[lkml]   [2011]   [Nov]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [patch 1/6] sched: use rt.nr_cpus_allowed to recover select_task_rq() cycles
On Fri, 2011-11-18 at 16:35 +0100, Peter Zijlstra wrote:
> On Fri, 2011-11-18 at 16:17 +0100, Mike Galbraith wrote:
> > kernel/sched_fair.c | 3 +++
> > kernel/sched_rt.c | 3 +++
> I hate to be the bringer of bad news, but you're going to have to respin
> all this, these files no longer exist ;-)

No biggie, will do.

I'm just showing what was in the kernel to go with the numbers.

The nohz thing I dislike, but the problem isn't going away. Don't like
the select_idle_sibling() ratelimit as is either, but I don't have a
better solution to this two faced little bugger.

> Please update your tip/master.

Yeah, these were master branch. Was twiddling stable as usual first, as
you then don't have to worry about stuff changing (especially unrelated
stuff that still affects you) day to day, and munging results.

> Also, < 2 ? nr_cpus_allowed being 0 is somewhat of a problem, so the
> only remaining option is == 1.

True, but see select_task_rq_rt() :)


 \ /
  Last update: 2011-11-18 18:37    [W:0.116 / U:0.504 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site