Re: [PATCH v5 3/7] sched: set initial value of runnable avg for newforked task

[Alex Shi]

On 05/09/2013 05:30 PM, Paul Turner wrote:
>> > With sched_slice, we need to set the runnable avg sum/period after new
>> > task assigned to a specific CPU.
>> > So, set them __sched_fork is meaningless. and then
> This is still a reasonable choice.
> 
> Assuming the system is well balanced, sched_slice() on the current CPU
> should be reasonably indicative of the slice wherever we end up.  The
> alternative is still to pick a constant.  We should do one of these.
> 
>> > __update_task_entity_contrib(&p->se) also no reason to use.
> Surely we'd still want it so that the right load is added by
> enqueue_entity_load_avg()?
> 

thanks for comments! 
the new patch attached, hope it is your liked this time. :)

---