[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAM9d7cj87QqoVNGJm5TJ1YS_QRoZpukQD8GbDg5VqJWc8euWFA@mail.gmail.com>
Date: Mon, 12 Oct 2020 14:04:48 +0900
From: Namhyung Kim <namhyung@...nel.org>
To: Jiri Olsa <jolsa@...hat.com>
Cc: Arnaldo Carvalho de Melo <acme@...nel.org>,
Ingo Molnar <mingo@...nel.org>,
Peter Zijlstra <a.p.zijlstra@...llo.nl>,
Mark Rutland <mark.rutland@....com>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
LKML <linux-kernel@...r.kernel.org>,
Stephane Eranian <eranian@...gle.com>,
Ian Rogers <irogers@...gle.com>,
Al Grant <al.grant@...s.arm.com>,
Adrian Hunter <adrian.hunter@...el.com>
Subject: Re: [PATCH] perf inject: Flush ordered events on FINISHED_ROUND
Hi Jiri,
On Thu, Oct 8, 2020 at 6:07 PM Jiri Olsa <jolsa@...hat.com> wrote:
>
> On Tue, Oct 06, 2020 at 02:40:32PM +0900, namhyung@...nel.org wrote:
> > This is the perf stat result:
> >
> > * Before
> >
> > 7,167,414,019 L1-dcache-loads
> > 337,471,761 L1-dcache-read-misses # 4.71% of all L1-dcache hits
> >
> > 11.011224671 seconds time elapsed
> >
> >
> > * After
> >
> > 7,075,556,792 L1-dcache-loads
> > 771,810,388 L1-dcache-read-misses # 10.91% of all L1-dcache hits
> >
> > 17.015901863 seconds time elapsed
> >
> >
> > Hmm.. it's a memory & time trade-off then. Maybe we need a switch to
> > select which one?
>
> I'd keep the faster one ;-) so the one before
Yeah, it's hard to argue when it slows things down much.
But I'd like to note that memory consumption also can be a
big issue in some environments. I'll bring this back later..
Thanks
Namhyung
Powered by blists - more mailing lists