Re: [Patch v6 1/7] sched/pelt.c: Add support to track thermal pressure

[Dietmar Eggemann]

On 08/01/2020 15:56, Thara Gopinath wrote:
> On Tue, 17 Dec 2019 at 07:54, Dietmar Eggemann <dietmar.eggemann@xxxxxxx>
> wrote:
> 
>> On 12/12/2019 05:11, Thara Gopinath wrote:
>>
>> minor: in subject: s/sched/pelt.c/sched/pelt
>>
>> [...]
>>
>>> diff --git a/kernel/sched/pelt.c b/kernel/sched/pelt.c
>>> index a96db50..9aac3b7 100644
>>> --- a/kernel/sched/pelt.c
>>> +++ b/kernel/sched/pelt.c
>>> @@ -353,6 +353,28 @@ int update_dl_rq_load_avg(u64 now, struct rq *rq,
>> int running)
>>>       return 0;
>>>  }
>>>
>>> +/*
>>> + * thermal:
>>> + *
>>> + *   load_sum = \Sum se->avg.load_sum
>>
>> Why not '\Sum rq->avg.load_sum' ?
>>
> Hi Dietmar,
> 
> The header for all other update_*_load_avg api use se->avg. and not rq->avg.

True but at least they (rt_rq, dl_rq, irq) also say 'but
se->avg.util_sum is not tracked'.

I guess this comment originally came from the
'__update_load_avg_blocked_se(), __update_load_avg_se(),
__update_load_avg_cfs_rq()' block:

* cfq_rq:
 *
 *   load_sum = \Sum se_weight(se) * se->avg.load_sum

but for rt_rq, dl_rq, irq and thermal we don't have a relationship like
between se and cfs_rq's so that's why this comment confuses me.