2013-07-05 17:06:36

by Stephane Eranian

Subject: [PATCH] perf stat: fix per-socket output bug for uncore events

This patch fixes a problem reported by Andi Kleen on perf
stat when measuring uncore events:

# perf stat --per-socket -e uncore_pcu/event=0x0/ -I1000 -a sleep 2

It would not report counts for the second socket. That was due to a
cpu mapping bug in print_aggr().

This patch also fixes the socket numbering bug for <not counted>
events.

Reported-by: Andi Kleen <[email protected]>
Signed-off-by: Stephane Eranian <[email protected]>
---

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 7e910ba..128e168 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -924,7 +924,7 @@ static void abs_printout(int cpu, int nr, struct perf_evsel *evsel, double avg)
static void print_aggr(char *prefix)
{
struct perf_evsel *counter;
- int cpu, s, s2, id, nr;
+ int cpu, cpu2, s, s2, id, nr;
u64 ena, run, val;

if (!(aggr_map || aggr_get_id))
@@ -936,7 +936,8 @@ static void print_aggr(char *prefix)
val = ena = run = 0;
nr = 0;
for (cpu = 0; cpu < perf_evsel__nr_cpus(counter); cpu++) {
- s2 = aggr_get_id(evsel_list->cpus, cpu);
+ cpu2 = perf_evsel__cpus(counter)->map[cpu];
+ s2 = aggr_get_id(evsel_list->cpus, cpu2);
if (s2 != id)
continue;
val += counter->counts->cpu[cpu].val;
@@ -948,7 +949,7 @@ static void print_aggr(char *prefix)
fprintf(output, "%s", prefix);

if (run == 0 || ena == 0) {
- aggr_printout(counter, cpu, nr);
+ aggr_printout(counter, id, nr);

fprintf(output, "%*s%s%*s",
csv_output ? 0 : 18,

2013-07-05 19:12:35

by Andi Kleen

Subject: Re: [PATCH] perf stat: fix per-socket output bug for uncore events

> Reported-by: Andi Kleen <[email protected]>
> Signed-off-by: Stephane Eranian <[email protected]>

Tested-by: Andi Kleen <[email protected]>

Works for me. Thanks.

-Andi

2013-07-05 20:37:52

by Arnaldo Carvalho de Melo

Subject: Re: [PATCH] perf stat: fix per-socket output bug for uncore events

Em Fri, Jul 05, 2013 at 07:06:45PM +0200, Stephane Eranian escreveu:
>
> This patch fixes a problem reported by Andi Kleen on perf
> stat when measuring uncore events:
>
> # perf stat --per-socket -e uncore_pcu/event=0x0/ -I1000 -a sleep 2
>
> It would not report counts for the second socket. That was due to a
> cpu mapping bug in print_aggr().
>
> This patch also fixes the socket numbering bug for <not counted>
> events.
>
> Reported-by: Andi Kleen <[email protected]>
> Signed-off-by: Stephane Eranian <[email protected]>

Thanks, applied.

- Arnaldo

2013-07-12 08:53:01

by tip-bot for Vasyl Gomonovych

Subject: [tip:perf/urgent] perf stat: Fix per-socket output bug for uncore events

Commit-ID: 582ec0829b3dd74d8c0f58403a3f9df8cbaa9c7d
Gitweb: http://git.kernel.org/tip/582ec0829b3dd74d8c0f58403a3f9df8cbaa9c7d
Author: Stephane Eranian <[email protected]>
AuthorDate: Fri, 5 Jul 2013 19:06:45 +0200
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Mon, 8 Jul 2013 18:01:46 -0300

perf stat: Fix per-socket output bug for uncore events

This patch fixes a problem reported by Andi Kleen on perf
stat when measuring uncore events:

# perf stat --per-socket -e uncore_pcu/event=0x0/ -I1000 -a sleep 2

It would not report counts for the second socket. That was due to a
cpu mapping bug in print_aggr().

This patch also fixes the socket numbering bug for <not counted>
events.

Reported-by: Andi Kleen <[email protected]>
Signed-off-by: Stephane Eranian <[email protected]>
Tested-by: Andi Kleen <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: [email protected]
Link: http://lkml.kernel.org/r/20130705170645.GA32519@quad
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/builtin-stat.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 95768af..352fbd7 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -924,7 +924,7 @@ static void abs_printout(int cpu, int nr, struct perf_evsel *evsel, double avg)
static void print_aggr(char *prefix)
{
struct perf_evsel *counter;
- int cpu, s, s2, id, nr;
+ int cpu, cpu2, s, s2, id, nr;
u64 ena, run, val;

if (!(aggr_map || aggr_get_id))
@@ -936,7 +936,8 @@ static void print_aggr(char *prefix)
val = ena = run = 0;
nr = 0;
for (cpu = 0; cpu < perf_evsel__nr_cpus(counter); cpu++) {
- s2 = aggr_get_id(evsel_list->cpus, cpu);
+ cpu2 = perf_evsel__cpus(counter)->map[cpu];
+ s2 = aggr_get_id(evsel_list->cpus, cpu2);
if (s2 != id)
continue;
val += counter->counts->cpu[cpu].val;
@@ -948,7 +949,7 @@ static void print_aggr(char *prefix)
fprintf(output, "%s", prefix);

if (run == 0 || ena == 0) {
- aggr_printout(counter, cpu, nr);
+ aggr_printout(counter, id, nr);

fprintf(output, "%*s%s%*s",
csv_output ? 0 : 18,