perf test: Speed up some tests using perf list
Namhyung Kim
maintainerPATCH 1/1 perf test: Speed up some tests using perf list
On my system, perf list is very slow to print the whole events. I think there's a performance issue in SDT and uprobes event listing. I noticed this issue while running perf test on x86 but it takes long to check some CoreSight event which should be skipped quickly.
Anyway, some test uses perf list to check whether the required event is available before running the test. The perf list command can take an argument to specify event class or (glob) pattern. But glob pattern is only to suppress output for unmatched ones after checking all events.
In this case, specifying event class is better to reduce the number of events it checks and to avoid buggy subsystems entirely.
No functional changes intended.
--- tools/perf/tests/shell/lib/coresight.sh | 2 +- tools/perf/tests/shell/lock_contention.sh | 2 +- tools/perf/tests/shell/record.sh | 2 +- tools/perf/tests/shell/test_arm_coresight.sh | 2 +- tools/perf/tests/shell/test_arm_coresight_disasm.sh | 2 +- tools/perf/tests/shell/test_arm_spe.sh | 2 +- tools/perf/tests/shell/test_arm_spe_fork.sh | 2 +- tools/perf/tests/shell/test_intel_pt.sh | 2 +- tools/perf/tests/shell/trace+probe_vfs_getname.sh | 2 +- 9 files changed, 9 insertions(+), 9 deletions(-)
tools/perf/tests/shell/lib/coresight.sh +1 -1
@@ -18,7 +18,7 @@ BIN="$DIR/$TEST" # If the test tool/binary does not exist and is executable then skip the test if ! test -x "$BIN"; then exit 2; fi # If CoreSight is not available, skip the test -perf list cs_etm | grep -q cs_etm || exit 2 +perf list pmu | grep -q cs_etm || exit 2 DATD="." # If the data dir env is set then make the data dir use that instead of ./ if test -n "$PERF_TEST_CORESIGHT_DATADIR"; then
tools/perf/tests/shell/lock_contention.sh +1 -1
@@ -27,7 +27,7 @@ check() { exit fi - if ! perf list | grep -q lock:contention_begin; then + if ! perf list tracepoint | grep -q lock:contention_begin; then echo "[Skip] No lock contention tracepoints" err=2 exit
tools/perf/tests/shell/record.sh +1 -1
@@ -94,7 +94,7 @@ test_per_thread() { test_register_capture() { echo "Register capture test" - if ! perf list | grep -q 'br_inst_retired.near_call' + if ! perf list pmu | grep -q 'br_inst_retired.near_call' then echo "Register capture test [Skipped missing event]" return
tools/perf/tests/shell/test_arm_coresight.sh +1 -1
@@ -12,7 +12,7 @@ glb_err=0 skip_if_no_cs_etm_event() { - perf list | grep -q 'cs_etm//' && return 0 + perf list pmu | grep -q 'cs_etm//' && return 0 # cs_etm event doesn't exist return 2
tools/perf/tests/shell/test_arm_coresight_disasm.sh +1 -1
@@ -8,7 +8,7 @@ # the script. Test all 3 parts are working correctly by running the script. skip_if_no_cs_etm_event() { - perf list | grep -q 'cs_etm//' && return 0 + perf list pmu | grep -q 'cs_etm//' && return 0 # cs_etm event doesn't exist return 2
tools/perf/tests/shell/test_arm_spe.sh +1 -1
@@ -9,7 +9,7 @@ # German Gomez <[email protected]>, 2021 skip_if_no_arm_spe_event() { - perf list | grep -E -q 'arm_spe_[0-9]+//' && return 0 + perf list pmu | grep -E -q 'arm_spe_[0-9]+//' && return 0 # arm_spe event doesn't exist return 2
tools/perf/tests/shell/test_arm_spe_fork.sh +1 -1
@@ -5,7 +5,7 @@ # German Gomez <[email protected]>, 2022 skip_if_no_arm_spe_event() { - perf list | grep -E -q 'arm_spe_[0-9]+//' && return 0 + perf list pmu | grep -E -q 'arm_spe_[0-9]+//' && return 0 return 2 }
tools/perf/tests/shell/test_intel_pt.sh +1 -1
@@ -5,7 +5,7 @@ set -e # Skip if no Intel PT -perf list | grep -q 'intel_pt//' || exit 2 +perf list pmu | grep -q 'intel_pt//' || exit 2 shelldir=$(dirname "$0") # shellcheck source=lib/waiting.sh
tools/perf/tests/shell/trace+probe_vfs_getname.sh +1 -1
@@ -19,7 +19,7 @@ skip_if_no_perf_trace || exit 2 . "$(dirname $0)"/lib/probe_vfs_getname.sh trace_open_vfs_getname() { - evts="$(echo "$(perf list syscalls:sys_enter_open* 2>/dev/null | grep -E 'open(at)? ' | sed -r 's/.*sys_enter_([a-z]+) +\[.*$/\1/')" | sed ':a;N;s:\n:,:g')" + evts="$(echo "$(perf list tracepoint 2>/dev/null | grep -E 'syscalls:sys_enter_open(at)? ' | sed -r 's/.*sys_enter_([a-z]+) +\[.*$/\1/')" | sed ':a;N;s:\n:,:g')" perf trace -e $evts touch $file 2>&1 | \ grep -E " +[0-9]+\.[0-9]+ +\( +[0-9]+\.[0-9]+ ms\): +touch/[0-9]+ open(at)?\((dfd: +CWD, +)?filename: +\"?${file}\"?, +flags: CREAT\|NOCTTY\|NONBLOCK\|WRONLY, +mode: +IRUGO\|IWUGO\) += +[0-9]+$" }
Signature · 1 line
2.47.0.rc1.288.g06298d1525-goog
Quoted 13 lines · Namhyung Kim
On my system, perf list is very slow to print the whole events. I think there's a performance issue in SDT and uprobes event listing. I noticed this issue while running perf test on x86 but it takes long to check some CoreSight event which should be skipped quickly.
Anyway, some test uses perf list to check whether the required event is available before running the test. The perf list command can take an argument to specify event class or (glob) pattern. But glob pattern is only to suppress output for unmatched ones after checking all events.
In this case, specifying event class is better to reduce the number of events it checks and to avoid buggy subsystems entirely.
Quoted 6 lines
No functional changes intended.
Cc German Gomez <[email protected]>Cc Carsten Haitzler <[email protected]>Signed-off-by Namhyung Kim <[email protected]>---
Quoted 17 lines · James Clark
Quoted 13 lines · Namhyung Kim
On my system, perf list is very slow to print the whole events. I think there's a performance issue in SDT and uprobes event listing. I noticed this issue while running perf test on x86 but it takes long to check some CoreSight event which should be skipped quickly.
Anyway, some test uses perf list to check whether the required event is available before running the test. The perf list command can take an argument to specify event class or (glob) pattern. But glob pattern is only to suppress output for unmatched ones after checking all events.
In this case, specifying event class is better to reduce the number of events it checks and to avoid buggy subsystems entirely.
Quoted 8 lines
Quoted 6 lines
No functional changes intended.
Cc German Gomez <[email protected]>Cc Carsten Haitzler <[email protected]>Signed-off-by Namhyung Kim <[email protected]>---
Reviewed-by James Clark <[email protected]>
There's a nice better than >2x speed up waiting in too :-)
https://lore.kernel.org/lkml/[email protected]/
Thanks,
Ian
Quoted 11 lines · Namhyung Kim
On my system, perf list is very slow to print the whole events. I think there's a performance issue in SDT and uprobes event listing. I noticed this issue while running perf test on x86 but it takes long to check some CoreSight event which should be skipped quickly.
Anyway, some test uses perf list to check whether the required event is available before running the test. The perf list command can take an argument to specify event class or (glob) pattern. But glob pattern is only to suppress output for unmatched ones after checking all events.
[...]
Applied to perf-tools-next, thanks!
Best regards,
Namhyung