lkml.org 
[lkml]   [2002]   [Oct]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: [PATCH] pooling NUMA scheduler with initial load balancing
Date
On Friday 11 October 2002 09:47, Erich Focht wrote:
> Hi Andrew,
>
> On Thursday 10 October 2002 19:34, Andrew Theurer wrote:
> > Thanks very much Erich. I did come across another problem here on
> > numa-q. In task_to_steal() there is a divide by cache_decay_ticks, which
> > apparantly is 0 on my system. This may have to do with notsc, but I am
> > not sure. I set cache_decay_ticks to 8, (I cannot boot without using
> > notsc) which is probably not correct, but I can now boot 16 processor
> > numa-q on 2.5.40-mm1 with your patches! I'll get some benchmark results
> > soon.
>
> oops... This is a bug in 2.5-i386. It means that the O(1) scheduler in
> 2.5 doesn't work well either because it doesn't take into account cache
> coolness. I'll post a fix to LKML in a few minutes.

Sorry, I thought the smp_tune_scheduling() call went lost during the
transition to the new cpu boot scheme. But it's there. And the problem
is indeed "notsc". So you'll have to fix it, I can't.

If you set the cache_decay_ticks to something non-zero, you should
_really_ do this for all the scheduler tests, otherwise your measurements
will not be comparable.

Regards,
Erich

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:29    [W:0.452 / U:0.120 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site