Re: [PATCH 2/4] sched/fair: Move avg_scan_cost calculations under SIS_PROP

[Mel Gorman]

On Thu, Dec 10, 2020 at 01:18:05PM +0800, Li, Aubrey wrote:
> > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > index ac7b34e7372b..5c41875aec23 100644
> > --- a/kernel/sched/fair.c
> > +++ b/kernel/sched/fair.c
> > @@ -6153,6 +6153,8 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t
> >  	if (!this_sd)
> >  		return -1;
> >  
> > +	cpumask_and(cpus, sched_domain_span(sd), p->cpus_ptr);
> > +
> >  	if (sched_feat(SIS_PROP)) {
> >  		u64 avg_cost, avg_idle, span_avg;
> >  
> > @@ -6168,11 +6170,9 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t
> >  			nr = div_u64(span_avg, avg_cost);
> >  		else
> >  			nr = 4;
> > -	}
> > -
> > -	time = cpu_clock(this);
> >  
> > -	cpumask_and(cpus, sched_domain_span(sd), p->cpus_ptr);
> > +		time = cpu_clock(this);
> > +	}
> >  
> >  	for_each_cpu_wrap(cpu, cpus, target) {
> >  		if (!--nr)
> >			return -1;
> 
> I thought about this again and here seems not to be consistent:
> - even if nr reduces to 0, shouldn't avg_scan_cost be updated as well before return -1?

You're right, but it's outside the scope
of this patch. I noted that this was a problem in
lore.kernel.org/r/lore.kernel.org/r/20201203141124.7391-8-mgorman@xxxxxxxxxxxxxxxxxxx
It's neither a consistent win or loss to always account for it and so
was dropped for this series to keep the number of controversial patches
to a minimum.

> - if avg_scan_cost is not updated because nr is throttled, the first 
> 	time = cpu_clock(this);
>   can be optimized. As nr is calculated and we already know which of the weight of cpumask and nr is greater.
> 

That is also outside the scope of this patch. To do that, cpumask_weight()
would have to be calculated but it's likely to be a net loss. Even under
light load, nr will be smaller than the domain weight incurring both the
cost of cpumask_weight and the clock read in the common case.

-- 
Mel Gorman
SUSE Labs