Re: [PATCH v3 3/7] sched: rotate the cpu search window for better spread

[Subhra Mazumdar]

On 6/28/19 4:54 AM, Srikar Dronamraju wrote:
> * subhra mazumdar <subhra.mazumdar@xxxxxxxxxx> [2019-06-26 18:29:15]:
>
> > Rotate the cpu search window for better spread of threads. This will ensure
> > an idle cpu will quickly be found if one exists.
> While rotating the cpu search window is good, not sure if this can find a
> idle cpu quickly. The probability of finding an idle cpu still should remain
> the same. No?
>
> > Signed-off-by: subhra mazumdar <subhra.mazumdar@xxxxxxxxxx>
> > ---
> >   kernel/sched/fair.c | 10 ++++++++--
> >   1 file changed, 8 insertions(+), 2 deletions(-)
> >
> > @@ -6219,9 +6219,15 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t
> >   		}
> >   	}
> >   
> > +	if (per_cpu(next_cpu, target) != -1)
> > +		target_tmp = per_cpu(next_cpu, target);
> > +	else
> > +		target_tmp = target;
> > +
> >   	time = local_clock();
> >   
> > -	for_each_cpu_wrap(cpu, sched_domain_span(sd), target) {
> > +	for_each_cpu_wrap(cpu, sched_domain_span(sd), target_tmp) {
> > +		per_cpu(next_cpu, target) = cpu;
> Shouldn't this assignment be outside the for loop.
> With the current code,
> 1. We keep reassigning multiple times.
> 2. The last assignment happes for idle_cpu and sometimes the
> assignment is for non-idle cpu.
We want the last assignment irrespective of it was an idle cpu or not since
in both cases we want to track the boundary of search.

Thanks,
Subhra