hybrid platforms expose different perf devices for P and E cores.
* Instead of one, "/sys/bus/event_source/devices/cpu" device, there are
* "/sys/bus/event_source/devices/{cpu_core,cpu_atom}".
*
* This makes it more complicated to the user, because most of the counters
* are available on both and have to be handled manually, otherwise.
*
* Code below, allow user to use the old "cpu" name, which is translated accordingly.
*/
const char *perf_device = pinfo->device;
if (strcmp(perf_device, "cpu") == 0 && perf_has_hybrid_devices()) {
switch (cpus[cpu].type) {
case INTEL_PCORE_TYPE:
perf_device = "cpu_core";
break;
case INTEL_ECORE_TYPE:
perf_device = "cpu_atom";
break;
default: /* Don't change, we will probably fail and report a problem soon. */
break;
}
}
perf_type = read_perf_type(perf_device);
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
if (perf_type == (unsigned int)-1) {
warnx("%s: perf/%s/%s: failed to read %s",
tools/power turbostat: Allow using cpu device in perf counters on hybrid platforms Intel hybrid platforms expose different perf devices for P and E cores. Instead of one, "/sys/bus/event_source/devices/cpu" device, there are "/sys/bus/event_source/devices/{cpu_core,cpu_atom}". This, however makes it more complicated for the user, because most of the counters are available on both and had to be handled manually. This patch allows users to use "virtual" cpu device that is seemingly translated to cpu_core and cpu_atom perf devices, depending on the type of a CPU we are opening the counter for. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-20 18:47:59 +02:00
__func__, perf_device, pinfo->event, "type");
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
continue;
}
tools/power turbostat: Allow using cpu device in perf counters on hybrid platforms Intel hybrid platforms expose different perf devices for P and E cores. Instead of one, "/sys/bus/event_source/devices/cpu" device, there are "/sys/bus/event_source/devices/{cpu_core,cpu_atom}". This, however makes it more complicated for the user, because most of the counters are available on both and had to be handled manually. This patch allows users to use "virtual" cpu device that is seemingly translated to cpu_core and cpu_atom perf devices, depending on the type of a CPU we are opening the counter for. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-20 18:47:59 +02:00
perf_config = read_perf_config(perf_device, pinfo->event);
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
if (perf_config == (unsigned int)-1) {
warnx("%s: perf/%s/%s: failed to read %s",
tools/power turbostat: Allow using cpu device in perf counters on hybrid platforms Intel hybrid platforms expose different perf devices for P and E cores. Instead of one, "/sys/bus/event_source/devices/cpu" device, there are "/sys/bus/event_source/devices/{cpu_core,cpu_atom}". This, however makes it more complicated for the user, because most of the counters are available on both and had to be handled manually. This patch allows users to use "virtual" cpu device that is seemingly translated to cpu_core and cpu_atom perf devices, depending on the type of a CPU we are opening the counter for. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-20 18:47:59 +02:00
__func__, perf_device, pinfo->event, "config");
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
continue;
}
/* Scale is not required, some counters just don't have it. */
tools/power turbostat: Allow using cpu device in perf counters on hybrid platforms Intel hybrid platforms expose different perf devices for P and E cores. Instead of one, "/sys/bus/event_source/devices/cpu" device, there are "/sys/bus/event_source/devices/{cpu_core,cpu_atom}". This, however makes it more complicated for the user, because most of the counters are available on both and had to be handled manually. This patch allows users to use "virtual" cpu device that is seemingly translated to cpu_core and cpu_atom perf devices, depending on the type of a CPU we are opening the counter for. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-20 18:47:59 +02:00
perf_scale = read_perf_scale(perf_device, pinfo->event);
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
if (perf_scale == 0.0)
perf_scale = 1.0;
fd_perf = open_perf_counter(cpu, perf_type, perf_config, -1, 0);
if (fd_perf == -1) {
warnx("%s: perf/%s/%s: failed to open counter on cpu%d",
tools/power turbostat: Allow using cpu device in perf counters on hybrid platforms Intel hybrid platforms expose different perf devices for P and E cores. Instead of one, "/sys/bus/event_source/devices/cpu" device, there are "/sys/bus/event_source/devices/{cpu_core,cpu_atom}". This, however makes it more complicated for the user, because most of the counters are available on both and had to be handled manually. This patch allows users to use "virtual" cpu device that is seemingly translated to cpu_core and cpu_atom perf devices, depending on the type of a CPU we are opening the counter for. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-20 18:47:59 +02:00
__func__, perf_device, pinfo->event, cpu);
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
continue;
}
domain_visited[next_domain] = 1;
pinfo->fd_perf_per_domain[next_domain] = fd_perf;
pinfo->scale = perf_scale;
if (debug)
tools/power turbostat: Move debug prints from stdout to stderr This leaves the stdout cleaner, having only counter data. It makes it easier for programs to parse the output of turbostat, for example selftests. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 11:34:38 +02:00
fprintf(stderr, "Add perf/%s/%s cpu%d: %d\n",
tools/power turbostat: Allow using cpu device in perf counters on hybrid platforms Intel hybrid platforms expose different perf devices for P and E cores. Instead of one, "/sys/bus/event_source/devices/cpu" device, there are "/sys/bus/event_source/devices/{cpu_core,cpu_atom}". This, however makes it more complicated for the user, because most of the counters are available on both and had to be handled manually. This patch allows users to use "virtual" cpu device that is seemingly translated to cpu_core and cpu_atom perf devices, depending on the type of a CPU we are opening the counter for. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-08-20 18:47:59 +02:00
perf_device, pinfo->event, cpu, pinfo->fd_perf_per_domain[next_domain]);
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
}
pinfo = pinfo->next;
}
free(domain_visited);
return 0;
}
void added_perf_counters_init(void)
{
if (added_perf_counters_init_(sys.perf_tp))
errx(1, "%s: %s", __func__, "thread");
if (added_perf_counters_init_(sys.perf_cp))
errx(1, "%s: %s", __func__, "core");
if (added_perf_counters_init_(sys.perf_pp))
errx(1, "%s: %s", __func__, "package");
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
int parse_telem_info_file(int fd_dir, const char *info_filename, const char *format, unsigned long *output)
{
int fd_telem_info;
FILE *file_telem_info;
unsigned long value;
fd_telem_info = openat(fd_dir, info_filename, O_RDONLY);
if (fd_telem_info == -1)
return -1;
file_telem_info = fdopen(fd_telem_info, "r");
if (file_telem_info == NULL) {
close(fd_telem_info);
return -1;
}
if (fscanf(file_telem_info, format, &value) != 1) {
fclose(file_telem_info);
return -1;
}
fclose(file_telem_info);
*output = value;
return 0;
}
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
struct pmt_mmio *pmt_mmio_open(unsigned int target_guid)
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
{
DIR *dirp;
struct dirent *entry;
struct stat st;
unsigned int telem_idx;
int fd_telem_dir, fd_pmt;
unsigned long guid, size, offset;
size_t mmap_size;
void *mmio;
struct pmt_mmio *ret = NULL;
if (stat(SYSFS_TELEM_PATH, &st) == -1)
return NULL;
dirp = opendir(SYSFS_TELEM_PATH);
if (dirp == NULL)
return NULL;
for (;;) {
entry = readdir(dirp);
if (entry == NULL)
break;
if (strcmp(entry->d_name, ".") == 0)
continue;
if (strcmp(entry->d_name, "..") == 0)
continue;
if (sscanf(entry->d_name, "telem%u", &telem_idx) != 1)
continue;
if (fstatat(dirfd(dirp), entry->d_name, &st, 0) == -1) {
break;
}
if (!S_ISDIR(st.st_mode))
continue;
fd_telem_dir = openat(dirfd(dirp), entry->d_name, O_RDONLY);
if (fd_telem_dir == -1) {
break;
}
if (parse_telem_info_file(fd_telem_dir, "guid", "%lx", &guid)) {
close(fd_telem_dir);
break;
}
if (parse_telem_info_file(fd_telem_dir, "size", "%lu", &size)) {
close(fd_telem_dir);
break;
}
if (guid != target_guid) {
close(fd_telem_dir);
continue;
}
if (parse_telem_info_file(fd_telem_dir, "offset", "%lu", &offset)) {
close(fd_telem_dir);
break;
}
assert(offset == 0);
fd_pmt = openat(fd_telem_dir, "telem", O_RDONLY);
if (fd_pmt == -1)
goto loop_cleanup_and_break;
mmap_size = (size + 0x1000UL) & (~0x1000UL);
mmio = mmap(0, mmap_size, PROT_READ, MAP_SHARED, fd_pmt, 0);
if (mmio != MAP_FAILED) {
if (debug)
fprintf(stderr, "%s: 0x%lx mmaped at: %p\n", __func__, guid, mmio);
ret = calloc(1, sizeof(*ret));
if (!ret) {
fprintf(stderr, "%s: Failed to allocate pmt_mmio\n", __func__);
exit(1);
}
ret->guid = guid;
ret->mmio_base = mmio;
ret->pmt_offset = offset;
ret->size = size;
ret->next = pmt_mmios;
pmt_mmios = ret;
}
loop_cleanup_and_break:
close(fd_pmt);
close(fd_telem_dir);
break;
}
closedir(dirp);
return ret;
}
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
struct pmt_mmio *pmt_mmio_find(unsigned int guid)
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
{
struct pmt_mmio *pmmio = pmt_mmios;
while (pmmio) {
if (pmmio->guid == guid)
return pmmio;
pmmio = pmmio->next;
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
return NULL;
}
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
void *pmt_get_counter_pointer(struct pmt_mmio *pmmio, unsigned long counter_offset)
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
{
char *ret;
/* Get base of mmaped PMT file. */
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
ret = (char *)pmmio->mmio_base;
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
/*
* Apply PMT MMIO offset to obtain beginning of the mmaped telemetry data.
* It's not guaranteed that the mmaped memory begins with the telemetry data
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
* - we might have to apply the offset first.
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
*/
ret += pmmio->pmt_offset;
/* Apply the counter offset to get the address to the mmaped counter. */
ret += counter_offset;
return ret;
}
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
struct pmt_mmio *pmt_add_guid(unsigned int guid)
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
{
struct pmt_mmio *ret;
ret = pmt_mmio_find(guid);
if (!ret)
ret = pmt_mmio_open(guid);
return ret;
}
enum pmt_open_mode {
PMT_OPEN_TRY, /* Open failure is not an error. */
PMT_OPEN_REQUIRED, /* Open failure is a fatal error. */
};
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
struct pmt_counter *pmt_find_counter(struct pmt_counter *pcounter, const char *name)
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
{
while (pcounter) {
if (strcmp(pcounter->name, name) == 0)
break;
pcounter = pcounter->next;
}
return pcounter;
}
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
struct pmt_counter **pmt_get_scope_root(enum counter_scope scope)
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
{
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
switch (scope) {
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
case SCOPE_CPU:
return &sys.pmt_tp;
case SCOPE_CORE:
return &sys.pmt_cp;
case SCOPE_PACKAGE:
return &sys.pmt_pp;
}
__builtin_unreachable();
}
void pmt_counter_add_domain(struct pmt_counter *pcounter, unsigned long *pmmio, unsigned int domain_id)
{
/* Make sure the new domain fits. */
if (domain_id >= pcounter->num_domains)
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
pmt_counter_resize(pcounter, domain_id + 1);
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
assert(pcounter->domains);
assert(domain_id < pcounter->num_domains);
pcounter->domains[domain_id].pcounter = pmmio;
}
int pmt_add_counter(unsigned int guid, const char *name, enum pmt_datatype type,
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
unsigned int lsb, unsigned int msb, unsigned int offset, enum counter_scope scope,
enum counter_format format, unsigned int domain_id, enum pmt_open_mode mode)
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
{
struct pmt_mmio *mmio;
struct pmt_counter *pcounter;
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
struct pmt_counter **const pmt_root = pmt_get_scope_root(scope);
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
bool new_counter = false;
int conflict = 0;
if (lsb > msb) {
fprintf(stderr, "%s: %s: `%s` must be satisfied\n", __func__, "lsb <= msb", name);
exit(1);
}
if (msb >= 64) {
fprintf(stderr, "%s: %s: `%s` must be satisfied\n", __func__, "msb < 64", name);
exit(1);
}
mmio = pmt_add_guid(guid);
if (!mmio) {
if (mode != PMT_OPEN_TRY) {
fprintf(stderr, "%s: failed to map PMT MMIO for guid %x\n", __func__, guid);
exit(1);
}
return 1;
}
if (offset >= mmio->size) {
if (mode != PMT_OPEN_TRY) {
fprintf(stderr, "%s: offset %u outside of PMT MMIO size %u\n", __func__, offset, mmio->size);
exit(1);
}
return 1;
}
pcounter = pmt_find_counter(*pmt_root, name);
if (!pcounter) {
pcounter = calloc(1, sizeof(*pcounter));
new_counter = true;
}
if (new_counter) {
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
strncpy(pcounter->name, name, ARRAY_SIZE(pcounter->name) - 1);
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
pcounter->type = type;
pcounter->scope = scope;
pcounter->lsb = lsb;
pcounter->msb = msb;
pcounter->format = format;
} else {
conflict += pcounter->type != type;
conflict += pcounter->scope != scope;
conflict += pcounter->lsb != lsb;
conflict += pcounter->msb != msb;
conflict += pcounter->format != format;
}
if (conflict) {
fprintf(stderr, "%s: conflicting parameters for the PMT counter with the same name %s\n",
__func__, name);
exit(1);
}
pmt_counter_add_domain(pcounter, pmt_get_counter_pointer(mmio, offset), domain_id);
if (new_counter) {
pcounter->next = *pmt_root;
*pmt_root = pcounter;
}
return 0;
}
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
void pmt_init(void)
{
if (BIC_IS_ENABLED(BIC_Diec6)) {
pmt_add_counter(PMT_MTL_DC6_GUID, "Die%c6", PMT_TYPE_XTAL_TIME, PMT_COUNTER_MTL_DC6_LSB,
PMT_COUNTER_MTL_DC6_MSB, PMT_COUNTER_MTL_DC6_OFFSET, SCOPE_PACKAGE, FORMAT_DELTA,
0, PMT_OPEN_TRY);
}
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
void turbostat_init()
{
tools/power/turbostat: Handle offlined CPUs in cpu_subset It is possible that the cpu_subset contains offlined CPUs. If this happens during start, exit immediately because this is likely an operator error that is best fixed by re-invoking. If this happens at runtime, give a warning only because turbostat should do its best effort to continue running. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 14:20:47 +08:00
setup_all_buffers(true);
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 08:34:28 -04:00
set_base_cpu();
tools/power turbostat: detect and disable unavailable BICs at runtime To allow unprivileged user to run turbostat seamlessly. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-15 19:04:21 +01:00
check_msr_access();
check_perf_access();
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-23 20:29:09 -04:00
process_cpuid();
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
counter_info_init();
tools/power/turbostat: Introduce probe_pm_features() Feature probe has nothing to do with CPUID, thus it should not be in process_cpuids(). Introduce probe_pm_features() and move all feature probing functions into it. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Reviewed-by: Len Brown <len.brown@intel.com>
2023-09-13 23:32:39 +08:00
probe_pm_features();
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
msr_perf_init();
tools/power turbostat: add built-in-counter for IPC -- Instructions per Cycle Use linux-perf to access the hardware instructions-retired counter. This is necessary because the counter is not enabled by default, and also the counter is prone to roll-over -- both of which perf manages. It is not necessary to use perf for the cycle counter, because turbostat already needs to collect delta-aperf to calcuate frequency. Signed-off-by: Len Brown <len.brown@intel.com>
2021-01-30 23:43:10 -05:00
linux_perf_init();
tools/power turbostat: read RAPL counters via perf Some of the future Intel platforms will require reading the RAPL counters via perf and not MSR. On current platforms we can still read them using both ways. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-05 12:27:27 +01:00
rapl_perf_init();
tools/power turbostat: Read Core-cstates via perf Reading the counters via perf can be done in bulk with a single syscall, making the counter values more accurate with respect to one another by minimizing the time gap between individual counter reads. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-03-11 18:06:16 +01:00
cstate_perf_init();
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
added_perf_counters_init();
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
pmt_init();
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: Fix Core C6 residency on Atom CPUs For Atom CPUs that have core cstate deeper than C6, MSR_CORE_C6_RESIDENCY actually returns the residency for both CC6 and deeper Core cstates. Thus, the real Core C6 residency should be the subtraction of MSR_CORE_C6_RESIDENCY return value and MSR_CORE_C6_RESIDENCY return value. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-21 22:22:47 +08:00
for_all_cpus(get_cpu_type, ODD_COUNTERS);
for_all_cpus(get_cpu_type, EVEN_COUNTERS);
tools/power turbostat: Group SMI counter with APERF and MPERF These three counters now are treated similar to other perf counters groups. This simplifies and gets rid of a lot of special cases for APERF and MPERF. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-24 14:31:52 +02:00
if (BIC_IS_ENABLED(BIC_IPC) && has_aperf_access && get_instr_count_fd(base_cpu) != -1)
BIC_PRESENT(BIC_IPC);
tools/power turbostat: Add reading aperf and mperf via perf API By using the perf API we spend less time in between the reads of the counters, resulting in more accurate calculations of the dependent metrics. Using perf API is also usually faster overall, although cache miss, if we get one, is more costly when using perf vs MSR driver. We would fallback to the msr reads if the sysfs isn't there or when in --no-perf mode. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 16:42:22 +01:00
/*
* If TSC tweak is needed, but couldn't get it,
* disable more BICs, since it can't be reported accurately.
*/
if (platform->enable_tsc_tweak && !has_base_hz) {
bic_enabled &= ~BIC_Busy;
bic_enabled &= ~BIC_Bzy_MHz;
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
int fork_it(char **argv)
{
pid_t child_pid;
tools/power turbostat: graceful fail on garbage input When invald MSR's are specified on the command line, turbostat should simply print an error and exit. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-01 00:08:19 -04:00
int status;
tools/power turbostat: fix un-intended affinity of forked program Linux 3.4 included a modification to turbostat to lower cross-call overhead by using scheduler affinity: 15aaa34654831e98dd76f7738b6c7f5d05a66430 (tools turbostat: reduce measurement overhead due to IPIs) In the use-case where turbostat forks a child program, that change had the un-intended side-effect of binding the child to the last cpu in the system. This change removed the binding before forking the child. This is a back-port of a fix already included in turbostat v2. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-03 23:24:00 -04:00
tools/power turbostat: fix zero IRQ count shown in one-shot command mode The IRQ column has been working for periodic mode, but not in one-shot command mode, it shows only 0. until now. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-14 22:07:52 -05:00
snapshot_proc_sysfs_files();
tools/power turbostat: graceful fail on garbage input When invald MSR's are specified on the command line, turbostat should simply print an error and exit. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-01 00:08:19 -04:00
status = for_all_cpus(get_counters, EVEN_COUNTERS);
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
first_counter_read = 0;
tools/power turbostat: graceful fail on garbage input When invald MSR's are specified on the command line, turbostat should simply print an error and exit. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-01 00:08:19 -04:00
if (status)
exit(status);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
gettimeofday(&tv_even, (struct timezone *)NULL);
child_pid = fork();
if (!child_pid) {
/* child */
execvp(argv[0], argv);
tools/power turbostat: show error on exec When turbostat is run in one-shot command mode, the parent takes the 'before' counter snapshot, fork/exec/wait for the child to exit, takes the 'after' counter snapshot, and prints the results. however, if the child fails to exec the command, it immediately returns, without indicating that anythign was wrong. Add an error message showing that exec failed: sudo turbostat sleeeep 4 ... turbostat: exec sleeeep: No such file or directory ... Note that the parent will still print out the statistics, because it can't tell the difference between the failed exec and a command that is purposefully returning the same status. Unfortunately, this may obscure the error message. However, if the --out parameter is used, the error message is evident on stderr. Reported-by: Wendy Wang <wendy.wang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-23 17:00:51 -05:00
err(errno, "exec %s", argv[0]);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
} else {
/* parent */
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:18 -07:00
if (child_pid == -1)
err(1, "fork");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
signal(SIGINT, SIG_IGN);
signal(SIGQUIT, SIG_IGN);
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-20 17:20:18 -07:00
if (waitpid(child_pid, &status, 0) == -1)
err(status, "waitpid");
tools/power turbostat: return the exit status of a command turbostat failed to return a non-zero exit status even though the supplied command (turbostat <command>) failed. Currently when turbostat forks a command it returns zero instead of the actual exit status of the command. Modify the code to return the exit status. Signed-off-by: David Arcari <darcari@redhat.com> Acked-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2019-02-12 09:34:39 -05:00
if (WIFEXITED(status))
status = WEXITSTATUS(status);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
/*
* n.b. fork_it() does not check for errors from for_all_cpus()
* because re-starting is problematic when forking
*/
tools/power turbostat: fix zero IRQ count shown in one-shot command mode The IRQ column has been working for periodic mode, but not in one-shot command mode, it shows only 0. until now. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-14 22:07:52 -05:00
snapshot_proc_sysfs_files();
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 00:56:40 -04:00
for_all_cpus(get_counters, ODD_COUNTERS);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
gettimeofday(&tv_odd, (struct timezone *)NULL);
timersub(&tv_odd, &tv_even, &tv_delta);
tools/power turbostat: Gracefully handle ACPI S3 turbostat gives valid results across suspend to idle, aka freeze, whether invoked in interval mode, or in command mode. Indeed, this can be used to measure suspend to idle: turbostat echo freeze > /sys/power/state But this does not work across suspend to ACPI S3, because the processor counters, including the TSC, are reset on resume. Further, when turbostat detects a problem, it does't forgive the hardware, and interval mode will print *'s from there on out. Instead, upon detecting counters going backwards, simply reset and start over. Interval mode across ACPI S3: (observe TSC going backwards) root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 1 0.06 858 2294 0x0000000000000000 0 0 0.06 847 2294 0x0000002a254b98ac 1 1 0.06 878 2294 0x0000002a254efa3a 2 1 0.07 843 2294 0x0000002a2551df65 3 0 0.05 863 2294 0x0000002a2553fea2 turbostat: re-initialized with num_cpus 4 CPU Avg_MHz Busy% Bzy_MHz TSC_MHz MSR 0x010 - 2 0.20 849 2294 0x0000000000000000 0 2 0.26 856 2294 0x0000000449abb60d 1 2 0.20 844 2294 0x0000000449b087ec 2 2 0.21 850 2294 0x0000000449b35d5d 3 1 0.12 839 2294 0x0000000449b5fd5a ^C Command mode across ACPI S3: root@sharkbay:/home/lenb/turbostat-src# ./turbostat -M 0x10 sleep 10 ./turbostat: Counter reset detected 14.196299 sec Signed-off-by: Len Brown <len.brown@intel.com>
2016-04-22 20:31:46 -04:00
if (for_all_cpus_2(delta_cpu, ODD_COUNTERS, EVEN_COUNTERS))
fprintf(outf, "%s: Counter reset detected\n", progname);
else {
compute_average(EVEN_COUNTERS);
format_all_counters(EVEN_COUNTERS);
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "%.6f sec\n", tv_delta.tv_sec + tv_delta.tv_usec / 1000000.0);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
flush_output_stderr();
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: graceful fail on garbage input When invald MSR's are specified on the command line, turbostat should simply print an error and exit. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-01 00:08:19 -04:00
return status;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
int get_and_dump_counters(void)
{
int status;
tools/power turbostat: fix zero IRQ count shown in one-shot command mode The IRQ column has been working for periodic mode, but not in one-shot command mode, it shows only 0. until now. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-14 22:07:52 -05:00
snapshot_proc_sysfs_files();
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
status = for_all_cpus(get_counters, ODD_COUNTERS);
if (status)
return status;
status = for_all_cpus(dump_counters, ODD_COUNTERS);
if (status)
return status;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
flush_output_stdout();
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
return status;
}
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
void print_version()
{
tools/power turbostat: 2024.11.30 since 2024.07.26: assorted minor bug fixes assorted platform specific tweaks initial RAPL PSYS (SysWatt) support Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-30 16:22:00 -05:00
fprintf(outf, "turbostat version 2024.11.30 - Len Brown <lenb@kernel.org>\n");
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
}
tools/power turbostat: print the kernel boot commandline It would be handy to have cmdline in turbostat output. For example, according to the turbostat output, there are no C-states requested. In this case the user is very curious if something like intel_idle.max_cstate=0 was used, or may be idle=none too. It is also curious whether things like intel_pstate=nohwp were used. Print the boot command line accordingly: turbostat version 21.05.04 - Len Brown <lenb@kernel.org> Kernel command line: BOOT_IMAGE=/boot/vmlinuz-5.16.0+ root=UUID= b42359ed-1e05-42eb-8757-6bf2a1c19070 ro quiet splash vt.handoff=7 Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-21 22:38:34 +08:00
#define COMMAND_LINE_SIZE 2048
void print_bootcmd(void)
{
char bootcmd[COMMAND_LINE_SIZE];
FILE *fp;
int ret;
memset(bootcmd, 0, COMMAND_LINE_SIZE);
fp = fopen("/proc/cmdline", "r");
if (!fp)
return;
ret = fread(bootcmd, sizeof(char), COMMAND_LINE_SIZE - 1, fp);
if (ret) {
bootcmd[ret] = '\0';
/* the last character is already '\n' */
fprintf(outf, "Kernel command line: %s", bootcmd);
}
fclose(fp);
}
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
struct msr_counter *find_msrp_by_name(struct msr_counter *head, char *name)
{
struct msr_counter *mp;
for (mp = head; mp; mp = mp->next) {
if (debug)
tools/power turbostat: Move debug prints from stdout to stderr This leaves the stdout cleaner, having only counter data. It makes it easier for programs to parse the output of turbostat, for example selftests. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 11:34:38 +02:00
fprintf(stderr, "%s: %s %s\n", __func__, name, mp->name);
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
if (!strncmp(name, mp->name, strlen(mp->name)))
return mp;
}
return NULL;
}
tools/power turbostat: extend --add option to accept /sys path Previously, the --add option could specify only an MSR. Here is is extended so an arbitrary /sys attribute, as specified by an absolute file path name. sudo ./turbostat --add /sys/devices/system/cpu/cpu0/cpuidle/state5/usage Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-08 02:41:51 -05:00
int add_counter(unsigned int msr_num, char *path, char *name,
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
unsigned int width, enum counter_scope scope,
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
enum counter_type type, enum counter_format format, int flags, int id)
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
{
struct msr_counter *msrp;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (no_msr && msr_num)
errx(1, "Requested MSR counter 0x%x, but in --no-msr mode", msr_num);
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
if (debug)
tools/power turbostat: Move debug prints from stdout to stderr This leaves the stdout cleaner, having only counter data. It makes it easier for programs to parse the output of turbostat, for example selftests. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 11:34:38 +02:00
fprintf(stderr, "%s(msr%d, %s, %s, width%d, scope%d, type%d, format%d, flags%x, id%d)\n",
__func__, msr_num, path, name, width, scope, type, format, flags, id);
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
switch (scope) {
case SCOPE_CPU:
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
msrp = find_msrp_by_name(sys.tp, name);
if (msrp) {
if (debug)
tools/power turbostat: Move debug prints from stdout to stderr This leaves the stdout cleaner, having only counter data. It makes it easier for programs to parse the output of turbostat, for example selftests. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 11:34:38 +02:00
fprintf(stderr, "%s: %s FOUND\n", __func__, name);
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
break;
}
if (sys.added_thread_counters++ >= MAX_ADDED_THREAD_COUNTERS) {
warnx("ignoring thread counter %s", name);
return -1;
tools/power turbostat: fix bugs in --add option When --add was used more than once, overflowed buffers caused some counters to be stored on top of others, corrupting the results. Simplify the code by simply reserving space for up to 16 added counters per each cpu, core, package. Per-cpu added counters were being printed only per-core. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 22:22:13 -05:00
}
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
break;
case SCOPE_CORE:
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
msrp = find_msrp_by_name(sys.cp, name);
if (msrp) {
if (debug)
tools/power turbostat: Move debug prints from stdout to stderr This leaves the stdout cleaner, having only counter data. It makes it easier for programs to parse the output of turbostat, for example selftests. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 11:34:38 +02:00
fprintf(stderr, "%s: %s FOUND\n", __func__, name);
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
break;
}
if (sys.added_core_counters++ >= MAX_ADDED_CORE_COUNTERS) {
warnx("ignoring core counter %s", name);
return -1;
tools/power turbostat: fix bugs in --add option When --add was used more than once, overflowed buffers caused some counters to be stored on top of others, corrupting the results. Simplify the code by simply reserving space for up to 16 added counters per each cpu, core, package. Per-cpu added counters were being printed only per-core. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 22:22:13 -05:00
}
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
break;
case SCOPE_PACKAGE:
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
msrp = find_msrp_by_name(sys.pp, name);
if (msrp) {
if (debug)
tools/power turbostat: Move debug prints from stdout to stderr This leaves the stdout cleaner, having only counter data. It makes it easier for programs to parse the output of turbostat, for example selftests. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 11:34:38 +02:00
fprintf(stderr, "%s: %s FOUND\n", __func__, name);
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
break;
}
if (sys.added_package_counters++ >= MAX_ADDED_PACKAGE_COUNTERS) {
warnx("ignoring package counter %s", name);
return -1;
tools/power turbostat: fix bugs in --add option When --add was used more than once, overflowed buffers caused some counters to be stored on top of others, corrupting the results. Simplify the code by simply reserving space for up to 16 added counters per each cpu, core, package. Per-cpu added counters were being printed only per-core. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 22:22:13 -05:00
}
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
break;
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
default:
warnx("ignoring counter %s with unknown scope", name);
return -1;
}
if (msrp == NULL) {
msrp = calloc(1, sizeof(struct msr_counter));
if (msrp == NULL)
err(-1, "calloc msr_counter");
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
msrp->msr_num = msr_num;
strncpy(msrp->name, name, NAME_BYTES - 1);
msrp->width = width;
msrp->type = type;
msrp->format = format;
msrp->flags = flags;
switch (scope) {
case SCOPE_CPU:
msrp->next = sys.tp;
sys.tp = msrp;
break;
case SCOPE_CORE:
msrp->next = sys.cp;
sys.cp = msrp;
break;
case SCOPE_PACKAGE:
msrp->next = sys.pp;
sys.pp = msrp;
break;
}
}
if (path) {
struct sysfs_path *sp;
sp = calloc(1, sizeof(struct sysfs_path));
if (sp == NULL) {
perror("calloc");
exit(1);
}
strncpy(sp->path, path, PATH_BYTES - 1);
sp->id = id;
sp->next = msrp->sp;
msrp->sp = sp;
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
return 0;
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
/*
* Initialize the fields used for identifying and opening the counter.
*
* Defer the initialization of any runtime buffers for actually reading
* the counters for when we initialize all perf counters, so we can later
* easily call re_initialize().
*/
struct perf_counter_info *make_perf_counter_info(const char *perf_device,
const char *perf_event,
const char *name,
unsigned int width,
enum counter_scope scope,
enum counter_type type, enum counter_format format)
{
struct perf_counter_info *pinfo;
pinfo = calloc(1, sizeof(*pinfo));
if (!pinfo)
errx(1, "%s: Failed to allocate %s/%s\n", __func__, perf_device, perf_event);
strncpy(pinfo->device, perf_device, ARRAY_SIZE(pinfo->device) - 1);
strncpy(pinfo->event, perf_event, ARRAY_SIZE(pinfo->event) - 1);
strncpy(pinfo->name, name, ARRAY_SIZE(pinfo->name) - 1);
pinfo->width = width;
pinfo->scope = scope;
pinfo->type = type;
pinfo->format = format;
return pinfo;
}
int add_perf_counter(const char *perf_device, const char *perf_event, const char *name_buffer, unsigned int width,
enum counter_scope scope, enum counter_type type, enum counter_format format)
{
struct perf_counter_info *pinfo;
switch (scope) {
case SCOPE_CPU:
if (sys.added_thread_perf_counters >= MAX_ADDED_THREAD_COUNTERS) {
warnx("ignoring thread counter perf/%s/%s", perf_device, perf_event);
return -1;
}
break;
case SCOPE_CORE:
if (sys.added_core_perf_counters >= MAX_ADDED_CORE_COUNTERS) {
warnx("ignoring core counter perf/%s/%s", perf_device, perf_event);
return -1;
}
break;
case SCOPE_PACKAGE:
if (sys.added_package_perf_counters >= MAX_ADDED_PACKAGE_COUNTERS) {
warnx("ignoring package counter perf/%s/%s", perf_device, perf_event);
return -1;
}
break;
}
pinfo = make_perf_counter_info(perf_device, perf_event, name_buffer, width, scope, type, format);
if (!pinfo)
return -1;
switch (scope) {
case SCOPE_CPU:
pinfo->next = sys.perf_tp;
sys.perf_tp = pinfo;
++sys.added_thread_perf_counters;
break;
case SCOPE_CORE:
pinfo->next = sys.perf_cp;
sys.perf_cp = pinfo;
++sys.added_core_perf_counters;
break;
case SCOPE_PACKAGE:
pinfo->next = sys.perf_pp;
sys.perf_pp = pinfo;
++sys.added_package_perf_counters;
break;
}
// FIXME: we might not have debug here yet
if (debug)
tools/power turbostat: Move debug prints from stdout to stderr This leaves the stdout cleaner, having only counter data. It makes it easier for programs to parse the output of turbostat, for example selftests. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-10 11:34:38 +02:00
fprintf(stderr, "%s: %s/%s, name: %s, scope%d\n",
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
__func__, pinfo->device, pinfo->event, pinfo->name, pinfo->scope);
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
return 0;
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
void parse_add_command_msr(char *add_command)
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
{
int msr_num = 0;
tools/power turbostat: extend --add option to accept /sys path Previously, the --add option could specify only an MSR. Here is is extended so an arbitrary /sys attribute, as specified by an absolute file path name. sudo ./turbostat --add /sys/devices/system/cpu/cpu0/cpuidle/state5/usage Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-08 02:41:51 -05:00
char *path = NULL;
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
char perf_device[PERF_DEV_NAME_BYTES] = "";
char perf_event[PERF_EVT_NAME_BYTES] = "";
char name_buffer[PERF_NAME_BYTES] = "";
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
int width = 64;
int fail = 0;
enum counter_scope scope = SCOPE_CPU;
enum counter_type type = COUNTER_CYCLES;
enum counter_format format = FORMAT_DELTA;
while (add_command) {
if (sscanf(add_command, "msr0x%x", &msr_num) == 1)
goto next;
if (sscanf(add_command, "msr%d", &msr_num) == 1)
goto next;
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
BUILD_BUG_ON(ARRAY_SIZE(perf_device) <= 31);
BUILD_BUG_ON(ARRAY_SIZE(perf_event) <= 31);
if (sscanf(add_command, "perf/%31[^/]/%31[^,]", &perf_device[0], &perf_event[0]) == 2)
goto next;
tools/power turbostat: extend --add option to accept /sys path Previously, the --add option could specify only an MSR. Here is is extended so an arbitrary /sys attribute, as specified by an absolute file path name. sudo ./turbostat --add /sys/devices/system/cpu/cpu0/cpuidle/state5/usage Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-08 02:41:51 -05:00
if (*add_command == '/') {
path = add_command;
goto next;
}
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
if (sscanf(add_command, "u%d", &width) == 1) {
if ((width == 32) || (width == 64))
goto next;
width = 64;
}
if (!strncmp(add_command, "cpu", strlen("cpu"))) {
scope = SCOPE_CPU;
goto next;
}
if (!strncmp(add_command, "core", strlen("core"))) {
scope = SCOPE_CORE;
goto next;
}
if (!strncmp(add_command, "package", strlen("package"))) {
scope = SCOPE_PACKAGE;
goto next;
}
if (!strncmp(add_command, "cycles", strlen("cycles"))) {
type = COUNTER_CYCLES;
goto next;
}
if (!strncmp(add_command, "seconds", strlen("seconds"))) {
type = COUNTER_SECONDS;
goto next;
}
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
if (!strncmp(add_command, "usec", strlen("usec"))) {
type = COUNTER_USEC;
goto next;
}
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
if (!strncmp(add_command, "raw", strlen("raw"))) {
format = FORMAT_RAW;
goto next;
}
if (!strncmp(add_command, "delta", strlen("delta"))) {
format = FORMAT_DELTA;
goto next;
}
if (!strncmp(add_command, "percent", strlen("percent"))) {
format = FORMAT_PERCENT;
goto next;
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
BUILD_BUG_ON(ARRAY_SIZE(name_buffer) <= 18);
if (sscanf(add_command, "%18s,%*s", name_buffer) == 1) {
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
char *eos;
eos = strchr(name_buffer, ',');
if (eos)
*eos = '\0';
goto next;
}
next:
add_command = strchr(add_command, ',');
tools/power turbostat: extend --add option to accept /sys path Previously, the --add option could specify only an MSR. Here is is extended so an arbitrary /sys attribute, as specified by an absolute file path name. sudo ./turbostat --add /sys/devices/system/cpu/cpu0/cpuidle/state5/usage Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-08 02:41:51 -05:00
if (add_command) {
*add_command = '\0';
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
add_command++;
tools/power turbostat: extend --add option to accept /sys path Previously, the --add option could specify only an MSR. Here is is extended so an arbitrary /sys attribute, as specified by an absolute file path name. sudo ./turbostat --add /sys/devices/system/cpu/cpu0/cpuidle/state5/usage Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-08 02:41:51 -05:00
}
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
if ((msr_num == 0) && (path == NULL) && (perf_device[0] == '\0' || perf_event[0] == '\0')) {
fprintf(stderr, "--add: (msrDDD | msr0xXXX | /path_to_counter | perf/device/event ) required\n");
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
fail++;
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
/* Test for non-empty perf_device and perf_event */
const bool is_perf_counter = perf_device[0] && perf_event[0];
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
/* generate default column header */
if (*name_buffer == '\0') {
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
if (is_perf_counter) {
snprintf(name_buffer, ARRAY_SIZE(name_buffer), "perf/%s", perf_event);
} else {
if (width == 32)
sprintf(name_buffer, "M0x%x%s", msr_num, format == FORMAT_PERCENT ? "%" : "");
else
sprintf(name_buffer, "M0X%x%s", msr_num, format == FORMAT_PERCENT ? "%" : "");
}
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
}
tools/power turbostat: Extend --add option with perf counters User can now read perf counters using "--add perf/<device>/<event>". Other details work similarly to how --add works with MSRs. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-05-28 15:46:10 +02:00
if (is_perf_counter) {
if (add_perf_counter(perf_device, perf_event, name_buffer, width, scope, type, format))
fail++;
} else {
if (add_counter(msr_num, path, name_buffer, width, scope, type, format, 0, 0))
fail++;
}
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
if (fail) {
help();
exit(1);
}
}
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
bool starts_with(const char *str, const char *prefix)
{
return strncmp(prefix, str, strlen(prefix)) == 0;
}
void parse_add_command_pmt(char *add_command)
{
char *name = NULL;
char *type_name = NULL;
char *format_name = NULL;
unsigned int offset;
unsigned int lsb;
unsigned int msb;
unsigned int guid;
unsigned int domain_id;
enum counter_scope scope = 0;
enum pmt_datatype type = PMT_TYPE_RAW;
enum counter_format format = FORMAT_RAW;
bool has_offset = false;
bool has_lsb = false;
bool has_msb = false;
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
bool has_format = true; /* Format has a default value. */
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
bool has_guid = false;
bool has_scope = false;
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
bool has_type = true; /* Type has a default value. */
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
/* Consume the "pmt," prefix. */
add_command = strchr(add_command, ',');
if (!add_command) {
help();
exit(1);
}
++add_command;
while (add_command) {
if (starts_with(add_command, "name=")) {
name = add_command + strlen("name=");
goto next;
}
if (starts_with(add_command, "type=")) {
type_name = add_command + strlen("type=");
goto next;
}
if (starts_with(add_command, "domain=")) {
const size_t prefix_len = strlen("domain=");
if (sscanf(add_command + prefix_len, "cpu%u", &domain_id) == 1) {
scope = SCOPE_CPU;
has_scope = true;
} else if (sscanf(add_command + prefix_len, "core%u", &domain_id) == 1) {
scope = SCOPE_CORE;
has_scope = true;
} else if (sscanf(add_command + prefix_len, "package%u", &domain_id) == 1) {
scope = SCOPE_PACKAGE;
has_scope = true;
}
if (!has_scope) {
printf("%s: invalid value for scope. Expected cpu%%u, core%%u or package%%u.\n",
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
__func__);
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
exit(1);
}
goto next;
}
if (starts_with(add_command, "format=")) {
format_name = add_command + strlen("format=");
goto next;
}
if (sscanf(add_command, "offset=%u", &offset) == 1) {
has_offset = true;
goto next;
}
if (sscanf(add_command, "lsb=%u", &lsb) == 1) {
has_lsb = true;
goto next;
}
if (sscanf(add_command, "msb=%u", &msb) == 1) {
has_msb = true;
goto next;
}
if (sscanf(add_command, "guid=%x", &guid) == 1) {
has_guid = true;
goto next;
}
next:
add_command = strchr(add_command, ',');
if (add_command) {
*add_command = '\0';
add_command++;
}
}
if (!name) {
printf("%s: missing %s\n", __func__, "name");
exit(1);
}
if (strlen(name) >= PMT_COUNTER_NAME_SIZE_BYTES) {
tools/power turbostat: version 2024.07.26 Release 2024.07.26: Enable turbostat extensions to add both perf and PMT (Intel Platform Monitoring Technology) counters from the cmdline. Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter. This commit: Clean up white-space nits introduced since version 2024.05.10 Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-26 14:16:28 -04:00
printf("%s: name has to be at most %d characters long\n", __func__, PMT_COUNTER_NAME_SIZE_BYTES);
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
exit(1);
}
if (format_name) {
has_format = false;
if (strcmp("raw", format_name) == 0) {
format = FORMAT_RAW;
has_format = true;
}
if (strcmp("delta", format_name) == 0) {
format = FORMAT_DELTA;
has_format = true;
}
if (!has_format) {
fprintf(stderr, "%s: Invalid format %s. Expected raw or delta\n", __func__, format_name);
exit(1);
}
}
if (type_name) {
has_type = false;
if (strcmp("raw", type_name) == 0) {
type = PMT_TYPE_RAW;
has_type = true;
}
tools/power turbostat: Add MTL's PMT DC6 builtin counter Provide a definition for metadata that allows reading DC6 residency counter via PMT and exposes it as a builtin counter. Note that this residency counter is updated and read via entirely different mechanisms vs the MSR-based residency counters. On MTL processors, there are times when Die%c6 will report above 100%. This is still useful, but don't expect 3 digits of precision... Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:12:22 +02:00
if (strcmp("txtal_time", type_name) == 0) {
type = PMT_TYPE_XTAL_TIME;
has_type = true;
}
tools/power turbostat: Add early support for PMT counters Allows users to read Intel PMT (Platform Monitoring Technology) counters, providing interface similar to one used to add MSR and perf counters. Because PMT is exposed as a raw MMIO range, without metadata, user has to supply the necessary information to find and correctly display the requested counter. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-07-22 22:11:03 +02:00
if (!has_type) {
printf("%s: invalid %s: %s\n", __func__, "type", type_name);
exit(1);
}
}
if (!has_offset) {
printf("%s : missing %s\n", __func__, "offset");
exit(1);
}
if (!has_lsb) {
printf("%s: missing %s\n", __func__, "lsb");
exit(1);
}
if (!has_msb) {
printf("%s: missing %s\n", __func__, "msb");
exit(1);
}
if (!has_guid) {
printf("%s: missing %s\n", __func__, "guid");
exit(1);
}
if (!has_scope) {
printf("%s: missing %s\n", __func__, "scope");
exit(1);
}
if (lsb > msb) {
printf("%s: lsb > msb doesn't make sense\n", __func__);
exit(1);
}
pmt_add_counter(guid, name, type, lsb, msb, offset, scope, format, domain_id, PMT_OPEN_REQUIRED);
}
void parse_add_command(char *add_command)
{
if (strncmp(add_command, "pmt", strlen("pmt")) == 0)
return parse_add_command_pmt(add_command);
return parse_add_command_msr(add_command);
}
tools/power turbostat: Allow -e for all names. Currently, there are a number of variables which are displayed by default, enabled with -e all, and listed by --list, but which you can not give to --enable/-e. So you can enable CPU0c1 (in the bic array), but you can't enable C1 or C1% (not in the bic array, but exists in sysfs). This runs counter to both the documentation and user expectations, and it's just not very user friendly. As such, the mechanism used by --hide has been duplicated, and is now also used by --enable, so we can handle unknown names gracefully. Note: One impact of this is that truly unknown fields given to --enable will no longer generate errors, they will be silently ignored, as --hide does. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:38 -07:00
int is_deferred_add(char *name)
{
int i;
for (i = 0; i < deferred_add_index; ++i)
if (!strcmp(name, deferred_add_names[i]))
return 1;
return 0;
}
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
int is_deferred_skip(char *name)
{
int i;
for (i = 0; i < deferred_skip_index; ++i)
if (!strcmp(name, deferred_skip_names[i]))
return 1;
return 0;
}
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
void probe_sysfs(void)
{
char path[64];
char name_buf[16];
FILE *input;
int state;
char *sp;
tools/power turbostat: add POLL and POLL% column Like the "C1" and "C1%" column, the new POLL and POLL% columns show invocations and residency% during the measurement interval. While it didn't seem important to track in the past, we've recently found some Linux cpuidle bugs related to POLL%. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:38:29 -04:00
for (state = 10; state >= 0; --state) {
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
sprintf(path, "/sys/devices/system/cpu/cpu%d/cpuidle/state%d/name", base_cpu, state);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
input = fopen(path, "r");
if (input == NULL)
continue;
tools/power turbostat: Add checks for failure of fgets() and fscanf() Most calls to fgets() and fscanf() are followed by error checks. Add an exit-on-error in the remaining cases. Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-20 23:01:03 -04:00
if (!fgets(name_buf, sizeof(name_buf), input))
err(1, "%s: failed to read file", path);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
/* truncate "C1-HSW\n" to "C1", or truncate "C1\n" to "C1" */
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
sp = strchr(name_buf, '-');
if (!sp)
sp = strchrnul(name_buf, '\n');
*sp = '%';
*(sp + 1) = '\0';
tools/power turbostat: Fix output formatting for ACPI CST enumeration turbostat formatting is broken with ACPI CST for enumeration. The problem is that the CX_ACPI% is eight characters long which does not work with tab formatting. One simple solution is to remove the underbar from the state name such that C1_ACPI will be displayed as C1ACPI. Signed-off-by: David Arcari <darcari@redhat.com> Cc: Len Brown <lenb@kernel.org> Cc: linux-kernel@vger.kernel.org Signed-off-by: Len Brown <len.brown@intel.com>
2020-08-10 10:43:30 -04:00
remove_underbar(name_buf);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
fclose(input);
sprintf(path, "cpuidle/state%d/time", state);
tools/power turbostat: Allow -e for all names. Currently, there are a number of variables which are displayed by default, enabled with -e all, and listed by --list, but which you can not give to --enable/-e. So you can enable CPU0c1 (in the bic array), but you can't enable C1 or C1% (not in the bic array, but exists in sysfs). This runs counter to both the documentation and user expectations, and it's just not very user friendly. As such, the mechanism used by --hide has been duplicated, and is now also used by --enable, so we can handle unknown names gracefully. Note: One impact of this is that truly unknown fields given to --enable will no longer generate errors, they will be silently ignored, as --hide does. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:38 -07:00
if (!DO_BIC(BIC_sysfs) && !is_deferred_add(name_buf))
continue;
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
if (is_deferred_skip(name_buf))
continue;
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
add_counter(0, path, name_buf, 64, SCOPE_CPU, COUNTER_USEC, FORMAT_PERCENT, SYSFS_PERCPU, 0);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
}
tools/power turbostat: add POLL and POLL% column Like the "C1" and "C1%" column, the new POLL and POLL% columns show invocations and residency% during the measurement interval. While it didn't seem important to track in the past, we've recently found some Linux cpuidle bugs related to POLL%. Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-01 12:38:29 -04:00
for (state = 10; state >= 0; --state) {
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
sprintf(path, "/sys/devices/system/cpu/cpu%d/cpuidle/state%d/name", base_cpu, state);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
input = fopen(path, "r");
if (input == NULL)
continue;
tools/power turbostat: Add checks for failure of fgets() and fscanf() Most calls to fgets() and fscanf() are followed by error checks. Add an exit-on-error in the remaining cases. Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Len Brown <len.brown@intel.com>
2019-03-20 23:01:03 -04:00
if (!fgets(name_buf, sizeof(name_buf), input))
err(1, "%s: failed to read file", path);
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
/* truncate "C1-HSW\n" to "C1", or truncate "C1\n" to "C1" */
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
sp = strchr(name_buf, '-');
if (!sp)
sp = strchrnul(name_buf, '\n');
*sp = '\0';
fclose(input);
tools/power turbostat: Fix output formatting for ACPI CST enumeration turbostat formatting is broken with ACPI CST for enumeration. The problem is that the CX_ACPI% is eight characters long which does not work with tab formatting. One simple solution is to remove the underbar from the state name such that C1_ACPI will be displayed as C1ACPI. Signed-off-by: David Arcari <darcari@redhat.com> Cc: Len Brown <lenb@kernel.org> Cc: linux-kernel@vger.kernel.org Signed-off-by: Len Brown <len.brown@intel.com>
2020-08-10 10:43:30 -04:00
remove_underbar(name_buf);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
sprintf(path, "cpuidle/state%d/usage", state);
tools/power turbostat: Allow -e for all names. Currently, there are a number of variables which are displayed by default, enabled with -e all, and listed by --list, but which you can not give to --enable/-e. So you can enable CPU0c1 (in the bic array), but you can't enable C1 or C1% (not in the bic array, but exists in sysfs). This runs counter to both the documentation and user expectations, and it's just not very user friendly. As such, the mechanism used by --hide has been duplicated, and is now also used by --enable, so we can handle unknown names gracefully. Note: One impact of this is that truly unknown fields given to --enable will no longer generate errors, they will be silently ignored, as --hide does. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:38 -07:00
if (!DO_BIC(BIC_sysfs) && !is_deferred_add(name_buf))
continue;
tools/power turbostat: support "--hide C1" etc. Originally, the only way to hide the sysfs C-state statistics columns was with "--hide sysfs". This was because we process "--hide" before we probe for those columns. hack --hide to remember deferred hide requests, and apply them when sysfs is probed. "--hide sysfs" is still available as short-hand to refer to the entire group of counters. The down-side of this change is that we no longer error check for bogus --hide column names. But the user will quickly figure that out if a column they mean to hide is still there... Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 23:21:13 -05:00
if (is_deferred_skip(name_buf))
continue;
tools/power turbostat: Add columns for clustered uncore frequency New machines have multiple uncore frequencies per package, visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/ turbostat now samples these frequencies each measurement interval. For each package, turbostat now prints "UMHzX.Y" columns, where X = domain_id, and Y = fabric_cluster_id. The system summary for each UMHzX.Y column is the average value for across all of the packages in the system. Signed-off-by: Len Brown <len.brown@intel.com>
2024-04-24 21:12:18 -04:00
add_counter(0, path, name_buf, 64, SCOPE_CPU, COUNTER_ITEMS, FORMAT_DELTA, SYSFS_PERCPU, 0);
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
}
}
tools/power turbostat: add --cpu parameter With the --cpu parameter, turbostat prints only lines for the specified set of CPUs: sudo ./turbostat --quiet --show Core,CPU --cpu 0,1,3..5,6-7 Core CPU - - 0 0 0 4 1 1 1 5 2 6 3 3 3 7 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 23:54:15 -05:00
/*
* parse cpuset with following syntax
* 1,2,4..6,8-10 and set bits in cpu_subset
*/
void parse_cpu_command(char *optarg)
{
tools/power turbostat: move --Package and --processor into the --cpu option --Package is now "--cpu package", which will display just the 1st CPU in each package --processor is not "--cpu core" which will display just the 1st CPU in each core Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 22:33:42 -05:00
if (!strcmp(optarg, "core")) {
if (cpu_subset)
goto error;
show_core_only++;
return;
}
if (!strcmp(optarg, "package")) {
if (cpu_subset)
goto error;
show_pkg_only++;
return;
}
if (show_core_only || show_pkg_only)
goto error;
tools/power turbostat: add --cpu parameter With the --cpu parameter, turbostat prints only lines for the specified set of CPUs: sudo ./turbostat --quiet --show Core,CPU --cpu 0,1,3..5,6-7 Core CPU - - 0 0 0 4 1 1 1 5 2 6 3 3 3 7 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 23:54:15 -05:00
cpu_subset = CPU_ALLOC(CPU_SUBSET_MAXCPUS);
if (cpu_subset == NULL)
err(3, "CPU_ALLOC");
cpu_subset_size = CPU_ALLOC_SIZE(CPU_SUBSET_MAXCPUS);
CPU_ZERO_S(cpu_subset_size, cpu_subset);
tools/power/turbostat: Abstrct function for parsing cpu string Abstract parse_cpu_str() which can update any specified cpu_set by a given cpu string. This can be used to handle further CPU limitations from other sources like cgroup. The cpu string parsing code is also enhanced to handle the strings that have an extra '\n' before string terminator. Signed-off-by: Zhang Rui <rui.zhang@intel.com>
2023-10-20 09:45:21 +08:00
if (parse_cpu_str(optarg, cpu_subset, cpu_subset_size))
goto error;
tools/power turbostat: add --cpu parameter With the --cpu parameter, turbostat prints only lines for the specified set of CPUs: sudo ./turbostat --quiet --show Core,CPU --cpu 0,1,3..5,6-7 Core CPU - - 0 0 0 4 1 1 1 5 2 6 3 3 3 7 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 23:54:15 -05:00
return;
error:
tools/power turbostat: move --Package and --processor into the --cpu option --Package is now "--cpu package", which will display just the 1st CPU in each package --processor is not "--cpu core" which will display just the 1st CPU in each core Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-21 22:33:42 -05:00
fprintf(stderr, "\"--cpu %s\" malformed\n", optarg);
help();
tools/power turbostat: add --cpu parameter With the --cpu parameter, turbostat prints only lines for the specified set of CPUs: sudo ./turbostat --quiet --show Core,CPU --cpu 0,1,3..5,6-7 Core CPU - - 0 0 0 4 1 1 1 5 2 6 3 3 3 7 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 23:54:15 -05:00
exit(-1);
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
void cmdline(int argc, char **argv)
{
int opt;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
int option_index = 0;
static struct option long_options[] = {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
{ "add", required_argument, 0, 'a' },
{ "cpu", required_argument, 0, 'c' },
{ "Dump", no_argument, 0, 'D' },
{ "debug", no_argument, 0, 'd' }, /* internal, not documented */
{ "enable", required_argument, 0, 'e' },
{ "interval", required_argument, 0, 'i' },
{ "IPC", no_argument, 0, 'I' },
{ "num_iterations", required_argument, 0, 'n' },
tools/power turbostat: Allow printing header every N iterations This gives the ability to reprint the header every N iterations, so you can ensure that a scrolling display always has the header visible somewhere on the screen. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:39 -07:00
{ "header_iterations", required_argument, 0, 'N' },
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
{ "help", no_argument, 0, 'h' },
{ "hide", required_argument, 0, 'H' }, // meh, -h taken by --help
{ "Joules", no_argument, 0, 'J' },
{ "list", no_argument, 0, 'l' },
{ "out", required_argument, 0, 'o' },
{ "quiet", no_argument, 0, 'q' },
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
{ "no-msr", no_argument, 0, 'M' },
tools/power turbostat: Add --no-perf option Add the --no-perf option to allow users to run turbostat without accessing perf. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:58:02 +01:00
{ "no-perf", no_argument, 0, 'P' },
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
{ "show", required_argument, 0, 's' },
{ "Summary", no_argument, 0, 'S' },
{ "TCC", required_argument, 0, 'T' },
{ "version", no_argument, 0, 'v' },
{ 0, 0, 0, 0 }
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
};
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
progname = argv[0];
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
/*
* Parse some options early, because they may make other options invalid,
* like adding the MSR counter with --add and at the same time using --no-msr.
*/
tools/power turbostat: Fix child's argument forwarding Add '+' to optstring when early scanning for --no-msr and --no-perf. It causes option processing to stop as soon as a nonoption argument is encountered, effectively skipping child's arguments. Fixes: 3e4048466c39 ("tools/power turbostat: Add --no-msr option") Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-11-13 15:48:22 +01:00
while ((opt = getopt_long_only(argc, argv, "+MPn:", long_options, &option_index)) != -1) {
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
switch (opt) {
case 'M':
no_msr = 1;
break;
tools/power turbostat: Add --no-perf option Add the --no-perf option to allow users to run turbostat without accessing perf. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:58:02 +01:00
case 'P':
no_perf = 1;
break;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
default:
break;
}
}
optind = 0;
while ((opt = getopt_long_only(argc, argv, "+C:c:Dde:hi:Jn:o:qMST:v", long_options, &option_index)) != -1) {
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
switch (opt) {
tools/power turbostat: Make extensible via the --add parameter Create the "--add" parameter. This can be used to teach an existing turbostat binary about any number of any type of counter. turbostat(8) details the syntax for --add. Signed-off-by: Len Brown <len.brown@intel.com>
2016-12-22 23:57:55 -05:00
case 'a':
parse_add_command(optarg);
break;
tools/power turbostat: add --cpu parameter With the --cpu parameter, turbostat prints only lines for the specified set of CPUs: sudo ./turbostat --quiet --show Core,CPU --cpu 0,1,3..5,6-7 Core CPU - - 0 0 0 4 1 1 1 5 2 6 3 3 3 7 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 23:54:15 -05:00
case 'c':
parse_cpu_command(optarg);
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
case 'D':
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
dump_only++;
tools/power turbostat: Force --no-perf in --dump mode Force the --no-perf early to prevent using it as a source. User asks for raw values, but perf returns them relative to the opening of the file descriptor. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-10-24 15:17:45 +02:00
/*
* Force the no_perf early to prevent using it as a source.
* User asks for raw values, but perf returns them relative
* to the opening of the file descriptor.
*/
no_perf = 1;
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
break;
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
case 'e':
/* --enable specified counter */
tools/power turbostat: add optional APIC X2APIC columns Add APIC and X2APIC columns to the topology section. They are disabled-by-default -- enable like so: --debug or --enable APIC,X2APIC Signed-off-by: Len Brown <len.brown@intel.com>
2018-06-06 17:44:48 -04:00
bic_enabled = bic_enabled | bic_lookup(optarg, SHOW_LIST);
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
case 'd':
debug++;
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
ENABLE_BIC(BIC_DISABLED_BY_DEFAULT);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
break;
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
case 'H':
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
/*
* --hide: do not show those specified
* multiple invocations simply clear more bits in enabled mask
*/
bic_enabled &= ~bic_lookup(optarg, HIDE_LIST);
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
case 'h':
default:
help();
exit(1);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
case 'i':
tools/power turbostat: allow sub-sec intervals turbostat -i interval_sec will sample and display statistics every interval_sec. interval_sec used to be a whole number of seconds, but now we accept a decimal, as small as 0.001 sec (1 ms). Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-12 22:44:48 -05:00
{
double interval = strtod(optarg, NULL);
if (interval < 0.001) {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "interval %f seconds is too small\n", interval);
tools/power turbostat: allow sub-sec intervals turbostat -i interval_sec will sample and display statistics every interval_sec. interval_sec used to be a whole number of seconds, but now we accept a decimal, as small as 0.001 sec (1 ms). Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-12 22:44:48 -05:00
exit(2);
}
tools/power turbostat: fix printing on input The recent patch that implements table printing on a keypress introduced a regression - turbostat prints the table almost continuously if it is run from a daemon program. The problem is also easy to reproduce like this: echo | turbostat The reason is that we cannot assume that stdin is always a TTY. It can be many things. This patch adds fixes the problem by limiting the new keypress functionality to TTYs only. If stdin is not a TTY, we just sleep for the full interval time. While on it, clean-up 'do_sleep()' to return no value, as callers do not expect that anyway. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-04 15:01:47 +03:00
interval_tv.tv_sec = interval_ts.tv_sec = interval;
tools/power turbostat: end current interval upon newline input In turbostat interval mode, a newline typed on standard input will now conclude the current interval. Data will immediately be collected and printed for that interval, and the next interval will be started. This is similar to the recently added SIGUSR1 feature. But that is for use by programs, while this is for interactive use. Signed-off-by: Len Brown <len.brown@intel.com>
2017-07-19 19:28:37 -04:00
interval_tv.tv_usec = (interval - interval_tv.tv_sec) * 1000000;
tools/power turbostat: fix printing on input The recent patch that implements table printing on a keypress introduced a regression - turbostat prints the table almost continuously if it is run from a daemon program. The problem is also easy to reproduce like this: echo | turbostat The reason is that we cannot assume that stdin is always a TTY. It can be many things. This patch adds fixes the problem by limiting the new keypress functionality to TTYs only. If stdin is not a TTY, we just sleep for the full interval time. While on it, clean-up 'do_sleep()' to return no value, as callers do not expect that anyway. Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-04 15:01:47 +03:00
interval_ts.tv_nsec = (interval - interval_ts.tv_sec) * 1000000000;
tools/power turbostat: allow sub-sec intervals turbostat -i interval_sec will sample and display statistics every interval_sec. interval_sec used to be a whole number of seconds, but now we accept a decimal, as small as 0.001 sec (1 ms). Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-12 22:44:48 -05:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
case 'J':
rapl_joules++;
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 01:25:08 -04:00
break;
tools/power turbostat: Add --list option to show available header names It is handy to know the list of column header names, so that they can be used with --add and --skip The new --list option shows them: sudo ./turbostat --list --hide sysfs ,Core,CPU,Avg_MHz,Busy%,Bzy_MHz,TSC_MHz,IRQ,SMI,CPU%c1,CPU%c3,CPU%c6,CPU%c7,CoreTmp,PkgTmp,GFX%rc6,GFXMHz,PkgWatt,CorWatt,GFXWatt Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 17:15:11 -05:00
case 'l':
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
ENABLE_BIC(BIC_DISABLED_BY_DEFAULT);
tools/power turbostat: Add --list option to show available header names It is handy to know the list of column header names, so that they can be used with --add and --skip The new --list option shows them: sudo ./turbostat --list --hide sysfs ,Core,CPU,Avg_MHz,Busy%,Bzy_MHz,TSC_MHz,IRQ,SMI,CPU%c1,CPU%c3,CPU%c6,CPU%c7,CoreTmp,PkgTmp,GFX%rc6,GFXMHz,PkgWatt,CorWatt,GFXWatt Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 17:15:11 -05:00
list_header_only++;
quiet++;
break;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
case 'o':
outf = fopen_or_die(optarg, "w");
break;
tools/power turbostat: print system config, unless --quiet Some users want turbostat to tell them everything, by default. Some users want turbostat to be quiet, by default. I find that I'm in the 1st camp, and so I've never liked needing to type the --debug parameter to decode the system configuration. So here we change the default and print the system configuration, by default. (The --debug option is now un-documented, though it does still exist for debugging turbostat internals) When you do not want to see the system configuration header, use the new "--quiet" option. Signed-off-by: Len Brown <len.brown@intel.com>
2017-01-21 02:26:00 -05:00
case 'q':
quiet = 1;
break;
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
case 'M':
tools/power turbostat: Add --no-perf option Add the --no-perf option to allow users to run turbostat without accessing perf. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:58:02 +01:00
case 'P':
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
/* Parsed earlier */
break;
tools/power turbostat: if --num_iterations, print for specific number of iterations There's a use case during test to only print specific round of iterations if --num_iterations is specified, for example, with this patch applied: turbostat -i 5 -n 4 will capture 4 samples with 5 seconds interval. [lenb: renamed to --num_iterations from --iterations] Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Reviewed-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-04-26 08:41:03 +08:00
case 'n':
num_iterations = strtod(optarg, NULL);
if (num_iterations <= 0) {
tools/power turbostat: formatting Spring is here... run a long overdue Lendent on turbostat.c no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-04 19:21:34 -04:00
fprintf(outf, "iterations %d should be positive number\n", num_iterations);
tools/power turbostat: if --num_iterations, print for specific number of iterations There's a use case during test to only print specific round of iterations if --num_iterations is specified, for example, with this patch applied: turbostat -i 5 -n 4 will capture 4 samples with 5 seconds interval. [lenb: renamed to --num_iterations from --iterations] Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Reviewed-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-04-26 08:41:03 +08:00
exit(2);
}
break;
tools/power turbostat: Allow printing header every N iterations This gives the ability to reprint the header every N iterations, so you can ensure that a scrolling display always has the header visible somewhere on the screen. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:39 -07:00
case 'N':
header_iterations = strtod(optarg, NULL);
if (header_iterations <= 0) {
tools/power turbostat: be more useful as non-root Don't exit if used this way: sudo setcap cap_sys_nice,cap_sys_rawio=+ep ./turbostat sudo chmod +r /dev/cpu/*/msr ./turbostat note: cap_sys_admin is now also needed for the perf IPC counter: sudo setcap cap_sys_admin,cap_sys_nice,cap_sys_rawio=+ep ./turbostat Reported-by: Artem S. Tashkinov <aros@gmx.com> Reported-by: Toby Broom <tbroom@outlook.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-16 22:25:45 -04:00
fprintf(outf, "iterations %d should be positive number\n", header_iterations);
tools/power turbostat: Allow printing header every N iterations This gives the ability to reprint the header every N iterations, so you can ensure that a scrolling display always has the header visible somewhere on the screen. Signed-off-by: Zephaniah E. Loss-Cutler-Hull <zephaniah@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-10-04 21:54:39 -07:00
exit(2);
tools/power turbostat: if --num_iterations, print for specific number of iterations There's a use case during test to only print specific round of iterations if --num_iterations is specified, for example, with this patch applied: turbostat -i 5 -n 4 will capture 4 samples with 5 seconds interval. [lenb: renamed to --num_iterations from --iterations] Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Reviewed-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2018-04-26 08:41:03 +08:00
}
break;
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
case 's':
tools/power turbostat: add --enable Time_Of_Day_Seconds Add a Time_Of_Day_Seconds column showing when measurement for each row was completed. Units are [sec.subsec] since Epoch, as reported by gettimeofday(2). While useful to correlate turbostat output with other tools, this built-in column is disabled, by default. Add the "--enable" option to enable such disabled-by-default built-in columns: "--enable Time_Of_Day_Seconds" "--enable usec" "--enable all", will enable all disabled-by-defauilt built-in counters. When "--debug" is used, all disabled-by-default columns are enabled, unless explicitly skipped using "--hide" Signed-off-by: Len Brown <len.brown@intel.com>
2017-10-17 15:42:56 -04:00
/*
* --show: show only those specified
* The 1st invocation will clear and replace the enabled mask
* subsequent invocations can add to it.
*/
if (shown == 0)
bic_enabled = bic_lookup(optarg, SHOW_LIST);
else
bic_enabled |= bic_lookup(optarg, SHOW_LIST);
shown = 1;
tools/power turbostat: Add --show and --hide parameters Add the "--show" and "--hide" cmdline parameters. By default, turbostat shows all columns. turbostat --hide counter_list will continue showing all columns, except for those listed. turbostat --show counter_list will show _only_ the listed columns These features work for built-in counters, and have no effect on columns added with the --add parameter. Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-10 00:25:41 -05:00
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
case 'S':
summary_only++;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
break;
case 'T':
tools/power turbostat: rename tcc variables There are two TCC activation temeprature. One is the default TCC activation temperature, also known as TJ_MAX. Another one is the effective TCC activation temperature, which is the subtraction of default TCC activation temperature and TCC offset. The name of variable tcc_activation_temp might be misleading here. Thus rename tcc_activation_temp to tj_max, and use tcc_default and tcc_offset to calculate the effective TCC activation temperature. No functional change in this patch. Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-04-26 18:49:26 +08:00
tj_max_override = atoi(optarg);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 00:48:57 -05:00
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
case 'v':
print_version();
exit(0);
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-16 10:23:41 -08:00
break;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
}
}
}
tools/power turbostat: Increase the limit for fd opened When running turbostat, a system with 512 cpus reaches the limit for maximum number of file descriptors that can be opened. To solve this problem, the limit is raised to 2^15, which is a large enough number. Below data is collected from AMD server systems while running turbostat: |-----------+-------------------------------| | # of cpus | # of opened fds for turbostat | |-----------+-------------------------------| | 128 | 260 | |-----------+-------------------------------| | 192 | 388 | |-----------+-------------------------------| | 512 | 1028 | |-----------+-------------------------------| So, the new max limit would be sufficient up to 2^14 cpus (but this also depends on how many counters are enabled). Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Wyes Karny <wyes.karny@amd.com> Signed-off-by: Len Brown <len.brown@intel.com>
2023-10-03 05:07:51 +00:00
void set_rlimit(void)
{
struct rlimit limit;
if (getrlimit(RLIMIT_NOFILE, &limit) < 0)
err(1, "Failed to get rlimit");
if (limit.rlim_max < MAX_NOFILE)
limit.rlim_max = MAX_NOFILE;
if (limit.rlim_cur < MAX_NOFILE)
limit.rlim_cur = MAX_NOFILE;
if (setrlimit(RLIMIT_NOFILE, &limit) < 0)
err(1, "Failed to set rlimit");
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
int main(int argc, char **argv)
{
tools/power/turbostat: Move process to root cgroup When available CPUs are reduced via cgroup cpuset controller, turbostat will exit with errors (For example): get_counters: Could not migrate to CPU 0 turbostat: re-initialized with num_cpus 20 get_counters: Could not migrate to CPU 0 turbostat: re-initialized with num_cpus 20 Move the turbostat to root cgroup, which has every CPU. Writing the value 0 to a cgroup.procs file causes the writing process to be moved to the corresponding cgroup. Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Tested-by: Zhang Rui <rui.zhang@intel.com>
2023-01-24 10:39:53 -08:00
int fd, ret;
fd = open("/sys/fs/cgroup/cgroup.procs", O_WRONLY);
if (fd < 0)
goto skip_cgroup_setting;
ret = write(fd, "0\n", 2);
if (ret == -1)
perror("Can't update cgroup\n");
close(fd);
skip_cgroup_setting:
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 23:36:17 -05:00
outf = stderr;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
cmdline(argc, argv);
tools/power turbostat: print the kernel boot commandline It would be handy to have cmdline in turbostat output. For example, according to the turbostat output, there are no C-states requested. In this case the user is very curious if something like intel_idle.max_cstate=0 was used, or may be idle=none too. It is also curious whether things like intel_pstate=nohwp were used. Print the boot command line accordingly: turbostat version 21.05.04 - Len Brown <lenb@kernel.org> Kernel command line: BOOT_IMAGE=/boot/vmlinuz-5.16.0+ root=UUID= b42359ed-1e05-42eb-8757-6bf2a1c19070 ro quiet splash vt.handoff=7 Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-21 22:38:34 +08:00
if (!quiet) {
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 01:56:38 -05:00
print_version();
tools/power turbostat: print the kernel boot commandline It would be handy to have cmdline in turbostat output. For example, according to the turbostat output, there are no C-states requested. In this case the user is very curious if something like intel_idle.max_cstate=0 was used, or may be idle=none too. It is also curious whether things like intel_pstate=nohwp were used. Print the boot command line accordingly: turbostat version 21.05.04 - Len Brown <lenb@kernel.org> Kernel command line: BOOT_IMAGE=/boot/vmlinuz-5.16.0+ root=UUID= b42359ed-1e05-42eb-8757-6bf2a1c19070 ro quiet splash vt.handoff=7 Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2022-04-21 22:38:34 +08:00
print_bootcmd();
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
tools/power turbostat: print sysfs C-state stats When turbostat shows % of time in a CPU idle power state, it has always been showing information from underlying hardware residency counters. While this reflects what the hardware is doing, and is thus useful for understanding the hardware, it doesn't directly tell us what Linux requested -- which is useful for tuning Linux itself. Here we add columns to turbostat to show the Linux cpuidle sub-system statistics: /sys/devices/system/cpu/cpu*/cpuidle/state*/* The first group of columns are the "usage", which is the number of times software requested that C-state in the measurement interval. eg C1 below. The second group of columns are the "time", which is the percentage of the measurement interval time that software has requested the specified C-state. eg C1% below. These software counters can be compared to the underlying hardware residency counters (eg CPU%c1 CPU%c3 CPU%c6 CPU%c7) to compare what sofware requested to what the hardware delivered. These sysfs attributes are discovered when turbostat starts, rather than being "built in". So the --show and --hide parameters do not know about these dynamic column names. However "--show sysfs" and "--hide sysfs" act on the entire group of columns: turbostat --show sysfs ... cpu4: POLL: CPUIDLE CORE POLL IDLE cpu4: C1: MWAIT 0x00 cpu4: C1E: MWAIT 0x01 cpu4: C3: MWAIT 0x10 cpu4: C6: MWAIT 0x20 cpu4: C7s: MWAIT 0x32 ... C1 C1E C3 C6 C7s C1% C1E% C3% C6% C7s% 3 6 5 1 188 0.00 0.02 0.00 0.00 99.93 0 6 5 0 58 0.00 0.16 0.02 0.00 99.70 0 0 0 0 9 0.00 0.00 0.00 0.00 99.96 0 0 0 1 24 0.00 0.00 0.00 0.02 99.93 0 0 0 0 9 0.00 0.00 0.00 0.00 99.97 0 0 0 0 32 0.00 0.00 0.00 0.00 99.96 0 0 0 0 7 0.00 0.00 0.00 0.00 99.98 2 0 0 0 36 0.00 0.00 0.00 0.00 99.97 1 0 0 0 13 0.00 0.00 0.00 0.00 99.98 Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-09 18:25:22 -05:00
probe_sysfs();
tools/power turbostat: Increase the limit for fd opened When running turbostat, a system with 512 cpus reaches the limit for maximum number of file descriptors that can be opened. To solve this problem, the limit is raised to 2^15, which is a large enough number. Below data is collected from AMD server systems while running turbostat: |-----------+-------------------------------| | # of cpus | # of opened fds for turbostat | |-----------+-------------------------------| | 128 | 260 | |-----------+-------------------------------| | 192 | 388 | |-----------+-------------------------------| | 512 | 1028 | |-----------+-------------------------------| So, the new max limit would be sufficient up to 2^14 cpus (but this also depends on how many counters are enabled). Reviewed-by: Doug Smythies <dsmythies@telus.net> Tested-by: Doug Smythies <dsmythies@telus.net> Signed-off-by: Wyes Karny <wyes.karny@amd.com> Signed-off-by: Len Brown <len.brown@intel.com>
2023-10-03 05:07:51 +00:00
if (!getuid())
set_rlimit();
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
turbostat_init();
tools/power turbostat: Add --no-msr option Add --no-msr option to allow users to run turbostat without accessing MSRs via the MSR driver. Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com> Reviewed-by: Len Brown <len.brown@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2024-01-11 15:48:09 +01:00
if (!no_msr)
msr_sum_record();
tools/power turbostat: fix dump for AMD cpus turbostat --Dump exits early with status 243 (-13) get_counters() calls get_msr_sum() on zen CPUS for MSR_PKG_ENERGY_STAT, but per_cpu_msr_sum has not been initialized. Signed-off-by: Dan Merillat <git@dan.eginity.com> Signed-off-by: Len Brown <len.brown@intel.com>
2021-05-09 05:08:55 -04:00
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 17:13:15 +02:00
/* dump counters and exit */
if (dump_only)
return get_and_dump_counters();
tools/power turbostat: Add --list option to show available header names It is handy to know the list of column header names, so that they can be used with --add and --skip The new --list option shows them: sudo ./turbostat --list --hide sysfs ,Core,CPU,Avg_MHz,Busy%,Bzy_MHz,TSC_MHz,IRQ,SMI,CPU%c1,CPU%c3,CPU%c6,CPU%c7,CoreTmp,PkgTmp,GFX%rc6,GFXMHz,PkgWatt,CorWatt,GFXWatt Signed-off-by: Len Brown <len.brown@intel.com>
2017-02-15 17:15:11 -05:00
/* list header and exit */
if (list_header_only) {
print_header(",");
flush_output_stdout();
return 0;
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-22 23:53:03 -04:00
/*
* if any params left, it must be a command to fork
*/
if (argc - optind)
return fork_it(argv + optind);
else
turbostat_loop();
return 0;
}
Reference in a new issue Copy permalink