[v1] sched/core: Reduce cost of sched_move_task when config autogroup

[PATCH] sched/core: Reduce cost of sched_move_task when config autogroup

Posted by wuchi 2 years, 10 months ago

Some sched_move_task calls of autogroup is useless when the
task_struct->sched_task_group isn't changed because of task_group
of cpu_cgroup overlay task_group of autogroup. The overlay key codes
are as follows:

sched_cgroup_fork->autogroup_task_group->task_wants_autogroup
sched_change_group->autogroup_task_group->autogroup_task_group

sched_move_task eg:
task A belongs to cpu_cgroup0 and autogroup0, it will always to
cpu_cgroup0 when doing exit. So there is no need to do {de|en}queue.
The call graph is as follow.

do_exit
  sched_autogroup_exit_task
    sched_move_task
      dequeue_task
        sched_change_group
	  A.sched_task_group = sched_get_task_group
      enqueue_task

So do some check before dequeue task in sched_move_task.

Signed-off-by: wuchi <wuchi.zero@gmail.com>
---
 kernel/sched/core.c | 29 +++++++++++++++++++++++++++--
 1 file changed, 27 insertions(+), 2 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index a380f34789a2..acc9a0e391f4 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -10330,7 +10330,7 @@ void sched_release_group(struct task_group *tg)
 	spin_unlock_irqrestore(&task_group_lock, flags);
 }
 
-static void sched_change_group(struct task_struct *tsk)
+static struct task_group *sched_get_task_group(struct task_struct *tsk)
 {
 	struct task_group *tg;
 
@@ -10342,7 +10342,28 @@ static void sched_change_group(struct task_struct *tsk)
 	tg = container_of(task_css_check(tsk, cpu_cgrp_id, true),
 			  struct task_group, css);
 	tg = autogroup_task_group(tsk, tg);
-	tsk->sched_task_group = tg;
+
+	return tg;
+}
+
+static bool sched_task_group_changed(struct task_struct *tsk)
+{
+	/*
+	 * Some sched_move_task calls of autogroup is useless when the
+	 * task_struct->sched_task_group isn't changed because of task_group
+	 * of cpu_cgroup overlay task_group of autogroup. so do some check
+	 * before dequeue task in sched_move_task.
+	 */
+#ifdef CONFIG_SCHED_AUTOGROUP
+	return sched_get_task_group(tsk) != tsk->sched_task_group;
+#else
+	return true;
+#endif /* CONFIG_SCHED_AUTOGROUP */
+}
+
+static void sched_change_group(struct task_struct *tsk)
+{
+	tsk->sched_task_group = sched_get_task_group(tsk);
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	if (tsk->sched_class->task_change_group)
@@ -10369,6 +10390,9 @@ void sched_move_task(struct task_struct *tsk)
 	rq = task_rq_lock(tsk, &rf);
 	update_rq_clock(rq);
 
+	if (!sched_task_group_changed(tsk))
+		goto unlock;
+
 	running = task_current(rq, tsk);
 	queued = task_on_rq_queued(tsk);
 
@@ -10391,6 +10415,7 @@ void sched_move_task(struct task_struct *tsk)
 		resched_curr(rq);
 	}
 
+unlock:
 	task_rq_unlock(rq, tsk, &rf);
 }
 
-- 
2.20.1

Re: [PATCH] sched/core: Reduce cost of sched_move_task when config autogroup

Posted by Peter Zijlstra 2 years, 10 months ago

On Sun, Mar 19, 2023 at 03:56:43PM +0800, wuchi wrote:
> Some sched_move_task calls of autogroup is useless when the
> task_struct->sched_task_group isn't changed because of task_group
> of cpu_cgroup overlay task_group of autogroup. The overlay key codes
> are as follows:
> 
> sched_cgroup_fork->autogroup_task_group->task_wants_autogroup
> sched_change_group->autogroup_task_group->autogroup_task_group
> 
> sched_move_task eg:
> task A belongs to cpu_cgroup0 and autogroup0, it will always to
> cpu_cgroup0 when doing exit. So there is no need to do {de|en}queue.
> The call graph is as follow.
> 
> do_exit
>   sched_autogroup_exit_task
>     sched_move_task
>       dequeue_task
>         sched_change_group
> 	  A.sched_task_group = sched_get_task_group
>       enqueue_task
> 
> So do some check before dequeue task in sched_move_task.

No immediate objection; but the $subject seems to suggest you did this
for performance reasons, yet there are no performance results in this
Changelog. How much does this save under what load?

Re: [PATCH] sched/core: Reduce cost of sched_move_task when config autogroup

Posted by chi wu 2 years, 10 months ago

Peter Zijlstra <peterz@infradead.org> 于2023年3月20日周一 16:28写道：
>
> On Sun, Mar 19, 2023 at 03:56:43PM +0800, wuchi wrote:
> > Some sched_move_task calls of autogroup is useless when the
> > task_struct->sched_task_group isn't changed because of task_group
> > of cpu_cgroup overlay task_group of autogroup. The overlay key codes
> > are as follows:
> >
> > sched_cgroup_fork->autogroup_task_group->task_wants_autogroup
> > sched_change_group->autogroup_task_group->autogroup_task_group
> >
> > sched_move_task eg:
> > task A belongs to cpu_cgroup0 and autogroup0, it will always to
> > cpu_cgroup0 when doing exit. So there is no need to do {de|en}queue.
> > The call graph is as follow.
> >
> > do_exit
> >   sched_autogroup_exit_task
> >     sched_move_task
> >       dequeue_task
> >         sched_change_group
> >         A.sched_task_group = sched_get_task_group
> >       enqueue_task
> >
> > So do some check before dequeue task in sched_move_task.
>
> No immediate objection; but the $subject seems to suggest you did this
> for performance reasons, yet there are no performance results in this
> Changelog. How much does this save under what load?

cpu:bogomips=4600.00
kernel:6.3.0-rc3
cpu cgroup: 6:cpu,cpuacct:/user.slice

run cmds:

#!/bin/bash
for i in {0..10000}; do
sleep 0 &
done
wait

bpftrace -e 'k:sched_move_task { @ts[tid] = nsecs; }
kr:sched_move_task /@ts[tid]/ { @ns += nsecs - @ts[tid];
delete(@ts[tid]); }'

cost time(ns):
 without patch: 43528033
 with      patch: 18934496
                  diff: -24593537  -56.5%

About the change, move sched_task_group_changed before
update_rq_clock as following and add the performance results in
the Changelog?

@@ -10369,6 +10390,9 @@ void sched_move_task(struct task_struct *tsk)
        rq = task_rq_lock(tsk, &rf);

+       if (!sched_task_group_changed(tsk))
+               goto unlock;
+
        update_rq_clock(rq);

Re: [PATCH] sched/core: Reduce cost of sched_move_task when config autogroup

Posted by Peter Zijlstra 2 years, 10 months ago

On Mon, Mar 20, 2023 at 08:40:51PM +0800, chi wu wrote:
> Peter Zijlstra <peterz@infradead.org> 于2023年3月20日周一 16:28写道：
> >
> > On Sun, Mar 19, 2023 at 03:56:43PM +0800, wuchi wrote:
> > > Some sched_move_task calls of autogroup is useless when the
> > > task_struct->sched_task_group isn't changed because of task_group
> > > of cpu_cgroup overlay task_group of autogroup. The overlay key codes
> > > are as follows:
> > >
> > > sched_cgroup_fork->autogroup_task_group->task_wants_autogroup
> > > sched_change_group->autogroup_task_group->autogroup_task_group
> > >
> > > sched_move_task eg:
> > > task A belongs to cpu_cgroup0 and autogroup0, it will always to
> > > cpu_cgroup0 when doing exit. So there is no need to do {de|en}queue.
> > > The call graph is as follow.
> > >
> > > do_exit
> > >   sched_autogroup_exit_task
> > >     sched_move_task
> > >       dequeue_task
> > >         sched_change_group
> > >         A.sched_task_group = sched_get_task_group
> > >       enqueue_task
> > >
> > > So do some check before dequeue task in sched_move_task.
> >
> > No immediate objection; but the $subject seems to suggest you did this
> > for performance reasons, yet there are no performance results in this
> > Changelog. How much does this save under what load?
> 
> cpu:bogomips=4600.00
> kernel:6.3.0-rc3
> cpu cgroup: 6:cpu,cpuacct:/user.slice
> 
> run cmds:
> 
> #!/bin/bash
> for i in {0..10000}; do
> sleep 0 &
> done
> wait
> 
> bpftrace -e 'k:sched_move_task { @ts[tid] = nsecs; }
> kr:sched_move_task /@ts[tid]/ { @ns += nsecs - @ts[tid];
> delete(@ts[tid]); }'
> 
> cost time(ns):
>  without patch: 43528033
>  with      patch: 18934496
>                   diff: -24593537  -56.5%
> 
> About the change, move sched_task_group_changed before
> update_rq_clock as following and add the performance results in
> the Changelog?

Yes please, that gives good motivation for the change.

Thanks!