[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 3 Mar 2016 15:21:07 +0100
From: Ingo Molnar <mingo@...nel.org>
To: Steve Muckle <steve.muckle@...aro.org>
Cc: Peter Zijlstra <peterz@...radead.org>,
Ingo Molnar <mingo@...hat.com>,
"Rafael J. Wysocki" <rafael@...nel.org>,
linux-kernel@...r.kernel.org, linux-pm@...r.kernel.org,
Vincent Guittot <vincent.guittot@...aro.org>,
Morten Rasmussen <morten.rasmussen@....com>,
Dietmar Eggemann <dietmar.eggemann@....com>,
Juri Lelli <Juri.Lelli@....com>,
Patrick Bellasi <patrick.bellasi@....com>,
Michael Turquette <mturquette@...libre.com>,
Ricky Liang <jcliang@...omium.org>
Subject: Re: [RFCv7 PATCH 03/10] sched: scheduler-driven cpu frequency
selection
* Steve Muckle <steve.muckle@...aro.org> wrote:
> From: Michael Turquette <mturquette@...libre.com>
>
> Scheduler-driven CPU frequency selection hopes to exploit both
> per-task and global information in the scheduler to improve frequency
> selection policy, achieving lower power consumption, improved
> responsiveness/performance, and less reliance on heuristics and
> tunables. For further discussion on the motivation of this integration
> see [0].
>
> This patch implements a shim layer between the Linux scheduler and the
> cpufreq subsystem. The interface accepts capacity requests from the
> CFS, RT and deadline sched classes. The requests from each sched class
> are summed on each CPU with a margin applied to the CFS and RT
> capacity requests to provide some headroom. Deadline requests are
> expected to be precise enough given their nature to not require
> headroom. The maximum total capacity request for a CPU in a frequency
> domain drives the requested frequency for that domain.
>
> Policy is determined by both the sched classes and this shim layer.
>
> Note that this algorithm is event-driven. There is no polling loop to
> check cpu idle time nor any other method which is unsynchronized with
> the scheduler, aside from an optional throttling mechanism.
>
> Thanks to Juri Lelli <juri.lelli@....com> for contributing design ideas,
> code and test results, and to Ricky Liang <jcliang@...omium.org>
> for initialization and static key inc/dec fixes.
>
> [0] http://article.gmane.org/gmane.linux.kernel/1499836
>
> [smuckle@...aro.org: various additions and fixes, revised commit text]
>
> CC: Ricky Liang <jcliang@...omium.org>
> Signed-off-by: Michael Turquette <mturquette@...libre.com>
> Signed-off-by: Juri Lelli <juri.lelli@....com>
> Signed-off-by: Steve Muckle <smuckle@...aro.org>
> ---
> drivers/cpufreq/Kconfig | 21 ++
> include/linux/cpufreq.h | 3 +
> include/linux/sched.h | 8 +
> kernel/sched/Makefile | 1 +
> kernel/sched/cpufreq_sched.c | 459 +++++++++++++++++++++++++++++++++++++++++++
Please rename this to kernel/sched/cpufreq.c - no need to say 'sched' twice! :-)
> kernel/sched/sched.h | 51 +++++
> 6 files changed, 543 insertions(+)
> create mode 100644 kernel/sched/cpufreq_sched.c
So I really like how you push all high level code into kernel/sched/cpufreq.c and
use the cpufreq drivers only for actual low level frequency switching.
It would be nice to converge this code with the code from Rafael:
[PATCH 0/6] cpufreq: schedutil governor
i.e. use scheduler internal metrics within the scheduler, and create a clear
interface between low level cpufreq drivers and the cpufreq code living in the
scheduler.
Thanks,
Ingo
Powered by blists - more mailing lists