Re: [PATCH] sched/fair: Speed-up energy-aware wake-ups

[Pavan Kondeti]

Hi Quentin,

On Fri, Sep 20, 2019 at 11:41:15AM +0200, Quentin Perret wrote:
> Hi Pavan,
> 
> On Friday 20 Sep 2019 at 08:32:15 (+0530), Pavan Kondeti wrote:
> > Earlier, we are not checking the spare capacity for the prev_cpu. Now that the
> > continue statement is removed, prev_cpu could also be the max_spare_cap_cpu.
> > Actually that makes sense. Because there is no reason why we want to select
> > another CPU which has less spare capacity than previous CPU.
> > 
> > Is this behavior intentional?
> 
> The intent was indeed to not compute the energy for another CPU in
> prev_cpu's perf domain if prev_cpu is the one with max spare cap -- it
> is useless to do so since this other CPU cannot 'beat' prev_cpu and
> will never be chosen in the end.

Yes. Selecting the prev_cpu is the correct decision.

> 
> But I did miss that we'd end up computing the energy for prev_cpu
> twice ... Harmless but useless. So yeah, let's optimize that case too :)
> 
> > When prev_cpu == max_spare_cap_cpu, we are evaluating the energy again for the
> > same CPU below. That could have been skipped by returning prev_cpu when
> > prev_cpu == max_spare_cap_cpu.
> 
> Right, something like the patch below ? My test results are still
> looking good with it applied.
> 
> Thanks for the careful review,
> Quentin
> ---
> From 7b8258287f180a2c383ebe397e8129f5f898ffbe Mon Sep 17 00:00:00 2001
> From: Quentin Perret <qperret@xxxxxxxxxxx>
> Date: Fri, 20 Sep 2019 09:07:20 +0100
> Subject: [PATCH] sched/fair: Avoid redundant EAS calculation
> 
> The EAS wake-up path computes the system energy for several CPU
> candidates: the CPU with maximum spare capacity in each performance
> domain, and the prev_cpu. However, if prev_cpu also happens to be the
> CPU with maximum spare capacity in its performance domain, the energy
> calculation is still done twice, unnecessarily.
> 
> Add a condition to filter out this corner case before doing the energy
> calculation.
> 
> Reported-by: Pavan Kondeti <pkondeti@xxxxxxxxxxxxxx>
> Signed-off-by: Quentin Perret <qperret@xxxxxxxxxxx>
> ---
>  kernel/sched/fair.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index d4bbf68c3161..7399382bc291 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -6412,7 +6412,7 @@ static int find_energy_efficient_cpu(struct task_struct *p, int prev_cpu)
>  		}
>  
>  		/* Evaluate the energy impact of using this CPU. */
> -		if (max_spare_cap_cpu >= 0) {
> +		if (max_spare_cap_cpu >= 0 && max_spare_cap_cpu != prev_cpu) {
>  			cur_delta = compute_energy(p, max_spare_cap_cpu, pd);
>  			cur_delta -= base_energy_pd;
>  			if (cur_delta < best_delta) {
> -- 
> 2.22.1
> 

+1. Looks good to me.

-- 
Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.