2023-12-13 07:12:19

by Wang Jinchao

Subject: [PATCH v2] sched/fair: merge same code in enqueue_task_fair

The code below is duplicated in two for loops and need to be
consolidated

Signed-off-by: WangJinchao <[email protected]>
---
kernel/sched/fair.c | 31 ++++++++-----------------------
1 file changed, 8 insertions(+), 23 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index d7a3c63a2171..e1373bfd4f2e 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -6681,30 +6681,15 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
cpufreq_update_util(rq, SCHED_CPUFREQ_IOWAIT);

for_each_sched_entity(se) {
- if (se->on_rq)
- break;
cfs_rq = cfs_rq_of(se);
- enqueue_entity(cfs_rq, se, flags);
-
- cfs_rq->h_nr_running++;
- cfs_rq->idle_h_nr_running += idle_h_nr_running;
-
- if (cfs_rq_is_idle(cfs_rq))
- idle_h_nr_running = 1;
-
- /* end evaluation on encountering a throttled cfs_rq */
- if (cfs_rq_throttled(cfs_rq))
- goto enqueue_throttle;
-
- flags = ENQUEUE_WAKEUP;
- }
-
- for_each_sched_entity(se) {
- cfs_rq = cfs_rq_of(se);
-
- update_load_avg(cfs_rq, se, UPDATE_TG);
- se_update_runnable(se);
- update_cfs_group(se);
+ if (se->on_rq) {
+ update_load_avg(cfs_rq, se, UPDATE_TG);
+ se_update_runnable(se);
+ update_cfs_group(se);
+ } else {
+ enqueue_entity(cfs_rq, se, flags);
+ flags = ENQUEUE_WAKEUP;
+ }

cfs_rq->h_nr_running++;
cfs_rq->idle_h_nr_running += idle_h_nr_running;
--
2.40.0

2023-12-13 08:19:51

by Abel Wu

Subject: Re: [PATCH v2] sched/fair: merge same code in enqueue_task_fair

Hi Jinchao,

On 12/13/23 3:12 PM, WangJinchao Wrote:
> The code below is duplicated in two for loops and need to be
> consolidated

It doesn't need to, but it can actually bring some benefit from
the point of view of text size, especially in warehouse-scale
computers where icache is extremely contended.

add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-56 (-56)
Function old new delta
enqueue_task_fair 936 880 -56
Total: Before=64899, After=64843, chg -0.09%

>
> Signed-off-by: WangJinchao <[email protected]>
> ---
> kernel/sched/fair.c | 31 ++++++++-----------------------
> 1 file changed, 8 insertions(+), 23 deletions(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index d7a3c63a2171..e1373bfd4f2e 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -6681,30 +6681,15 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
> cpufreq_update_util(rq, SCHED_CPUFREQ_IOWAIT);
>
> for_each_sched_entity(se) {
> - if (se->on_rq)
> - break;
> cfs_rq = cfs_rq_of(se);
> - enqueue_entity(cfs_rq, se, flags);
> -
> - cfs_rq->h_nr_running++;
> - cfs_rq->idle_h_nr_running += idle_h_nr_running;
> -
> - if (cfs_rq_is_idle(cfs_rq))
> - idle_h_nr_running = 1;
> -
> - /* end evaluation on encountering a throttled cfs_rq */
> - if (cfs_rq_throttled(cfs_rq))
> - goto enqueue_throttle;
> -
> - flags = ENQUEUE_WAKEUP;
> - }
> -
> - for_each_sched_entity(se) {
> - cfs_rq = cfs_rq_of(se);
> -
> - update_load_avg(cfs_rq, se, UPDATE_TG);
> - se_update_runnable(se);
> - update_cfs_group(se);
> + if (se->on_rq) {
> + update_load_avg(cfs_rq, se, UPDATE_TG);
> + se_update_runnable(se);
> + update_cfs_group(se);
> + } else {
> + enqueue_entity(cfs_rq, se, flags);
> + flags = ENQUEUE_WAKEUP;
> + }
>
> cfs_rq->h_nr_running++;
> cfs_rq->idle_h_nr_running += idle_h_nr_running;

I have no strong opinon about this 'cleanup', but the same pattern
can also be found in dequeue_task_fair() and I think it would be
better get them synchronized.

Thanks,
Abel

2023-12-13 08:25:00

by Vincent Guittot

Subject: Re: [PATCH v2] sched/fair: merge same code in enqueue_task_fair

On Wed, 13 Dec 2023 at 09:19, Abel Wu <[email protected]> wrote:
>
> Hi Jinchao,
>
> On 12/13/23 3:12 PM, WangJinchao Wrote:
> > The code below is duplicated in two for loops and need to be
> > consolidated
>
> It doesn't need to, but it can actually bring some benefit from
> the point of view of text size, especially in warehouse-scale
> computers where icache is extremely contended.
>
> add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-56 (-56)
> Function old new delta
> enqueue_task_fair 936 880 -56
> Total: Before=64899, After=64843, chg -0.09%
>
> >
> > Signed-off-by: WangJinchao <[email protected]>
> > ---
> > kernel/sched/fair.c | 31 ++++++++-----------------------
> > 1 file changed, 8 insertions(+), 23 deletions(-)
> >
> > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > index d7a3c63a2171..e1373bfd4f2e 100644
> > --- a/kernel/sched/fair.c
> > +++ b/kernel/sched/fair.c
> > @@ -6681,30 +6681,15 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
> > cpufreq_update_util(rq, SCHED_CPUFREQ_IOWAIT);
> >
> > for_each_sched_entity(se) {
> > - if (se->on_rq)
> > - break;
> > cfs_rq = cfs_rq_of(se);
> > - enqueue_entity(cfs_rq, se, flags);
> > -
> > - cfs_rq->h_nr_running++;
> > - cfs_rq->idle_h_nr_running += idle_h_nr_running;
> > -
> > - if (cfs_rq_is_idle(cfs_rq))
> > - idle_h_nr_running = 1;
> > -
> > - /* end evaluation on encountering a throttled cfs_rq */
> > - if (cfs_rq_throttled(cfs_rq))
> > - goto enqueue_throttle;
> > -
> > - flags = ENQUEUE_WAKEUP;
> > - }
> > -
> > - for_each_sched_entity(se) {
> > - cfs_rq = cfs_rq_of(se);
> > -
> > - update_load_avg(cfs_rq, se, UPDATE_TG);
> > - se_update_runnable(se);
> > - update_cfs_group(se);
> > + if (se->on_rq) {
> > + update_load_avg(cfs_rq, se, UPDATE_TG);
> > + se_update_runnable(se);
> > + update_cfs_group(se);
> > + } else {
> > + enqueue_entity(cfs_rq, se, flags);
> > + flags = ENQUEUE_WAKEUP;
> > + }
> >
> > cfs_rq->h_nr_running++;
> > cfs_rq->idle_h_nr_running += idle_h_nr_running;
>
> I have no strong opinon about this 'cleanup', but the same pattern
> can also be found in dequeue_task_fair() and I think it would be
> better get them synchronized.

I agree, I don't see any benefit from this change

>
> Thanks,
> Abel