Re: [PATCH v2 3/6] lib/group_cpus: optimize inner loop in grp_spread_init_one()

[Yury Norov <yury.norov@gmail.com>]

On Thu, Dec 07, 2023 at 01:45:21PM -0800, Andrew Morton wrote:
> On Thu,  7 Dec 2023 12:38:57 -0800 Yury Norov <yury.norov@gmail.com> wrote:
> 
> > The loop starts from the beginning every time we switch to the next
> > sibling mask. This is the Schlemiel the Painter's style of coding
> > because we know for sure that nmsk is clear up to current CPU, and we
> > can just continue from the next CPU.
> > 
> > Also, we can do it nicer if leverage the dedicated for_each() iterator.
> > 
> > --- a/lib/group_cpus.c
> > +++ b/lib/group_cpus.c
> > @@ -30,13 +30,13 @@ static void grp_spread_init_one(struct cpumask *irqmsk, struct cpumask *nmsk,
> >  
> >  		/* If the cpu has siblings, use them first */
> >  		siblmsk = topology_sibling_cpumask(cpu);
> > -		for (sibl = -1; cpus_per_grp > 0; ) {
> > -			sibl = cpumask_next(sibl, siblmsk);
> > -			if (sibl >= nr_cpu_ids)
> > -				break;
> 
> I assume this test goes away because the iterator takes care of it?

Yes, correct.

> 
> > +		sibl = cpu + 1;
> > +
> > +		for_each_cpu_and_from(sibl, siblmsk, nmsk) {
> >  			__cpumask_clear_cpu(sibl, nmsk);
> >  			__cpumask_set_cpu(sibl, irqmsk);
> > -			cpus_per_grp--;
> > +			if (cpus_per_grp-- == 0)
> > +				return;
> >  		}
> >  	}
> >  }