Re: [PATCH 2/4] sched/fair: Do not replace recent_used_cpu with the new target

[Mel Gorman]

On Tue, Dec 08, 2020 at 10:57:29AM +0100, Dietmar Eggemann wrote:
> On 07/12/2020 10:15, Mel Gorman wrote:
> > After select_idle_sibling, p->recent_used_cpu is set to the
> > new target. However on the next wakeup, prev will be the same as
> 
> I'm confused here. Isn't current->recent_used_cpu set to 'cpu =
> smp_processor_id()' after sis()? Looking at v5.10-rc6.

If you are referring to this;

               if (want_affine)
                       current->recent_used_cpu = cpu;

then it gets removed by the path. That replaces recent_used_cpu with the
wakers CPU which still works but the hit rate is lower.

> 
> [...]
> 
> > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > index 23934dbac635..01b38fc17bca 100644
> > --- a/kernel/sched/fair.c
> > +++ b/kernel/sched/fair.c
> > @@ -6274,6 +6274,7 @@ static int select_idle_sibling(struct task_struct *p, int prev, int target)
> >  
> >  	/* Check a recently used CPU as a potential idle candidate: */
> >  	recent_used_cpu = p->recent_used_cpu;
> > +	p->recent_used_cpu = prev;
> >  	if (recent_used_cpu != prev &&
> >  	    recent_used_cpu != target &&
> >  	    cpus_share_cache(recent_used_cpu, target) &&
> 
> p->recent_used_cpu is already set to prev in this if condition.
> 

That can be removed as redundant, I'll fix it.

-- 
Mel Gorman
SUSE Labs