Re: [Patch v4 2/6] sched: Add infrastructure to store and update instantaneous thermal pressure

[Vincent Guittot]

On Mon, 4 Nov 2019 at 18:29, Dietmar Eggemann <dietmar.eggemann@xxxxxxx> wrote:
>
> On 01/11/2019 21:57, Thara Gopinath wrote:
> > On 11/01/2019 08:17 AM, Dietmar Eggemann wrote:
> >> On 22.10.19 22:34, Thara Gopinath wrote:
> >>
> >> [...]
> >>
> >>> +/**
> >>> + * trigger_thermal_pressure_average: Trigger the thermal pressure accumulate
> >>> + *                              and average algorithm
> >>> + */
> >>> +void trigger_thermal_pressure_average(struct rq *rq)
> >>> +{
> >>> +   update_thermal_load_avg(rq_clock_task(rq), rq,
> >>> +                           per_cpu(delta_capacity, cpu_of(rq)));
> >>> +}
> >>
> >> Why not call update_thermal_load_avg() directly in fair.c? We do this for all
> >> the other update_foo_load_avg() functions (foo eq. irq, rt_rq, dl_rq ...)
> > thermal.c is going away in next version and I am moving everything to
> > fair.c. So this is taken care of
> >
> >>
> >> You don't have to pass 'u64 now', so you can hide it plus the
> >
> > You still need now.All the update_*_avg apis take now as a parameter.
>
> You do need it for the ___update_load_sum() call inside the
> foo_load_avg() functions. But that doesn't mean you have to pass it into
> foo_load_avg(). Look at update_irq_load_avg() for example. We don't pass
> rq->clock as now in there.

update_irq_load_avg is the exception but having now as a parameter is
the default behavior that update_thermal_load_avg have to follow

>
> -int update_thermal_load_avg(u64 now, struct rq *rq, u64 capacity)
> +extern int sched_thermal_decay_coeff;
> +
> +int update_thermal_load_avg(struct rq *rq, u64 capacity)
>  {
> +       u64 now = rq_clock_task(rq) >> sched_thermal_decay_coeff;
> +
>         if (___update_load_sum(now, &rq->avg_thermal,
>                                capacity,
>                                capacity,