Re: [PATCH 1/4] sched/topology: Store root domain CPU capacity sum

[Vincent Guittot]

On Tue, 14 Apr 2020 at 17:27, Dietmar Eggemann <dietmar.eggemann@xxxxxxx> wrote:
>
> On 14.04.20 14:45, Quentin Perret wrote:
> > On Wednesday 08 Apr 2020 at 19:03:57 (+0200), Vincent Guittot wrote:
> >> Or we can do the opposite and only use capacity_orig_of()/rq->cpu_capacity_orig.
> >>
> >> Is there a case where the max cpu capacity changes over time ? So I
> >> would prefer to use cpu_capacity_orig which is a field of scheduler
> >> instead of always calling an external arch specific function
> >
> > Note however that using arch_scale_cpu_capacity() would be more
> > efficient, especially on non-arm/arm64 systems where it is a
> > compile-time constant.
>
> or essentially all ARCHs not defining it.
>
> >
> > It's probably a matter of personal taste, but I find rq->cpu_capacity_orig
> > superfluous. It wastes space without actually giving you anything no?
> > Anybody remembers why it was introduced in the first place?
>
> v3.18 arm providing arch_scale_cpu_capacity()
> v4.1  introduction of rq->cpu_capacity_orig
> v4.10 arm64 providing arch_scale_cpu_capacity()
> ...
>
> So it's down to the question of 'minimizing the scheduler calls to an
> external arch function' vs 'efficiency'.

Using cpu_capacity_orig will provide more consistency because it will
be set while setting the sched_domain topology