Re: [PATCH 18/18 v2] sched/fair: Take into account slice in EAS

[Kayra Cizmeci]

> When the cost is the same, take into account the slice of a task to try to
> select a CPU where is will run first.

'is'? It gives 'it' vibes more of 'is' vibes... (:>)

>  static bool update_best_cpu(struct energy_cpu_stat *target,
>  			    struct energy_cpu_stat *min,
> -			    int prev, struct sched_domain *sd)
> +			    int prev, struct sched_domain *sd,
> +			    struct task_struct *p)
>  {
> +	unsigned long task_slice = p->se.slice;
> +
> +	/*  Select the one where you can run first */

Again. A typo. An extra blank. I'll normally make a joke but,
I'm tired. ;-)

> +	if (task_slice < get_rq_min_slice(cpu_rq(target->cpu)) &&
> +	    task_slice >= get_rq_min_slice(cpu_rq(min->cpu)))
> +		return true;
> +
> +	/* Favor previous CPU */
>  	if (target->cpu == prev)
>  		return true;
>  	if (min->cpu == prev)
> @@ -9762,7 +9771,7 @@ static int find_energy_efficient_cpu(struct task_struct *p, int prev_cpu)
>  			 */
>  			if (target_perf < min_stat.min_perf)
>  				find_pd_cost(pd->em_pd, target_perf, &target_stat);
> -			else if (!update_best_cpu(&target_stat, &min_stat, prev_cpu, sd))
> +			else if (!update_best_cpu(&target_stat, &min_stat, prev_cpu, sd, p))
>  				continue;
>  
> 			/* Save the new most efficient CPU of the PD */


Also,

Scene:

On the reworked version of feec(), we have this check:

  if (p->on_rq && !p->se.sched_delayed && cpu == prev_cpu)

A little after and outside of this check, we call update_best_cpu().
Now, if this check can get true, on this state:

prev's min_slice cannot be bigger than p's task_slice.

So when target is prev the first check is always false,
and when min->cpu is prev, the second check is always true.

Also also, I have a really important issue abot the patch
series that can finish the whole patch series and end the world
itself...

The issue is...

I need to sleep.

Thanks,
Kayra