Re: [RFC PATCH 2/6] sched/dl: Capacity-aware migrations

[luca abeni <luca.abeni@santannapisa.it>]

Hi,

On Tue, 7 May 2019 15:10:50 +0100
Quentin Perret <quentin.perret@arm.com> wrote:

> On Monday 06 May 2019 at 06:48:32 (+0200), Luca Abeni wrote:
> >  static inline unsigned long cpu_bw_dl(struct rq *rq)
> >  {
> > -	return (rq->dl.running_bw * SCHED_CAPACITY_SCALE) >>
> > BW_SHIFT;
> > +	unsigned long res;
> > +
> > +	res = (rq->dl.running_bw * SCHED_CAPACITY_SCALE) >>
> > BW_SHIFT; +
> > +	return (res << SCHED_CAPACITY_SHIFT) /
> > +	       arch_scale_cpu_capacity(NULL, rq->cpu);  
> 
> The only user of cpu_bw_dl() is schedutil right ? If yes, we probably
> don't want to scale things here -- schedutil already does this I
> believe.

I think I added this modification because I arrived to the conclusion
that schedutils does not perform this rescaling (at least, I think it
did not perform it when I wrote the patch :)


BTW, while re-reading the patch I do not understand why this change was
added in this specific patch... I suspect it should have gone in a
separate patch.



				Luca

> 
> Thanks,
> Quentin
> 
> >  }
> >  
> >  static inline unsigned long cpu_util_dl(struct rq *rq)
> > -- 
> > 2.20.1
> >