linux/tools/perf/tests/shell/stat_bpf_counters.sh

75 lines
2 KiB
Bash
Raw Permalink Normal View History

#!/bin/bash
# perf stat --bpf-counters test (exclusive)
# SPDX-License-Identifier: GPL-2.0
set -e
workload="perf test -w sqrtloop"
# check whether $2 is within +/- 20% of $1
compare_number()
{
first_num=$1
second_num=$2
# upper bound is first_num * 120%
upper=$(expr $first_num + $first_num / 5 )
# lower bound is first_num * 80%
lower=$(expr $first_num - $first_num / 5 )
if [ $second_num -gt $upper ] || [ $second_num -lt $lower ]; then
echo "The difference between $first_num and $second_num are greater than 20%."
exit 1
fi
}
check_counts()
{
perf test stat_bpf_counter.sh: Stabilize the test results The test has been failing for some time when two separate runs of perf benchmarks are recorded for cycles events and their counts are compared, while once the recording was done with option --bpf-counters and once without it. It is expected that the count of the samples should be within a certain range, firstly the difference was set to be within 10%, which was then later raised to 20%. However, the test case keeps failing on certain architectures as recording the provided benchmark can produce completely different counts based on the current load of the system. Sampling two separate runs on intel-eaglestream-spr-13 of "perf stat --no-big-num -e cycles -- perf bench sched messaging -g 1 -l 100 -t": Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 396782898 cycles 0.010051983 seconds time elapsed 0.008664000 seconds user 0.097058000 seconds sys Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 1431133032 cycles 0.021803714 seconds time elapsed 0.023377000 seconds user 0.349918000 seconds sys , which is ranging from 400mil to 1400mil samples. Instead of recording the cycles use instructions event, which provides more stable values. At the same time change the tested workload to one of the provided testing workloads by perf that is not based on a scheduler, which can provide another dependency on the current load. Sampling instructions event with the new workload provide much more stable results on intel-eaglestream-spr-13 of "perf stat --no-big-num -e instructions -- perf test -w brstack": Performance counter stats for 'perf test -w brstack': 64584494 instructions 0.009173945 seconds time elapsed 0.007262000 seconds user 0.002071000 seconds sys Performance counter stats for 'perf test -w brstack': 64672669 instructions 0.008888135 seconds time elapsed 0.005018000 seconds user 0.004018000 seconds sys Signed-off-by: Veronika Molnarova <vmolnaro@redhat.com> Acked-by: Namhyung Kim <namhyung@kernel.org> Cc: mpetlan@redhat.com Signed-off-by: Namhyung Kim <namhyung@kernel.org> Link: https://lore.kernel.org/r/20240625092001.10909-1-vmolnaro@redhat.com
2024-06-25 11:20:01 +02:00
base_instructions=$1
bpf_instructions=$2
perf test stat_bpf_counter.sh: Stabilize the test results The test has been failing for some time when two separate runs of perf benchmarks are recorded for cycles events and their counts are compared, while once the recording was done with option --bpf-counters and once without it. It is expected that the count of the samples should be within a certain range, firstly the difference was set to be within 10%, which was then later raised to 20%. However, the test case keeps failing on certain architectures as recording the provided benchmark can produce completely different counts based on the current load of the system. Sampling two separate runs on intel-eaglestream-spr-13 of "perf stat --no-big-num -e cycles -- perf bench sched messaging -g 1 -l 100 -t": Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 396782898 cycles 0.010051983 seconds time elapsed 0.008664000 seconds user 0.097058000 seconds sys Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 1431133032 cycles 0.021803714 seconds time elapsed 0.023377000 seconds user 0.349918000 seconds sys , which is ranging from 400mil to 1400mil samples. Instead of recording the cycles use instructions event, which provides more stable values. At the same time change the tested workload to one of the provided testing workloads by perf that is not based on a scheduler, which can provide another dependency on the current load. Sampling instructions event with the new workload provide much more stable results on intel-eaglestream-spr-13 of "perf stat --no-big-num -e instructions -- perf test -w brstack": Performance counter stats for 'perf test -w brstack': 64584494 instructions 0.009173945 seconds time elapsed 0.007262000 seconds user 0.002071000 seconds sys Performance counter stats for 'perf test -w brstack': 64672669 instructions 0.008888135 seconds time elapsed 0.005018000 seconds user 0.004018000 seconds sys Signed-off-by: Veronika Molnarova <vmolnaro@redhat.com> Acked-by: Namhyung Kim <namhyung@kernel.org> Cc: mpetlan@redhat.com Signed-off-by: Namhyung Kim <namhyung@kernel.org> Link: https://lore.kernel.org/r/20240625092001.10909-1-vmolnaro@redhat.com
2024-06-25 11:20:01 +02:00
if [ "$base_instructions" = "<not" ]; then
echo "Skipping: instructions event not counted"
exit 2
fi
perf test stat_bpf_counter.sh: Stabilize the test results The test has been failing for some time when two separate runs of perf benchmarks are recorded for cycles events and their counts are compared, while once the recording was done with option --bpf-counters and once without it. It is expected that the count of the samples should be within a certain range, firstly the difference was set to be within 10%, which was then later raised to 20%. However, the test case keeps failing on certain architectures as recording the provided benchmark can produce completely different counts based on the current load of the system. Sampling two separate runs on intel-eaglestream-spr-13 of "perf stat --no-big-num -e cycles -- perf bench sched messaging -g 1 -l 100 -t": Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 396782898 cycles 0.010051983 seconds time elapsed 0.008664000 seconds user 0.097058000 seconds sys Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 1431133032 cycles 0.021803714 seconds time elapsed 0.023377000 seconds user 0.349918000 seconds sys , which is ranging from 400mil to 1400mil samples. Instead of recording the cycles use instructions event, which provides more stable values. At the same time change the tested workload to one of the provided testing workloads by perf that is not based on a scheduler, which can provide another dependency on the current load. Sampling instructions event with the new workload provide much more stable results on intel-eaglestream-spr-13 of "perf stat --no-big-num -e instructions -- perf test -w brstack": Performance counter stats for 'perf test -w brstack': 64584494 instructions 0.009173945 seconds time elapsed 0.007262000 seconds user 0.002071000 seconds sys Performance counter stats for 'perf test -w brstack': 64672669 instructions 0.008888135 seconds time elapsed 0.005018000 seconds user 0.004018000 seconds sys Signed-off-by: Veronika Molnarova <vmolnaro@redhat.com> Acked-by: Namhyung Kim <namhyung@kernel.org> Cc: mpetlan@redhat.com Signed-off-by: Namhyung Kim <namhyung@kernel.org> Link: https://lore.kernel.org/r/20240625092001.10909-1-vmolnaro@redhat.com
2024-06-25 11:20:01 +02:00
if [ "$bpf_instructions" = "<not" ]; then
echo "Failed: instructions not counted with --bpf-counters"
exit 1
fi
}
test_bpf_counters()
{
printf "Testing --bpf-counters "
perf test stat_bpf_counter.sh: Stabilize the test results The test has been failing for some time when two separate runs of perf benchmarks are recorded for cycles events and their counts are compared, while once the recording was done with option --bpf-counters and once without it. It is expected that the count of the samples should be within a certain range, firstly the difference was set to be within 10%, which was then later raised to 20%. However, the test case keeps failing on certain architectures as recording the provided benchmark can produce completely different counts based on the current load of the system. Sampling two separate runs on intel-eaglestream-spr-13 of "perf stat --no-big-num -e cycles -- perf bench sched messaging -g 1 -l 100 -t": Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 396782898 cycles 0.010051983 seconds time elapsed 0.008664000 seconds user 0.097058000 seconds sys Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 1431133032 cycles 0.021803714 seconds time elapsed 0.023377000 seconds user 0.349918000 seconds sys , which is ranging from 400mil to 1400mil samples. Instead of recording the cycles use instructions event, which provides more stable values. At the same time change the tested workload to one of the provided testing workloads by perf that is not based on a scheduler, which can provide another dependency on the current load. Sampling instructions event with the new workload provide much more stable results on intel-eaglestream-spr-13 of "perf stat --no-big-num -e instructions -- perf test -w brstack": Performance counter stats for 'perf test -w brstack': 64584494 instructions 0.009173945 seconds time elapsed 0.007262000 seconds user 0.002071000 seconds sys Performance counter stats for 'perf test -w brstack': 64672669 instructions 0.008888135 seconds time elapsed 0.005018000 seconds user 0.004018000 seconds sys Signed-off-by: Veronika Molnarova <vmolnaro@redhat.com> Acked-by: Namhyung Kim <namhyung@kernel.org> Cc: mpetlan@redhat.com Signed-off-by: Namhyung Kim <namhyung@kernel.org> Link: https://lore.kernel.org/r/20240625092001.10909-1-vmolnaro@redhat.com
2024-06-25 11:20:01 +02:00
base_instructions=$(perf stat --no-big-num -e instructions -- $workload 2>&1 | awk '/instructions/ {print $1}')
bpf_instructions=$(perf stat --no-big-num --bpf-counters -e instructions -- $workload 2>&1 | awk '/instructions/ {print $1}')
check_counts $base_instructions $bpf_instructions
compare_number $base_instructions $bpf_instructions
echo "[Success]"
}
test_bpf_modifier()
{
printf "Testing bpf event modifier "
perf test stat_bpf_counter.sh: Stabilize the test results The test has been failing for some time when two separate runs of perf benchmarks are recorded for cycles events and their counts are compared, while once the recording was done with option --bpf-counters and once without it. It is expected that the count of the samples should be within a certain range, firstly the difference was set to be within 10%, which was then later raised to 20%. However, the test case keeps failing on certain architectures as recording the provided benchmark can produce completely different counts based on the current load of the system. Sampling two separate runs on intel-eaglestream-spr-13 of "perf stat --no-big-num -e cycles -- perf bench sched messaging -g 1 -l 100 -t": Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 396782898 cycles 0.010051983 seconds time elapsed 0.008664000 seconds user 0.097058000 seconds sys Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 1431133032 cycles 0.021803714 seconds time elapsed 0.023377000 seconds user 0.349918000 seconds sys , which is ranging from 400mil to 1400mil samples. Instead of recording the cycles use instructions event, which provides more stable values. At the same time change the tested workload to one of the provided testing workloads by perf that is not based on a scheduler, which can provide another dependency on the current load. Sampling instructions event with the new workload provide much more stable results on intel-eaglestream-spr-13 of "perf stat --no-big-num -e instructions -- perf test -w brstack": Performance counter stats for 'perf test -w brstack': 64584494 instructions 0.009173945 seconds time elapsed 0.007262000 seconds user 0.002071000 seconds sys Performance counter stats for 'perf test -w brstack': 64672669 instructions 0.008888135 seconds time elapsed 0.005018000 seconds user 0.004018000 seconds sys Signed-off-by: Veronika Molnarova <vmolnaro@redhat.com> Acked-by: Namhyung Kim <namhyung@kernel.org> Cc: mpetlan@redhat.com Signed-off-by: Namhyung Kim <namhyung@kernel.org> Link: https://lore.kernel.org/r/20240625092001.10909-1-vmolnaro@redhat.com
2024-06-25 11:20:01 +02:00
stat_output=$(perf stat --no-big-num -e instructions/name=base_instructions/,instructions/name=bpf_instructions/b -- $workload 2>&1)
base_instructions=$(echo "$stat_output"| awk '/base_instructions/ {print $1}')
bpf_instructions=$(echo "$stat_output"| awk '/bpf_instructions/ {print $1}')
check_counts $base_instructions $bpf_instructions
compare_number $base_instructions $bpf_instructions
echo "[Success]"
}
# skip if --bpf-counters is not supported
perf test stat_bpf_counter.sh: Stabilize the test results The test has been failing for some time when two separate runs of perf benchmarks are recorded for cycles events and their counts are compared, while once the recording was done with option --bpf-counters and once without it. It is expected that the count of the samples should be within a certain range, firstly the difference was set to be within 10%, which was then later raised to 20%. However, the test case keeps failing on certain architectures as recording the provided benchmark can produce completely different counts based on the current load of the system. Sampling two separate runs on intel-eaglestream-spr-13 of "perf stat --no-big-num -e cycles -- perf bench sched messaging -g 1 -l 100 -t": Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 396782898 cycles 0.010051983 seconds time elapsed 0.008664000 seconds user 0.097058000 seconds sys Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 1431133032 cycles 0.021803714 seconds time elapsed 0.023377000 seconds user 0.349918000 seconds sys , which is ranging from 400mil to 1400mil samples. Instead of recording the cycles use instructions event, which provides more stable values. At the same time change the tested workload to one of the provided testing workloads by perf that is not based on a scheduler, which can provide another dependency on the current load. Sampling instructions event with the new workload provide much more stable results on intel-eaglestream-spr-13 of "perf stat --no-big-num -e instructions -- perf test -w brstack": Performance counter stats for 'perf test -w brstack': 64584494 instructions 0.009173945 seconds time elapsed 0.007262000 seconds user 0.002071000 seconds sys Performance counter stats for 'perf test -w brstack': 64672669 instructions 0.008888135 seconds time elapsed 0.005018000 seconds user 0.004018000 seconds sys Signed-off-by: Veronika Molnarova <vmolnaro@redhat.com> Acked-by: Namhyung Kim <namhyung@kernel.org> Cc: mpetlan@redhat.com Signed-off-by: Namhyung Kim <namhyung@kernel.org> Link: https://lore.kernel.org/r/20240625092001.10909-1-vmolnaro@redhat.com
2024-06-25 11:20:01 +02:00
if ! perf stat -e instructions --bpf-counters true > /dev/null 2>&1; then
perf test: Remove bash construct from stat_bpf_counters.sh test Currently the test skips with an error because == only works in bash: $ ./perf test 91 -v Couldn't bump rlimit(MEMLOCK), failures may take place when creating BPF maps, etc 91: perf stat --bpf-counters test : --- start --- test child forked, pid 44586 ./tests/shell/stat_bpf_counters.sh: 26: [: -v: unexpected operator test child finished with -2 ---- end ---- perf stat --bpf-counters test: Skip Changing == to = does the same thing, but doesn't result in an error: ./perf test 91 -v Couldn't bump rlimit(MEMLOCK), failures may take place when creating BPF maps, etc 91: perf stat --bpf-counters test : --- start --- test child forked, pid 45833 Skipping: --bpf-counters not supported Error: unknown option `bpf-counters' [...] test child finished with -2 ---- end ---- perf stat --bpf-counters test: Skip Signed-off-by: James Clark <james.clark@arm.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Florian Fainelli <f.fainelli@gmail.com> Cc: Ian Rogers <irogers@google.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: John Fastabend <john.fastabend@gmail.com> Cc: KP Singh <kpsingh@kernel.org> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Martin KaFai Lau <kafai@fb.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Song Liu <songliubraving@fb.com> Cc: Sumanth Korikkar <sumanthk@linux.ibm.com> Cc: Thomas Richter <tmricht@linux.ibm.com> Cc: Yonghong Song <yhs@fb.com> Cc: bpf@vger.kernel.org Cc: netdev@vger.kernel.org Link: https://lore.kernel.org/r/20211028134828.65774-2-james.clark@arm.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2021-10-28 14:48:25 +01:00
if [ "$1" = "-v" ]; then
echo "Skipping: --bpf-counters not supported"
perf test stat_bpf_counter.sh: Stabilize the test results The test has been failing for some time when two separate runs of perf benchmarks are recorded for cycles events and their counts are compared, while once the recording was done with option --bpf-counters and once without it. It is expected that the count of the samples should be within a certain range, firstly the difference was set to be within 10%, which was then later raised to 20%. However, the test case keeps failing on certain architectures as recording the provided benchmark can produce completely different counts based on the current load of the system. Sampling two separate runs on intel-eaglestream-spr-13 of "perf stat --no-big-num -e cycles -- perf bench sched messaging -g 1 -l 100 -t": Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 396782898 cycles 0.010051983 seconds time elapsed 0.008664000 seconds user 0.097058000 seconds sys Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 1431133032 cycles 0.021803714 seconds time elapsed 0.023377000 seconds user 0.349918000 seconds sys , which is ranging from 400mil to 1400mil samples. Instead of recording the cycles use instructions event, which provides more stable values. At the same time change the tested workload to one of the provided testing workloads by perf that is not based on a scheduler, which can provide another dependency on the current load. Sampling instructions event with the new workload provide much more stable results on intel-eaglestream-spr-13 of "perf stat --no-big-num -e instructions -- perf test -w brstack": Performance counter stats for 'perf test -w brstack': 64584494 instructions 0.009173945 seconds time elapsed 0.007262000 seconds user 0.002071000 seconds sys Performance counter stats for 'perf test -w brstack': 64672669 instructions 0.008888135 seconds time elapsed 0.005018000 seconds user 0.004018000 seconds sys Signed-off-by: Veronika Molnarova <vmolnaro@redhat.com> Acked-by: Namhyung Kim <namhyung@kernel.org> Cc: mpetlan@redhat.com Signed-off-by: Namhyung Kim <namhyung@kernel.org> Link: https://lore.kernel.org/r/20240625092001.10909-1-vmolnaro@redhat.com
2024-06-25 11:20:01 +02:00
perf --no-pager stat -e instructions --bpf-counters true || true
fi
exit 2
fi
test_bpf_counters
test_bpf_modifier
exit 0