Re: [PATCH v3 1/4] perf stat: Fix wrong per-thread runtime stat for interval mode

[Jin, Yao]

Hi Jiri,

On 5/7/2020 11:19 PM, Jiri Olsa wrote:
> On Thu, May 07, 2020 at 02:58:19PM +0800, Jin Yao wrote:
>
> SNIP
>
> > diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
> > index e0c1ad23c768..97ee941649e6 100644
> > --- a/tools/perf/builtin-stat.c
> > +++ b/tools/perf/builtin-stat.c
> > @@ -351,6 +351,16 @@ static void read_counters(struct timespec *rs)
> >   	}
> >   }
> >   
> > +static void thread_stats_reset(struct perf_stat_config *config)
> > +{
> > +	int i;
> > +
> > +	if (config->stats) {
> > +		for (i = 0; i < config->stats_num; i++)
> > +			perf_stat__reset_shadow_per_stat(&config->stats[i]);
> > +	}
> > +}
> > +
> >   static void process_interval(void)
> >   {
> >   	struct timespec ts, rs;
> > @@ -359,6 +369,7 @@ static void process_interval(void)
> >   	diff_timespec(&rs, &ts, &ref_time);
> >   
> >   	perf_stat__reset_shadow_per_stat(&rt_stat);
> > +	thread_stats_reset(&stat_config);
>
> can't you call in here perf_stat__reset_stats?

If we call perf_stat__reset_stat here, it will reset the evsel->counts, but I 
don't think it's necessary. The counts will be updated in read_counts() soon.

> and if not, I know it's threads related, but new
> and delete functions are:
>
>    runtime_stat_new, runtime_stat_delete
>
> so let's call it runtime_stat_reset and place it next to
> the new/delete functions

Yes, that's good idea. I will create runtime_stat_reset and place it next to
untime_stat_new/runtime_stat_delete.

> other than that it looks ok, thanks
>
> jirka

Thanks!

Thanks
Jin Yao

> >   	read_counters(&rs);
> >   
> >   	if (STAT_RECORD) {
> > --
> > 2.17.1