On Wed, Jul 31, 2013 at 06:39:03PM +0200, Peter Zijlstra wrote:
> On Wed, Jul 31, 2013 at 05:11:41PM +0100, Mel Gorman wrote:
> > RSS was another option it felt as arbitrary as a plain delay.
>
> Right, it would avoid 'small' programs getting scanning done with the
> rationale that their cost
On Wed, Jul 31, 2013 at 06:39:03PM +0200, Peter Zijlstra wrote:
On Wed, Jul 31, 2013 at 05:11:41PM +0100, Mel Gorman wrote:
RSS was another option it felt as arbitrary as a plain delay.
Right, it would avoid 'small' programs getting scanning done with the
rationale that their cost isn't
On Wed, Jul 31, 2013 at 05:11:41PM +0100, Mel Gorman wrote:
> RSS was another option it felt as arbitrary as a plain delay.
Right, it would avoid 'small' programs getting scanning done with the
rationale that their cost isn't that large since they don't have much
memory to begin with.
The same
On Wed, Jul 31, 2013 at 05:30:18PM +0200, Peter Zijlstra wrote:
> On Wed, Jul 31, 2013 at 12:57:19PM +0100, Mel Gorman wrote:
>
> > > Right, so what Ingo did is have the scan rate depend on the convergence.
> > > What exactly did you dislike about that?
> > >
> >
> > It depended entirely on
On Wed, Jul 31, 2013 at 12:57:19PM +0100, Mel Gorman wrote:
> > Right, so what Ingo did is have the scan rate depend on the convergence.
> > What exactly did you dislike about that?
> >
>
> It depended entirely on properly detecting if we are converged or not. As
> things like false share
On Wed, Jul 31, 2013 at 12:48:14PM +0200, Peter Zijlstra wrote:
> On Wed, Jul 31, 2013 at 11:30:52AM +0100, Mel Gorman wrote:
> > I'm not sure I understand your point. The scan rate is decreased again if
> > the page is found to be properly placed in the future. It's in the next
> > hunk you
On Wed, Jul 31, 2013 at 11:30:52AM +0100, Mel Gorman wrote:
> I'm not sure I understand your point. The scan rate is decreased again if
> the page is found to be properly placed in the future. It's in the next
> hunk you modify although the periodically reset comment is now out of date.
Yeah its
On Thu, Jul 25, 2013 at 12:36:20PM +0200, Peter Zijlstra wrote:
>
> Subject: sched, numa: Break stuff..
> From: Peter Zijlstra
> Date: Tue Jul 23 14:58:41 CEST 2013
>
> This patch is mostly a comment in code. I don't believe the current
> scan period adjustment scheme can work properly nor do I
On Thu, Jul 25, 2013 at 12:36:20PM +0200, Peter Zijlstra wrote:
Subject: sched, numa: Break stuff..
From: Peter Zijlstra pet...@infradead.org
Date: Tue Jul 23 14:58:41 CEST 2013
This patch is mostly a comment in code. I don't believe the current
scan period adjustment scheme can work
On Wed, Jul 31, 2013 at 11:30:52AM +0100, Mel Gorman wrote:
I'm not sure I understand your point. The scan rate is decreased again if
the page is found to be properly placed in the future. It's in the next
hunk you modify although the periodically reset comment is now out of date.
Yeah its
On Wed, Jul 31, 2013 at 12:48:14PM +0200, Peter Zijlstra wrote:
On Wed, Jul 31, 2013 at 11:30:52AM +0100, Mel Gorman wrote:
I'm not sure I understand your point. The scan rate is decreased again if
the page is found to be properly placed in the future. It's in the next
hunk you modify
On Wed, Jul 31, 2013 at 12:57:19PM +0100, Mel Gorman wrote:
Right, so what Ingo did is have the scan rate depend on the convergence.
What exactly did you dislike about that?
It depended entirely on properly detecting if we are converged or not. As
things like false share detection
On Wed, Jul 31, 2013 at 05:30:18PM +0200, Peter Zijlstra wrote:
On Wed, Jul 31, 2013 at 12:57:19PM +0100, Mel Gorman wrote:
Right, so what Ingo did is have the scan rate depend on the convergence.
What exactly did you dislike about that?
It depended entirely on properly detecting
On Wed, Jul 31, 2013 at 05:11:41PM +0100, Mel Gorman wrote:
RSS was another option it felt as arbitrary as a plain delay.
Right, it would avoid 'small' programs getting scanning done with the
rationale that their cost isn't that large since they don't have much
memory to begin with.
The same
Subject: sched, numa: Break stuff..
From: Peter Zijlstra
Date: Tue Jul 23 14:58:41 CEST 2013
This patch is mostly a comment in code. I don't believe the current
scan period adjustment scheme can work properly nor do I think it a
good idea to ratelimit the numa faults as a whole based on
Subject: sched, numa: Break stuff..
From: Peter Zijlstra pet...@infradead.org
Date: Tue Jul 23 14:58:41 CEST 2013
This patch is mostly a comment in code. I don't believe the current
scan period adjustment scheme can work properly nor do I think it a
good idea to ratelimit the numa faults as a
Summary:
Seeing improvement on a 2 node when running autonumabenchmark .
But seeing regression for specjbb for the same box.
Also seeing huge regression when running autonumabenchmark
both on 4 node and 8 node box.
Below is the autonuma benchmark results on a 2 node machine.
Autonuma
Summary:
Seeing improvement on a 2 node when running autonumabenchmark .
But seeing regression for specjbb for the same box.
Also seeing huge regression when running autonumabenchmark
both on 4 node and 8 node box.
Below is the autonuma benchmark results on a 2 node machine.
Autonuma
On Mon, Jul 15, 2013 at 04:20:02PM +0100, Mel Gorman wrote:
>
> specjbb
> 3.9.0 3.9.0 3.9.0
> 3.9.0
> vanilla accountload-v5 retrymigrate-v5
>swaptasks-v5
> TPut 1
This continues to build on the previous feedback and further testing and
I'm hoping this can be finalised relatively soon. False sharing is still
a major problem but I still think it deserves its own series. Minimally I
think the fact that we are now scanning shared pages without much additional
This continues to build on the previous feedback and further testing and
I'm hoping this can be finalised relatively soon. False sharing is still
a major problem but I still think it deserves its own series. Minimally I
think the fact that we are now scanning shared pages without much additional
On Mon, Jul 15, 2013 at 04:20:02PM +0100, Mel Gorman wrote:
specjbb
3.9.0 3.9.0 3.9.0
3.9.0
vanilla accountload-v5 retrymigrate-v5
swaptasks-v5
TPut 1 24474.00 (
22 matches
Mail list logo