lkml.org 
[lkml]   [2009]   [Jun]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH] perf stat: fix hardcoded nr_counter
From
Date
On Wed, 2009-06-24 at 10:58 +0200, Ingo Molnar wrote:
> * Jaswinder Singh Rajput <jaswinder@kernel.org> wrote:
>
> > nr_counter should be based on number of default_attrs entries.
>
> The problem is, this will cause a _lot_ of counters to be used by
> default, overloading the PMU quite significantly and reducing
> precision. The number of hw counters is chosen intentionally the way
> it is now, so that on common hardware there's no over-commit.
>
> As i suggested before, your change does make sense if introduced as
> extended event specifications to 'perf stat':
>
> -e all
> -e all-sw-counters
> -e all-hw-counters
>
> with regex wildcard support too.
>

Ok I will support it now.

> Instead, you chose to ignore my feedback and wasted my time by
> re-sending the same broken patch. If you do such things i'll have to
> start ignoring you - it is very ineffective for me to repeat every
> argument and suggestion two or more times and no other contributor
> does this ...
>

oops, I am sorry my AMD box was not able to boot because of the
regression the only machine which supports performance counter. I was
doing git-bisect on AMD box in the mean time I prepare this patch on old
Intel box.

> > Remove dead code and some code alignment.
>
> Could you please send this separately?
>

Send [PATCH -tip] perf stat: remove dead code

Thanks,
--
JSR



\
 
 \ /
  Last update: 2009-06-24 15:01    [W:0.295 / U:0.000 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site