lkml.org 
[lkml]   [2022]   [Jan]   [17]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
From
Subject[PATCH 3/4] perf evlist: No need to setup affinities when enabling events for pid targets
Date
From: Arnaldo Carvalho de Melo <acme@redhat.com>

When the target is a pid, not started by 'perf stat' we need to enable
the events, and in that case there is no need to setup affinities as we
use a dummy CPU map, with just one entry set to -1.

So stop doing it to avoid this needless call to sched_getaffinity():

# strace -ke sched_getaffinity perf stat -e cycles -p 241957 sleep 1
<SNIP>
sched_getaffinity(0, 512, [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31]) = 8
> /usr/lib64/libc-2.33.so(sched_getaffinity@@GLIBC_2.3.4+0x1a) [0xe6eea]
> /var/home/acme/bin/perf(affinity__setup+0x6a) [0x5329ca]
> /var/home/acme/bin/perf(__evlist__enable.constprop.0+0x23) [0x4b9693]
> /var/home/acme/bin/perf(enable_counters+0x14d) [0x42de5d]
> /var/home/acme/bin/perf(cmd_stat+0x2358) [0x4310c8]
> /var/home/acme/bin/perf(run_builtin+0x6a) [0x4a2cfa]
> /var/home/acme/bin/perf(main+0x612) [0x40f8c2]
> /usr/lib64/libc-2.33.so(__libc_start_main+0xd4) [0x27b74]
> /var/home/acme/bin/perf(_start+0x2d) [0x40fadd]
<SNIP>

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
tools/perf/util/evlist.c | 14 +++++++++-----
1 file changed, 9 insertions(+), 5 deletions(-)

diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 6e88d404b5b3e96f..ae6d4363da76ec56 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -487,12 +487,16 @@ static void __evlist__enable(struct evlist *evlist, char *evsel_name)
{
struct evsel *pos;
struct evlist_cpu_iterator evlist_cpu_itr;
- struct affinity affinity;
+ struct affinity saved_affinity, *affinity = NULL;

- if (affinity__setup(&affinity) < 0)
- return;
+ // See explanation in evlist__close()
+ if (!cpu_map__is_dummy(evlist->core.cpus)) {
+ if (affinity__setup(&saved_affinity) < 0)
+ return;
+ affinity = &saved_affinity;
+ }

- evlist__for_each_cpu(evlist_cpu_itr, evlist, &affinity) {
+ evlist__for_each_cpu(evlist_cpu_itr, evlist, affinity) {
pos = evlist_cpu_itr.evsel;
if (evsel__strcmp(pos, evsel_name))
continue;
@@ -500,7 +504,7 @@ static void __evlist__enable(struct evlist *evlist, char *evsel_name)
continue;
evsel__enable_cpu(pos, evlist_cpu_itr.cpu_map_idx);
}
- affinity__cleanup(&affinity);
+ affinity__cleanup(affinity);
evlist__for_each_entry(evlist, pos) {
if (evsel__strcmp(pos, evsel_name))
continue;
--
2.34.1
\
 
 \ /
  Last update: 2022-01-17 17:10    [W:0.071 / U:0.064 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site