[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAP-5=fUV5yt4c-PrvTFqW6ehuudwkqdO7QuQpvbUBDm6aT6+dQ@mail.gmail.com>
Date: Mon, 10 Mar 2025 16:40:36 -0700
From: Ian Rogers <irogers@...gle.com>
To: Namhyung Kim <namhyung@...nel.org>
Cc: Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...hat.com>,
Arnaldo Carvalho de Melo <acme@...nel.org>, Mark Rutland <mark.rutland@....com>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>, Jiri Olsa <jolsa@...nel.org>,
Adrian Hunter <adrian.hunter@...el.com>, Kan Liang <kan.liang@...ux.intel.com>,
Yicong Yang <yangyicong@...ilicon.com>, James Clark <james.clark@...aro.org>,
"Dr. David Alan Gilbert" <linux@...blig.org>, Levi Yun <yeoreum.yun@....com>, Ze Gao <zegao2021@...il.com>,
Weilin Wang <weilin.wang@...el.com>, Xu Yang <xu.yang_2@....com>,
linux-perf-users@...r.kernel.org, linux-kernel@...r.kernel.org,
Howard Chu <howardchu95@...il.com>
Subject: Re: [PATCH v2 11/11] perf python tracepoint: Switch to using parse_events
On Mon, Mar 10, 2025 at 3:17 PM Namhyung Kim <namhyung@...nel.org> wrote:
>
> On Fri, Feb 28, 2025 at 02:23:08PM -0800, Ian Rogers wrote:
> > Rather than manually configuring an evsel, switch to using
> > parse_events for greater commonality with the rest of the perf code.
> >
> > Reviewed-by: Howard Chu <howardchu95@...il.com>
> > Signed-off-by: Ian Rogers <irogers@...gle.com>
> > ---
> > tools/perf/python/tracepoint.py | 23 +++++++++++------------
> > 1 file changed, 11 insertions(+), 12 deletions(-)
> >
> > diff --git a/tools/perf/python/tracepoint.py b/tools/perf/python/tracepoint.py
> > index bba68a6d4515..38b2b6d11f64 100755
> > --- a/tools/perf/python/tracepoint.py
> > +++ b/tools/perf/python/tracepoint.py
> > @@ -5,24 +5,23 @@
> >
> > import perf
> >
> > -class tracepoint(perf.evsel):
> > - def __init__(self, sys, name):
> > - config = perf.tracepoint(sys, name)
> > - perf.evsel.__init__(self,
> > - type = perf.TYPE_TRACEPOINT,
> > - config = config,
> > - freq = 0, sample_period = 1, wakeup_events = 1,
> > - sample_type = perf.SAMPLE_PERIOD | perf.SAMPLE_TID | perf.SAMPLE_CPU | perf.SAMPLE_RAW | perf.SAMPLE_TIME)
> > -
> > def main():
> > - tp = tracepoint("sched", "sched_switch")
> > cpus = perf.cpu_map()
> > threads = perf.thread_map(-1)
> > + evlist = perf.parse_events("sched:sched_switch", cpus, threads)
> > + # Disable tracking of mmaps and similar that are unnecessary.
> > + for ev in evlist:
> > + ev.tracking = False
> > + # Configure evsels with default record options.
> > + evlist.config()
>
> I think the default option uses frequency of 4000 but tracepoints want
> to use period of 1. Also I'm not sure if it sets the proper sample type
> bits namely PERF_SAMPLE_RAW.
I used trace to ensure they matched. Fwiw, the sample_period for a
tracepoint is set to 1 in evsel__newtp_idx:
https://web.git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/util/evsel.c?h=perf-tools-next#n621
and the evsel__config won't overwrite an already set sample_period:
https://web.git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/util/evsel.c?h=perf-tools-next#n1341
Thanks, Ian
> Thanks,
> Namhyung
>
>
> > + # Simplify the sample_type and read_format of evsels
> > + for ev in evlist:
> > + ev.sample_type = ev.sample_type & ~perf.SAMPLE_IP
> > + ev.read_format = 0
> >
> > - evlist = perf.evlist(cpus, threads)
> > - evlist.add(tp)
> > evlist.open()
> > evlist.mmap()
> > + evlist.enable();
> >
> > while True:
> > evlist.poll(timeout = -1)
> > --
> > 2.48.1.711.g2feabab25a-goog
> >
Powered by blists - more mailing lists