[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <Z4FjMwlU6F2RSO4c@x1>
Date: Fri, 10 Jan 2025 15:13:07 -0300
From: Arnaldo Carvalho de Melo <acme@...nel.org>
To: Gabriele Monaco <gmonaco@...hat.com>
Cc: Namhyung Kim <namhyung@...nel.org>, Ian Rogers <irogers@...gle.com>,
Kan Liang <kan.liang@...ux.intel.com>, Jiri Olsa <jolsa@...nel.org>,
Adrian Hunter <adrian.hunter@...el.com>,
Peter Zijlstra <peterz@...radead.org>,
Ingo Molnar <mingo@...nel.org>, LKML <linux-kernel@...r.kernel.org>,
linux-perf-users@...r.kernel.org
Subject: Re: [PATCH 1/2] perf ftrace: Check min/max latency only with bucket
range
On Fri, Jan 10, 2025 at 04:11:05PM +0100, Gabriele Monaco wrote:
>
> On Fri, 2025-01-10 at 11:03 -0300, Arnaldo Carvalho de Melo wrote:
> > On Fri, Jan 10, 2025 at 10:09:14AM +0000, Gabriele Monaco wrote:
> > > 2025-01-10T00:46:49Z Namhyung Kim <namhyung@...nel.org>:
> > >
> > > > On Thu, Jan 09, 2025 at 08:53:02AM +0100, Gabriele Monaco wrote:
> > > > > On Wed, 2025-01-08 at 13:00 -0800, Namhyung Kim wrote:
> > > > > > It's an optional feature and remains 0 when bucket range is
> > > > > > not
> > > > > > given.
> > > > > > And it makes the histogram goes to the last entry always
> > > > > > because any
> > > > > > latency (num) is greater than or equal to 0.
> > > > >
> > > > > Thanks Namhyung for fixing this, something definitely slipped
> > > > > while
> > > > > testing..
> > > > >
> > > > > I confirm your patches work well also when the bucket range is
> > > > > provided but the
> > > > > min latency isn't.
> > > > >
> > > > > I'm wondering if it would be cleaner to propagate your changes
> > > > > (using
> > > > > min/max latency only if bucket_range is provided) also to
> > > > > make_histogram. That function currently works since we assume
> > > > > min_latency to be always 0, which is the case but probably not
> > > > > considering it altogether would look a bit better and prevent
> > > > > some
> > > > > headache in the future.
> > > >
> > > > It looks good. One thing I concern is 'num += min_latency'
> > > > before
> > > > do_inc. I put it there to make it symmetric to 'num -=
> > > > min_latency'
> > > > so it should go to inside the block too.
> > > >
> > > > Or you could factor it out as a function like 'i =
> > > > get_bucket_index(num)'
> > > > so that it can keep the original num for the stats.
> > > >
> > >
> > > Good point, I can have a deeper look at that. But I'd say it can
> > > come as a cleanup patch later.
> > > I have a couple more changes in mind and this would be no longer
> > > related to your changes.
> >
> > I'm tentatively taking this as an:
> >
> > Acked-by: Gabriele Monaco <gmonaco@...hat.com>
> >
> > But it would be great to have it as a Reviewed-by and perhaps a
> > Tested-by, provided explicitely in response to this thread, ok?
> >
> > Thanks,
> >
> > - Arnaldo
> >
>
> I did test after applying both patches, went through the code and
> confirm my test worked as expected, and I confirm the issue is there
> before patching. I tested also in between, so feel free to add to both
> patches:
>
> Reviewed-by: Gabriele Monaco <gmonaco@...hat.com>
> Tested-by: Gabriele Monaco <gmonaco@...hat.com>
Thanks, added.
> (I'm assuming you are referring to 1/2 and 2/2 and not the little patch
Yeah, Namhyung's 1/2 and 2/2 patches, these:
https://lore.kernel.org/r/20250108210015.1188531-1-namhyung@kernel.org
https://lore.kernel.org/r/20250108210015.1188531-2-namhyung@kernel.org
- Arnaldo
> I sent in the first answer)
>
> Thanks,
> Gabriele
Powered by blists - more mailing lists