From: Kyrylo Tkachov <[email protected]>
The autoprofile bootstrap support probes for Arm SPE from gcc/configure,
gcc-auto-profile, and the profile-optimization testsuite support. Each
bare "perf list" scans every event class even though arm_spe is a PMU.
This can make the probe unnecessarily slow on systems with many events.
Use "perf list pmu" at all three call sites. The PMU event class filter
has long been supported by perf and still reports arm_spe. Stubbed BRBE,
SPE, and fallback paths select the same profiler as before.
This patch actually saves a few minutes of end-to-end bootstrap time on my SPE
AArch64 many-core system.
Ok for trunk?
Thanks,
Kyrill
gcc/ChangeLog:
* config/aarch64/gcc-auto-profile: Probe for arm_spe with
"perf list pmu" instead of "perf list".
* configure.ac: Likewise.
* configure: Regenerate.
gcc/testsuite/ChangeLog:
* lib/profopt.exp: Likewise.
Signed-off-by: Kyrylo Tkachov <[email protected]>
---
gcc/config/aarch64/gcc-auto-profile | 2 +-
gcc/configure | 2 +-
gcc/configure.ac | 2 +-
gcc/testsuite/lib/profopt.exp | 2 +-
4 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/gcc/config/aarch64/gcc-auto-profile
b/gcc/config/aarch64/gcc-auto-profile
index 8f7502b30e3..6e33047698c 100755
--- a/gcc/config/aarch64/gcc-auto-profile
+++ b/gcc/config/aarch64/gcc-auto-profile
@@ -45,7 +45,7 @@ if [ "$use_brbe" = true ] ; then
set -x
perf record --inherit -o perf.data -j any,$FLAGS "$@"
set +x
-elif [ -n "$(perf list | grep arm_spe)" ] ; then
+elif [ -n "$(perf list pmu | grep arm_spe)" ]; then
echo >&2 "Info: Using SPE to collect branch profiles"
set -x
perf record --inherit -o perf.data -e arm_spe_0/branch_filter=1/ "$@"
diff --git a/gcc/configure b/gcc/configure
index 8f2ecfcffdc..cf2dfd99235 100755
--- a/gcc/configure
+++ b/gcc/configure
@@ -35189,7 +35189,7 @@ $as_echo_n "checking AArch64 build supports SPE
profiling for autoprofiledbootst
if test ! -n "$($PERF record -j any,u -o /dev/null /bin/true 2>&1 |
grep "PMU Hardware or event type doesn't support branch stack sampling.")"; then
AFDO_PROFILER=perf
aarch64_spe_support="using BRBE instead"
- elif test -n "$($PERF list | grep arm_spe)"; then
+ elif test -n "$($PERF list pmu | grep arm_spe)"; then
AFDO_PROFILER=perf_spe
aarch64_spe_support=yes
fi
diff --git a/gcc/configure.ac b/gcc/configure.ac
index 030ee373a6a..77ec150c3e2 100644
--- a/gcc/configure.ac
+++ b/gcc/configure.ac
@@ -7881,7 +7881,7 @@ if test x$PERF != xno; then
if test ! -n "$($PERF record -j any,u -o /dev/null /bin/true 2>&1 |
grep "PMU Hardware or event type doesn't support branch stack sampling.")"; then
AFDO_PROFILER=perf
aarch64_spe_support="using BRBE instead"
- elif test -n "$($PERF list | grep arm_spe)"; then
+ elif test -n "$($PERF list pmu | grep arm_spe)"; then
AFDO_PROFILER=perf_spe
aarch64_spe_support=yes
fi
diff --git a/gcc/testsuite/lib/profopt.exp b/gcc/testsuite/lib/profopt.exp
index 4f3f61958aa..cd80d5e9402 100644
--- a/gcc/testsuite/lib/profopt.exp
+++ b/gcc/testsuite/lib/profopt.exp
@@ -456,7 +456,7 @@ proc profopt-execute { src } {
set bprefix "afdo."
set compiler [lindex $GCC_UNDER_TEST 0]
set autofdo_version [string trim [lindex [remote_exec target
"$compiler --print-autofdo-gcov-version"] 1] ]
- set profiler_check [string trim [lindex [remote_exec target
"perf --no-pager list | grep arm_spe"] 1]]
+ set profiler_check [string trim [lindex [remote_exec target
"perf --no-pager list pmu | grep arm_spe"] 1]]
if {$profiler_check ne ""} {
set profiler "perf_spe"
} else {
--
2.50.1 (Apple Git-155)
