lkml.org 
[lkml]   [2001]   [Jul]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: CPU affinity & IPI latency

On 16-Jul-2001 Hubertus Franke wrote:
>
> David, you are preaching to choir.
>
> One can not have it both ways, at least without "#ifdef"s.
> As Mike stated, we made the decision to adhere to current scheduling
> semantics
> purely for the purspose of comparision and increased adaptation chances.
> As shown with the LoadBalancing addition to MQ, there are simple ways to
> relax and completely eliminate the feedback between the queues, if one so
> desires.
>
> As for the solutions you proposed for the "switching problem", namely the
> wakeup
> list. I don't think you want a list here. A list would basically mean that
> you
> would need to walk it and come up with a single decision again. I think
> what
> I proposed, namely a per-CPU reschedule reservation that can be overwritten
> taking
> PROC_CHANGE_PENALTY or some form of it into account, seems a better
> solution.
> Open to discussions...

No, when you're going to decide ( reschedule_idle ) which idle to spin out, you
can inspect the wake list and, based on the content of the list, one can take a
better decision about which idle to wake.
I think that a list, instead of a single task pointer, is a more open solution
that could drive to a more sophisticated choice of the CPU to stock the task to.




- Davide

-
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 12:57    [W:0.039 / U:0.060 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site