[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20210625152603.25960-1-lukasz.luba@arm.com>
Date: Fri, 25 Jun 2021 16:26:00 +0100
From: Lukasz Luba <lukasz.luba@....com>
To: linux-kernel@...r.kernel.org
Cc: Chris.Redpath@....com, lukasz.luba@....com,
dietmar.eggemann@....com, morten.rasmussen@....com,
qperret@...gle.com, linux-pm@...r.kernel.org, peterz@...radead.org,
rjw@...ysocki.net, viresh.kumar@...aro.org,
vincent.guittot@...aro.org, mingo@...hat.com,
juri.lelli@...hat.com, rostedt@...dmis.org, segall@...gle.com,
mgorman@...e.de, bristot@...hat.com, CCj.Yeh@...iatek.com
Subject: [PATCH 0/3] Improve EAS energy estimation and increase precision
Hi all,
The patch set aims to address a scenario for Energy Aware Scheduler,
where we estimate and compare energy values and miss a more precised results.
In some use cases estimations for two CPUs might give the same values
for a given task and it's utilization. Those values would be different
when we have a better precision and avoid this rounding error.
Thus, the decision of choosing a CPU for a waking-up task might also
be better.
We have received this feedback from our partners.
Address this rounding error issue and increase the precision of Energy Model
em_perf_state::cost values. This change should not affect other
subsystems in kernel: thermal IPA, PowerCap DTPM, etc, since they use
em_perf_state::power field, which is not touched. It also doesn't
trigger the need for updating all existing platforms to register EM and report
power values in different scale.
Regards,
Lukasz
Lukasz Luba (3):
sched/fair: Prepare variables for increased precision of EAS estimated
energy
PM: EM: Make em_cpu_energy() able to return bigger values
PM: EM: Increase energy calculation precision
include/linux/energy_model.h | 11 +++++++----
kernel/power/energy_model.c | 3 ++-
kernel/sched/fair.c | 13 +++++++------
3 files changed, 16 insertions(+), 11 deletions(-)
--
2.17.1
Powered by blists - more mailing lists