On Tue, 2016-09-06 at 15:42 +0200, Vincent Guittot wrote:
> On 6 September 2016 at 15:07, Mike Galbraith <[email protected]> wrote:
> > On Tue, 2016-09-06 at 15:01 +0200, Vincent Guittot wrote:
> > > Le Monday 05 Sep 2016 à 18:26:53 (+0200), Mike Galbraith a écrit :
> > > > Coming back to this, how about this instead, only increase the group
> > > > imbalance threshold when sd_llc_size == 2. Newer L3 equipped
> > > > processors then aren't affected.
> > > >
> > >
> > > Not sure that all systems with sd_llc_size == 2 wants this behavior.
> > >
> > > Why not adding a sched_feature for changing the 2nd half of the test
> > > for some systems ?
> >
> > Because users won't know, and shouldn't need to know that they need to
> > flip that switch.
>
> fair enough
>
> so how can we detect this specific system configuration ?
>
> sd_llc_size == 2 is not enough IMHO
Seems to me that _is_ the pertinent configuration. If that isn't
enough, I guess we just say too bad about old multiple LLC CPUs.
-Mike
