[PATCH] sched/fair: Simplify task_numa_find_cpu()

[PATCH] sched/fair: Simplify task_numa_find_cpu()

[Yuan ZhaoXiong]

Combine the 'cpumask_of_node()' with 'env->p->cpus_ptr' and drop the
cpumask_test_cpu().

Signed-off-by: Yuan ZhaoXiong <yuanzhaoxiong@baidu.com>
---
 kernel/sched/fair.c | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index f6a05d9..dc064d7 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -1950,11 +1950,7 @@ static void task_numa_find_cpu(struct task_numa_env *env,
 		maymove = !load_too_imbalanced(src_load, dst_load, env);
 	}
 
-	for_each_cpu(cpu, cpumask_of_node(env->dst_nid)) {
-		/* Skip this CPU if the source task cannot migrate */
-		if (!cpumask_test_cpu(cpu, env->p->cpus_ptr))
-			continue;
-
+	for_each_cpu_and(cpu, cpumask_of_node(env->dst_nid), env->p->cpus_ptr) {
 		env->dst_cpu = cpu;
 		if (task_numa_compare(env, taskimp, groupimp, maymove))
 			break;
-- 
1.8.3.1

Re: [PATCH] sched/fair: Simplify task_numa_find_cpu()

[Mel Gorman]

On Sat, Oct 30, 2021 at 03:25:59PM +0800, Yuan ZhaoXiong wrote:
> Combine the 'cpumask_of_node()' with 'env->p->cpus_ptr' and drop the
> cpumask_test_cpu().
> 
> Signed-off-by: Yuan ZhaoXiong <yuanzhaoxiong@baidu.com>

This potentially creates of a temporary cpumask variable as noted in the
comment for for_each_cpu_and.

 * This saves a temporary CPU mask in many places.  It is equivalent to:
 *      struct cpumask tmp;
 *      cpumask_and(&tmp, &mask1, &mask2);
 *      for_each_cpu(cpu, &tmp)
 *              ...

task_numa_find_cpu() is a relatively deep function. Did you check
the stack usage to make sure it's not pushing too close to the stack
boundary? While there are other users of for_each_cpu_and, they are mostly
shallow although find_energy_efficient_cpu() is a bit questionable and
probably should have used select_idle_mask.

Does the patch have a noticable performance impact? 

-- 
Mel Gorman
SUSE Labs

[PATCH] sched/fair: simplify task_numa_find_cpu()

[Yury Norov]

From: Yury Norov (NVIDIA) <yury.norov@gmail.com>

Use for_each_cpu_and() and drop some housekeeping code.

Signed-off-by: Yury Norov (NVIDIA) <yury.norov@gmail.com>
---
 kernel/sched/fair.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 0990ed90c14e..c48c0975fe7a 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2474,11 +2474,8 @@ static void task_numa_find_cpu(struct task_numa_env *env,
 		maymove = !load_too_imbalanced(src_load, dst_load, env);
 	}
 
-	for_each_cpu(cpu, cpumask_of_node(env->dst_nid)) {
-		/* Skip this CPU if the source task cannot migrate */
-		if (!cpumask_test_cpu(cpu, env->p->cpus_ptr))
-			continue;
-
+	/* Skip CPUs if the source task cannot migrate */
+	for_each_cpu_and(cpu, cpumask_of_node(env->dst_nid), env->p->cpus_ptr) {
 		env->dst_cpu = cpu;
 		if (task_numa_compare(env, taskimp, groupimp, maymove))
 			break;
-- 
2.43.0

Re: [PATCH] sched/fair: simplify task_numa_find_cpu()

[Phil Auld]

On Thu, Sep 11, 2025 at 04:31:35PM -0400 Yury Norov wrote:
> From: Yury Norov (NVIDIA) <yury.norov@gmail.com>
> 
> Use for_each_cpu_and() and drop some housekeeping code.
> 
> Signed-off-by: Yury Norov (NVIDIA) <yury.norov@gmail.com>

The "housekeeping" part threw me off for a second. I was looking for
isolation related code being dropped :)


Reviewed-by: Phil Auld <pauld@redhat.com>


> ---
>  kernel/sched/fair.c | 7 ++-----
>  1 file changed, 2 insertions(+), 5 deletions(-)
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 0990ed90c14e..c48c0975fe7a 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -2474,11 +2474,8 @@ static void task_numa_find_cpu(struct task_numa_env *env,
>  		maymove = !load_too_imbalanced(src_load, dst_load, env);
>  	}
>  
> -	for_each_cpu(cpu, cpumask_of_node(env->dst_nid)) {
> -		/* Skip this CPU if the source task cannot migrate */
> -		if (!cpumask_test_cpu(cpu, env->p->cpus_ptr))
> -			continue;
> -
> +	/* Skip CPUs if the source task cannot migrate */
> +	for_each_cpu_and(cpu, cpumask_of_node(env->dst_nid), env->p->cpus_ptr) {
>  		env->dst_cpu = cpu;
>  		if (task_numa_compare(env, taskimp, groupimp, maymove))
>  			break;
> -- 
> 2.43.0
> 
> 

-- 

Re: [PATCH] sched/fair: simplify task_numa_find_cpu()

[K Prateek Nayak]

Hello Yury,

On 9/12/2025 2:01 AM, Yury Norov wrote:
> From: Yury Norov (NVIDIA) <yury.norov@gmail.com>
> 
> Use for_each_cpu_and() and drop some housekeeping code.

Please feel free to include:

Reviewed-by: K Prateek Nayak <kprateek.nayak@amd.com>

> 
> Signed-off-by: Yury Norov (NVIDIA) <yury.norov@gmail.com>
> ---
>  kernel/sched/fair.c | 7 ++-----
>  1 file changed, 2 insertions(+), 5 deletions(-)
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 0990ed90c14e..c48c0975fe7a 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -2474,11 +2474,8 @@ static void task_numa_find_cpu(struct task_numa_env *env,
>  		maymove = !load_too_imbalanced(src_load, dst_load, env);
>  	}
>  
> -	for_each_cpu(cpu, cpumask_of_node(env->dst_nid)) {
> -		/* Skip this CPU if the source task cannot migrate */
> -		if (!cpumask_test_cpu(cpu, env->p->cpus_ptr))
> -			continue;
> -
> +	/* Skip CPUs if the source task cannot migrate */
> +	for_each_cpu_and(cpu, cpumask_of_node(env->dst_nid), env->p->cpus_ptr) {
>  		env->dst_cpu = cpu;
>  		if (task_numa_compare(env, taskimp, groupimp, maymove))
>  			break;

-- 
Thanks and Regards,
Prateek