2022-12-05 23:26:51

by Namhyung Kim

Subject: [PATCH] perf stat: Update event skip condition

In print_counter_aggrdata(), it skips some events that has no aggregate
count. It's actually for system-wide per-thread mode and merged uncore
and hybrid events.

Let's update the condition to check them explicitly.

Fixes: 91f85f98da7a ("Display event stats using aggr counts")
Reported-by: Athira Rajeev <[email protected]>
Signed-off-by: Namhyung Kim <[email protected]>
---
Athira, could you please check this fixes the problem?

tools/perf/util/stat-display.c | 16 +++++++++-------
1 file changed, 9 insertions(+), 7 deletions(-)

diff --git a/tools/perf/util/stat-display.c b/tools/perf/util/stat-display.c
index 847acdb5dc40..6c0de52ac4be 100644
--- a/tools/perf/util/stat-display.c
+++ b/tools/perf/util/stat-display.c
@@ -814,7 +814,8 @@ static void print_counter_aggrdata(struct perf_stat_config *config,
os->nr = aggr->nr;
os->evsel = counter;

- if (counter->supported && aggr->nr == 0)
+ /* Skip already merged uncore/hybrid events */
+ if (counter->merged_stat)
return;

uniquify_counter(config, counter);
@@ -823,6 +824,13 @@ static void print_counter_aggrdata(struct perf_stat_config *config,
ena = aggr->counts.ena;
run = aggr->counts.run;

+ /*
+ * Skip value 0 when enabling --per-thread globally, otherwise it will
+ * have too many 0 output.
+ */
+ if (val == 0 && config->aggr_mode == AGGR_THREAD && config->system_wide)
+ return;
+
if (!metric_only) {
if (config->json_output)
fputc('{', output);
@@ -899,9 +907,6 @@ static void print_aggr(struct perf_stat_config *config,
print_metric_begin(config, evlist, os, s);

evlist__for_each_entry(evlist, counter) {
- if (counter->merged_stat)
- continue;
-
print_counter_aggrdata(config, counter, s, os);
}
print_metric_end(config, os);
@@ -928,9 +933,6 @@ static void print_aggr_cgroup(struct perf_stat_config *config,
print_metric_begin(config, evlist, os, s);

evlist__for_each_entry(evlist, counter) {
- if (counter->merged_stat)
- continue;
-
if (counter->cgrp != os->cgrp)
continue;

--
2.39.0.rc0.267.gcb52ba06e7-goog

2022-12-06 15:07:56

by Athira Rajeev

Subject: Re: [PATCH] perf stat: Update event skip condition

> On 06-Dec-2022, at 4:46 AM, Namhyung Kim <[email protected]> wrote:
>
> In print_counter_aggrdata(), it skips some events that has no aggregate
> count. It's actually for system-wide per-thread mode and merged uncore
> and hybrid events.
>
> Let's update the condition to check them explicitly.
>
> Fixes: 91f85f98da7a ("Display event stats using aggr counts")
> Reported-by: Athira Rajeev <[email protected]>
> Signed-off-by: Namhyung Kim <[email protected]>
> ---
> Athira, could you please check this fixes the problem?
>
> tools/perf/util/stat-display.c | 16 +++++++++-------
> 1 file changed, 9 insertions(+), 7 deletions(-)
>
> diff --git a/tools/perf/util/stat-display.c b/tools/perf/util/stat-display.c
> index 847acdb5dc40..6c0de52ac4be 100644
> --- a/tools/perf/util/stat-display.c
> +++ b/tools/perf/util/stat-display.c
> @@ -814,7 +814,8 @@ static void print_counter_aggrdata(struct perf_stat_config *config,
> os->nr = aggr->nr;
> os->evsel = counter;
>
> - if (counter->supported && aggr->nr == 0)
> + /* Skip already merged uncore/hybrid events */
> + if (counter->merged_stat)
> return;

Hi Namhyung,

Thanks for the fix.

This patch removes the merged_stat checks from print_aggr and print_aggr_cgroup.
We also have this check in print_counter which needs to be removed.

With that change,

Acked-by: Athira Rajeev <[email protected]>

>
> uniquify_counter(config, counter);
> @@ -823,6 +824,13 @@ static void print_counter_aggrdata(struct perf_stat_config *config,
> ena = aggr->counts.ena;
> run = aggr->counts.run;
>
> + /*
> + * Skip value 0 when enabling --per-thread globally, otherwise it will
> + * have too many 0 output.
> + */
> + if (val == 0 && config->aggr_mode == AGGR_THREAD && config->system_wide)
> + return;
> +
> if (!metric_only) {
> if (config->json_output)
> fputc('{', output);
> @@ -899,9 +907,6 @@ static void print_aggr(struct perf_stat_config *config,
> print_metric_begin(config, evlist, os, s);
>
> evlist__for_each_entry(evlist, counter) {
> - if (counter->merged_stat)
> - continue;
> -
> print_counter_aggrdata(config, counter, s, os);
> }
> print_metric_end(config, os);
> @@ -928,9 +933,6 @@ static void print_aggr_cgroup(struct perf_stat_config *config,
> print_metric_begin(config, evlist, os, s);
>
> evlist__for_each_entry(evlist, counter) {
> - if (counter->merged_stat)
> - continue;
> -
> if (counter->cgrp != os->cgrp)
> continue;
>
> --
> 2.39.0.rc0.267.gcb52ba06e7-goog
>

2022-12-06 16:24:30

by Arnaldo Carvalho de Melo

Subject: Re: [PATCH] perf stat: Update event skip condition

Em Tue, Dec 06, 2022 at 07:45:29PM +0530, Athira Rajeev escreveu:
>
>
> > On 06-Dec-2022, at 4:46 AM, Namhyung Kim <[email protected]> wrote:
> >
> > In print_counter_aggrdata(), it skips some events that has no aggregate
> > count. It's actually for system-wide per-thread mode and merged uncore
> > and hybrid events.
> >
> > Let's update the condition to check them explicitly.
> >
> > Fixes: 91f85f98da7a ("Display event stats using aggr counts")
> > Reported-by: Athira Rajeev <[email protected]>
> > Signed-off-by: Namhyung Kim <[email protected]>
> > ---
> > Athira, could you please check this fixes the problem?
> >
> > tools/perf/util/stat-display.c | 16 +++++++++-------
> > 1 file changed, 9 insertions(+), 7 deletions(-)
> >
> > diff --git a/tools/perf/util/stat-display.c b/tools/perf/util/stat-display.c
> > index 847acdb5dc40..6c0de52ac4be 100644
> > --- a/tools/perf/util/stat-display.c
> > +++ b/tools/perf/util/stat-display.c
> > @@ -814,7 +814,8 @@ static void print_counter_aggrdata(struct perf_stat_config *config,
> > os->nr = aggr->nr;
> > os->evsel = counter;
> >
> > - if (counter->supported && aggr->nr == 0)
> > + /* Skip already merged uncore/hybrid events */
> > + if (counter->merged_stat)
> > return;
>
> Hi Namhyung,
>
> Thanks for the fix.
>
> This patch removes the merged_stat checks from print_aggr and print_aggr_cgroup.
> We also have this check in print_counter which needs to be removed.
>
> With that change,

Waiting for v2 then.

- Arnaldo

> Acked-by: Athira Rajeev <[email protected]>
>
> >
> > uniquify_counter(config, counter);
> > @@ -823,6 +824,13 @@ static void print_counter_aggrdata(struct perf_stat_config *config,
> > ena = aggr->counts.ena;
> > run = aggr->counts.run;
> >
> > + /*
> > + * Skip value 0 when enabling --per-thread globally, otherwise it will
> > + * have too many 0 output.
> > + */
> > + if (val == 0 && config->aggr_mode == AGGR_THREAD && config->system_wide)
> > + return;
> > +
> > if (!metric_only) {
> > if (config->json_output)
> > fputc('{', output);
> > @@ -899,9 +907,6 @@ static void print_aggr(struct perf_stat_config *config,
> > print_metric_begin(config, evlist, os, s);
> >
> > evlist__for_each_entry(evlist, counter) {
> > - if (counter->merged_stat)
> > - continue;
> > -
> > print_counter_aggrdata(config, counter, s, os);
> > }
> > print_metric_end(config, os);
> > @@ -928,9 +933,6 @@ static void print_aggr_cgroup(struct perf_stat_config *config,
> > print_metric_begin(config, evlist, os, s);
> >
> > evlist__for_each_entry(evlist, counter) {
> > - if (counter->merged_stat)
> > - continue;
> > -
> > if (counter->cgrp != os->cgrp)
> > continue;
> >
> > --
> > 2.39.0.rc0.267.gcb52ba06e7-goog
> >

--

- Arnaldo