lkml.org 
[lkml]   [2010]   [Jan]   [23]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 00/12] perf lock: New subcommand "perf lock", for analyzing lock statistics

* Peter Zijlstra <peterz@infradead.org> wrote:

> On Fri, 2010-01-22 at 17:48 +0100, Peter Zijlstra wrote:
> > On Fri, 2010-01-22 at 11:42 -0500, Jason Baron wrote:
> > > On Fri, Jan 22, 2010 at 10:39:01PM +0900, Hitoshi Mitake wrote:
> > > > Adding new subcommand "perf lock" to perf.
> > > >
> > >
> > > looks like a really intresting patch! Does this patch mean that we can
> > > potentially remove CONFIG_LOCK_STAT, now that the functionality is
> > > available via 'perf'? Another desire for me has been to able to run
> > > lockstat on production box. This has not been possible before without
> > > incurring the performance penalty of lockdep. however, with 'perf' tools
> > > ability to be turned on an off, perhaps there is (or we can build) a
> > > mode for lockdep that is lower overhead for this usecase?
> >
> > boot with: prove_locking=0
> >
> > This patch still needs all the hooks and bits from lockdep, no way you
> > want all that in a prodction kernel.
>
> Also, I suspect lockstat to still be lots faster, since it doesn't need to
> transfer gobs of data but builds the histogram in-kernel.
>
> Having the ability to do in-kernel histrograms with perf is something I've
> thought about but not yet managed to think of a nice way of doing it.

Yeah, that has come up in the past and it would be excellent to have. That
would allow the removal of the special-purpose lockstat /proc output.

Ingo


\
 
 \ /
  Last update: 2010-01-23 08:01    [W:0.084 / U:4.664 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site