[PATCH] sched/fair: Remove the cost of a redundant cpumask_next_wrap in select_idle_cpu

[PATCH] sched/fair: Remove the cost of a redundant cpumask_next_wrap in select_idle_cpu

[Barry Song]

From: Barry Song <song.bao.hua@hisilicon.com>

This patch keeps the same scanning amount, but drops a redundant loop
of cpumask_next_wrap.
The original code did for_each_cpu_wrap(cpu, cpus, target + 1), then
checked --nr; this patch does --nr before doing the next loop, thus,
it can remove a cpumask_next_wrap() which costs a little bit.

Signed-off-by: Barry Song <song.bao.hua@hisilicon.com>
---
 kernel/sched/fair.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index ff69f24..e2fb3e0 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -6298,9 +6298,9 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, bool
 
 		span_avg = sd->span_weight * avg_idle;
 		if (span_avg > 4*avg_cost)
-			nr = div_u64(span_avg, avg_cost);
+			nr = div_u64(span_avg, avg_cost) - 1;
 		else
-			nr = 4;
+			nr = 3;
 
 		time = cpu_clock(this);
 	}
@@ -6312,11 +6312,11 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, bool
 				return i;
 
 		} else {
-			if (!--nr)
-				return -1;
 			idle_cpu = __select_idle_cpu(cpu, p);
 			if ((unsigned int)idle_cpu < nr_cpumask_bits)
 				break;
+			if (!--nr)
+				return -1;
 		}
 	}
 
-- 
1.8.3.1

Re: [PATCH] sched/fair: Remove the cost of a redundant cpumask_next_wrap in select_idle_cpu

[Peter Zijlstra]

On Tue, Nov 23, 2021 at 07:22:29PM +0800, Barry Song wrote:
> From: Barry Song <song.bao.hua@hisilicon.com>
> 
> This patch keeps the same scanning amount, but drops a redundant loop
> of cpumask_next_wrap.
> The original code did for_each_cpu_wrap(cpu, cpus, target + 1), then
> checked --nr; this patch does --nr before doing the next loop, thus,
> it can remove a cpumask_next_wrap() which costs a little bit.
> 
> Signed-off-by: Barry Song <song.bao.hua@hisilicon.com>
> ---
>  kernel/sched/fair.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index ff69f24..e2fb3e0 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -6298,9 +6298,9 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, bool
>  
>  		span_avg = sd->span_weight * avg_idle;
>  		if (span_avg > 4*avg_cost)
> -			nr = div_u64(span_avg, avg_cost);
> +			nr = div_u64(span_avg, avg_cost) - 1;
>  		else
> -			nr = 4;
> +			nr = 3;
>  
>  		time = cpu_clock(this);
>  	}
> @@ -6312,11 +6312,11 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, bool
>  				return i;
>  
>  		} else {
> -			if (!--nr)
> -				return -1;
>  			idle_cpu = __select_idle_cpu(cpu, p);
>  			if ((unsigned int)idle_cpu < nr_cpumask_bits)
>  				break;
> +			if (!--nr)
> +				return -1;
>  		}
>  	}

That's just confusing code. Isn't it much clearer to write the whole
thing like so ?

	nr--;
	for_each_cpu_wrap(cpu, cpus, target+1) {
		...
		if (!nr--)
			return -1;
	}

Re: [PATCH] sched/fair: Remove the cost of a redundant cpumask_next_wrap in select_idle_cpu

[Barry Song]

On Wed, Nov 24, 2021 at 10:07 AM Peter Zijlstra <peterz@infradead.org> wrote:
>
> On Tue, Nov 23, 2021 at 07:22:29PM +0800, Barry Song wrote:
> > From: Barry Song <song.bao.hua@hisilicon.com>
> >
> > This patch keeps the same scanning amount, but drops a redundant loop
> > of cpumask_next_wrap.
> > The original code did for_each_cpu_wrap(cpu, cpus, target + 1), then
> > checked --nr; this patch does --nr before doing the next loop, thus,
> > it can remove a cpumask_next_wrap() which costs a little bit.
> >
> > Signed-off-by: Barry Song <song.bao.hua@hisilicon.com>
> > ---
> >  kernel/sched/fair.c | 8 ++++----
> >  1 file changed, 4 insertions(+), 4 deletions(-)
> >
> > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > index ff69f24..e2fb3e0 100644
> > --- a/kernel/sched/fair.c
> > +++ b/kernel/sched/fair.c
> > @@ -6298,9 +6298,9 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, bool
> >
> >               span_avg = sd->span_weight * avg_idle;
> >               if (span_avg > 4*avg_cost)
> > -                     nr = div_u64(span_avg, avg_cost);
> > +                     nr = div_u64(span_avg, avg_cost) - 1;
> >               else
> > -                     nr = 4;
> > +                     nr = 3;
> >
> >               time = cpu_clock(this);
> >       }
> > @@ -6312,11 +6312,11 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, bool
> >                               return i;
> >
> >               } else {
> > -                     if (!--nr)
> > -                             return -1;
> >                       idle_cpu = __select_idle_cpu(cpu, p);
> >                       if ((unsigned int)idle_cpu < nr_cpumask_bits)
> >                               break;
> > +                     if (!--nr)
> > +                             return -1;
> >               }
> >       }
>
> That's just confusing code. Isn't it much clearer to write the whole
> thing like so ?
>
>         nr--;

this is fine to avoid the code of setting 4 to 3 and setting
div_u64(span_avg, avg_cost) to
div_u64(span_avg, avg_cost) - 1;

>         for_each_cpu_wrap(cpu, cpus, target+1) {
>                 ...
>                 if (!nr--)

I guess you mean if(!--nr).

For example, if nr=4, the original code will only check 3 cpus for
__select_idle_cpu.
the new code "nr--" will check 4 cpus for __select_idle_cpu. to keep the amount
untouched,  the code should be --nr.

>                         return -1;
>         }
>

Thanks
Barry