Re: [PATCH 3/6] sched: Report the different kinds of imbalances in /proc/schedstat

[Sapkal, Swapnil]

Hello Peter,

Thanks for the review.

On 12/18/2024 5:10 PM, Peter Zijlstra wrote:
> On Wed, Dec 18, 2024 at 04:36:26AM +0000, Swapnil Sapkal wrote:
>
> > +static void update_lb_imbalance_stat(struct lb_env *env, struct sched_domain *sd,
> > +				     enum cpu_idle_type idle)
> > +{
> > +	switch (env->migration_type) {
> > +	case migrate_load:
> > +		schedstat_add(sd->lb_imbalance_load[idle], env->imbalance);
> > +		break;
> > +	case migrate_util:
> > +		schedstat_add(sd->lb_imbalance_util[idle], env->imbalance);
> > +		break;
> > +	case migrate_task:
> > +		schedstat_add(sd->lb_imbalance_task[idle], env->imbalance);
> > +		break;
> > +	case migrate_misfit:
> > +		schedstat_add(sd->lb_imbalance_misfit[idle], env->imbalance);
> > +		break;
> > +	}
> > +}
>
>
> Can you please write that like:
>
> 	if (!schedstat_enabled())
> 		return;
>
> 	switch () {
> 	case ...
> 		__schedstat_add();
> 	}
>
> It makes no sense to have 4 copies of schedstat_enabled() inside the
> switch statement -- esp. since afaik the compilers aren't able to CSE
> static keys :/

This makes sense. I will update this change in v2.

--
Thanks and Regards,
Swapnil