lkml.org 
[lkml]   [2013]   [Oct]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 6/8] perf tools: Add new comm infrastructure
Em Mon, Oct 28, 2013 at 05:05:30PM +0100, Frederic Weisbecker escreveu:
> On Mon, Oct 28, 2013 at 11:29:12AM -0300, Arnaldo Carvalho de Melo wrote:
> > Em Mon, Oct 28, 2013 at 09:43:09AM -0300, Arnaldo Carvalho de Melo escreveu:
> > > Em Mon, Oct 28, 2013 at 11:12:50AM +0100, Frederic Weisbecker escreveu:
> > > > On Mon, Oct 28, 2013 at 06:15:26PM +0900, Namhyung Kim wrote:
> > > > > On Mon, 28 Oct 2013 10:09:46 +0100, Frederic Weisbecker wrote:
> > > > > > Ah cool! Could you please remind me the name of that branch so that I
> > > > > > can do some tests and work on top of it?
> >
> > > > > git://git.kernel.org/pub/scm/linux/kernel/git/namhyung/linux-perf.git
> > > > > perf/callchain-v5
> >
> > > > Hmm, I tried to rebase on top of tip:perf/core and git is getting confused I guess with
> > > > which is applied and which is out of tree. There have been a lot of changes since
> > > > then and thus quite some conflicts.
> > > >
> > > > If you don't mind, I would love if you rebase against latest tip:perf/core as you
> > > > know better than me what has been applied and what hasn't.
> >
> > > I'll try to do it now.
> >
> > Can you please try with:
> >
> > git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git
> > perf/comm
> >
> > And share your results? I fixed up everything wrt recent flux in my
> > perf/core branch, did just basic testing here, but so far it looks ok
> > (and fast!).
>
> Ok I tried perf report, perf script and perf top and it looks good.
> In fact it looks better than the tip tree. I fixes some bugs that I can

tip.tip has the borked disabling of MMAP2 that may explain some of the
problems you described here

> see in tip:/perf/core:
>
> $ perf record perf bench sched messaging
> $ perf report --stdio -s comm
> write failure on standard output: Bad file descriptor
>
> Not sure where that comes from, but it's fixed in your tree. May be that's
> on the fixes before the comm patches in your tree.
>
> Also it differentiate between pre-exec and post-fork events, which looks
> more precise (and it fixes some comm mangling as well):
>
>
> Before:
>
> # Overhead Command
> # ........ ...............
> #
> 100.00% sched-me

I noticed the above, just on the --stdio tho, checking why this is so...
Can you try without --stdio, even using --tui?

- Arnaldo

>
> After:
>
> # Overhead Command
> # ........ ...............
> #
> 99.89% sched-messaging
> 0.11% perf
>
>
> Thanks!


\
 
 \ /
  Last update: 2013-10-28 18:21    [W:0.099 / U:2.392 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site