Re: [PATCH v9 1/4] sched: Remove rq's runnable avg

[Dietmar Eggemann <dietmar.eggemann@arm.com>]

On 23/06/15 01:08, Yuyang Du wrote:
> The per-rq runnable averages (rq->avg) was introduced by Ben with this commit:
> 
> commit 18bf2805d9b30cb823d4919b42cd230f59c7ce1f
> Author: Ben Segall <bsegall@google.com>
> Date:   Thu Oct 4 12:51:20 2012 +0200
> 
>     sched: Maintain per-rq runnable averages
> 
>     Since runqueues do not have a corresponding sched_entity we instead embed a
>     sched_avg structure directly.
> 
>     Signed-off-by: Ben Segall <bsegall@google.com>
>     Reviewed-by: Paul Turner <pjt@google.com>
>     Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
>     Link: http://lkml.kernel.org/r/20120823141506.442637130@google.com
>     Signed-off-by: Ingo Molnar <mingo@kernel.org>
> 
> With CONFIG_FAIR_GROUP_SCHED=y, the rq->avg (runnable_avg_sum, avg_period) is
> used to calc the contrib in __update_tg_runnable_avg() for the root group.
> 
> But the root group's load data is never consumed since its merge into kernel
> in 2012. The basic reason is we have just one root group, and it is
> meaningless to give it weight. As the extra code is in the scheduler's
> hot path, so it is better to remove it.
> 
> Signed-off-by: Yuyang Du <yuyang.du@intel.com>

Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Tested-by: Dietmar Eggemann <dietmar.eggemann@arm.com>

There is an additional clean-up possible since idle_[enter|exit]_fair()
are empty now.

[...]