[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <0d7cbbb0-9c35-c6c8-839b-2f12d6ed2808@linux.intel.com>
Date: Fri, 8 May 2020 10:45:25 +0800
From: "Jin, Yao" <yao.jin@...ux.intel.com>
To: Jiri Olsa <jolsa@...hat.com>
Cc: acme@...nel.org, jolsa@...nel.org, peterz@...radead.org,
mingo@...hat.com, alexander.shishkin@...ux.intel.com,
Linux-kernel@...r.kernel.org, ak@...ux.intel.com,
kan.liang@...el.com, yao.jin@...el.com
Subject: Re: [PATCH v3 2/4] perf counts: Reset prev_raw_counts counts
Hi Jiri,
On 5/7/2020 11:19 PM, Jiri Olsa wrote:
> On Thu, May 07, 2020 at 02:58:20PM +0800, Jin Yao wrote:
>> The evsel->prev_raw_counts is updated in perf_evsel__compute_deltas:
>>
>> perf_evsel__compute_deltas()
>> {
>> tmp = *perf_counts(evsel->prev_raw_counts, cpu, thread);
>> *perf_counts(evsel->prev_raw_counts, cpu, thread) = *count;
>> }
>>
>> When we want to reset the evsel->prev_raw_counts in
>> perf_evsel__reset_prev_raw_counts, zeroing the aggr is not enough,
>> we need to reset the counts too.
>>
>> Signed-off-by: Jin Yao <yao.jin@...ux.intel.com>
>> ---
>> tools/perf/util/counts.c | 5 +++++
>> tools/perf/util/counts.h | 2 ++
>> tools/perf/util/stat.c | 1 +
>> 3 files changed, 8 insertions(+)
>>
>> diff --git a/tools/perf/util/counts.c b/tools/perf/util/counts.c
>> index f94e1a23dad6..af3bf36f7c63 100644
>> --- a/tools/perf/util/counts.c
>> +++ b/tools/perf/util/counts.c
>> @@ -64,3 +64,8 @@ void perf_evsel__free_counts(struct evsel *evsel)
>> perf_counts__delete(evsel->counts);
>> evsel->counts = NULL;
>> }
>> +
>> +void perf_evsel__reset_raw_counts(struct evsel *evsel)
>> +{
>> + perf_counts__reset(evsel->prev_raw_counts);
>> +}
>> diff --git a/tools/perf/util/counts.h b/tools/perf/util/counts.h
>> index 92196df4945f..15bb9acb7cb0 100644
>> --- a/tools/perf/util/counts.h
>> +++ b/tools/perf/util/counts.h
>> @@ -42,4 +42,6 @@ void perf_evsel__reset_counts(struct evsel *evsel);
>> int perf_evsel__alloc_counts(struct evsel *evsel, int ncpus, int nthreads);
>> void perf_evsel__free_counts(struct evsel *evsel);
>>
>> +void perf_evsel__reset_raw_counts(struct evsel *evsel);
>> +
>> #endif /* __PERF_COUNTS_H */
>> diff --git a/tools/perf/util/stat.c b/tools/perf/util/stat.c
>> index 774468341851..89e541564ed5 100644
>> --- a/tools/perf/util/stat.c
>> +++ b/tools/perf/util/stat.c
>> @@ -168,6 +168,7 @@ static void perf_evsel__reset_prev_raw_counts(struct evsel *evsel)
>> evsel->prev_raw_counts->aggr.val = 0;
>> evsel->prev_raw_counts->aggr.ena = 0;
>> evsel->prev_raw_counts->aggr.run = 0;
>> + perf_evsel__reset_raw_counts(evsel);
>
> that seems needed, but we have it scathered all over the place,
> could you centralize the reset? the way I see it the perf_counts__reset
> should zero all the members of struct perf_counts.. so also
> the aggr values
>
> it could also check for counts != NULL and we could call it
> instead of:
> perf_evsel__reset_prev_raw_counts
> perf_evsel__reset_raw_counts
> perf_counts__reset
>
> jirka
>
While sometime, we may want to only reset evsel->counts and sometime we just
want to reset the evsel->prev_raw_counts. So maybe we improve
perf_evsel__reset_counts to:
void perf_evsel__reset_counts(struct evsel *evsel, bool counts, bool
prev_raw_counts)
{
if (counts)
perf_counts__reset(evsel->counts);
if (prev_raw_counts && evsel->prev_raw_counts);
perf_counts__reset(evsel->prev_raw_counts);
}
static void perf_counts__reset(struct perf_counts *counts)
{
counts->aggr.val = 0;
counts->aggr.ena = 0;
counts->aggr.run = 0;
xyarray__reset(counts->loaded);
xyarray__reset(counts->values);
}
So we will only keep perf_evsel__reset_counts and remove
perf_evsel__reset_prev_raw_counts.
Thanks
Jin Yao
Powered by blists - more mailing lists