Re: [RFC PATCH v4 1/2] sched/fair: Introduce short duration task check

[Chen Yu]

Hi Dietmar,
thanks for reviewing the patch!
On 2023-01-05 at 12:33:16 +0100, Dietmar Eggemann wrote:
> On 16/12/2022 07:11, Chen Yu wrote:
> 
> [...]
> 
> > @@ -5995,6 +6005,18 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
> >  
> >  static void set_next_buddy(struct sched_entity *se);
> >  
> > +static inline void dur_avg_update(struct task_struct *p, bool task_sleep)
> > +{
> > +	u64 dur;
> > +
> > +	if (!task_sleep)
> > +		return;
> > +
> > +	dur = p->se.sum_exec_runtime - p->se.prev_sum_exec_runtime_vol;
> > +	p->se.prev_sum_exec_runtime_vol = p->se.sum_exec_runtime;
> 
> Shouldn't se->prev_sum_exec_runtime_vol be set in enqueue_task_fair()
> and not in dequeue_task_fair()->dur_avg_update()? Otherwise `dur` will
> contain sleep time.
> 
After the task p is dequeued, p's sum_exec_runtime will not be increased.
Unless task p is switched in again, p's sum_exec_runtime will continue to
increase. So dur should not include the sleep time, because we substract
between the sum_exec_runtime rather than rq->clock_task. Not sure if I understand
this correctly?

My original thought was that, record the average run time of every section:
Only consider that task voluntarily relinquishes the CPU.
For example, suppose on CPU1, task p1 and p2 run alternatively:

 --------------------> time

 | p1 runs 1ms | p2 preempt p1 | p1 switch in, runs 0.5ms and blocks |
               ^               ^                                     ^
 |_____________|               |_____________________________________|
                                                                     ^
                                                                     |
                                                                  p1 dequeued

p1's duration in one section is (1 + 0.5)ms. Because if p2 does not
preempt p1, p1 can run 1.5ms. This reflects the nature of a task,
how long it wishes to run at most.

> Like we do for se->prev_sum_exec_runtime in set_next_entity() but for
> one `set_next_entity()-put_prev_entity()` run section.
> 
> AFAICS, you want to measure the exec_runtime sum over all run sections
> between enqueue and dequeue.
Yes, we tried to record the 'decayed' average exec_runtime for each section.
Say, task p runs for a ms , then p is dequeued and blocks for b ms, and then
runs for c ms, its average duration is 0.875 * a + 0.125 * c , which is
what update_avg() does.

thanks,
Chenyu