[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20231217212312.igyy5ydjwywmk5x3@airbuntu>
Date: Sun, 17 Dec 2023 21:23:12 +0000
From: Qais Yousef <qyousef@...alina.io>
To: Dietmar Eggemann <dietmar.eggemann@....com>
Cc: Ingo Molnar <mingo@...nel.org>, Peter Zijlstra <peterz@...radead.org>,
"Rafael J. Wysocki" <rafael@...nel.org>,
Viresh Kumar <viresh.kumar@...aro.org>,
Vincent Guittot <vincent.guittot@...aro.org>,
linux-kernel@...r.kernel.org, linux-pm@...r.kernel.org,
Lukasz Luba <lukasz.luba@....com>, Wei Wang <wvw@...gle.com>,
Rick Yiu <rickyiu@...gle.com>, Chung-Kai Mei <chungkai@...gle.com>,
Hongyan Xia <hongyan.xia2@....com>
Subject: Re: [PATCH 0/4] sched: cpufreq: Remove uclamp max-aggregation
On 12/18/23 09:19, Dietmar Eggemann wrote:
> On 08/12/2023 02:52, Qais Yousef wrote:
>
> [...]
>
> > ===
> >
> > This patch is based on remove margins series [1] and data is collected it
> > against it as a baseline.
> >
> > Testing on pixel 6 with mainline(ish) kernel
>
> How is the Pixel6 configured in terms of per-policy rate_limit_us and
> response_time_ms ? Is this the now default 2ms and whatever the systems
> calculates for response_time_ms ?
Yes.
>
> Pixel6 is still a slow switching device, rigth?
>
> root 297 2 1 08:58:01 ? 00:00:13 [sugov:0]
> root 298 2 0 08:58:01 ? 00:00:03 [sugov:4]
> root 299 2 1 08:58:01 ? 00:00:05 [sugov:6]
Yes.
>
> > ==
> >
> > Speedometer browser benchmark
> >
> > | baseline | 1.25 headroom | patch | patch + 1.25 headroom
> > -------+-----------+---------------+-----------+---------------------
> > score | 108.03 | 135.72 | 108.09 | 137.47
> > -------+-----------+---------------+-----------+---------------------
> > power | 1204.75 | 1451.79 | 1216.17 | 1484.73
> > -------+-----------+---------------+-----------+---------------------
>
> What's the difference between baseline & 1.25 headroom. IMHO, we have:
Baseline is the remove-margins patches as specified in the quoted text above
> > This patch is based on remove margins series [1] and data is collected it
> > against it as a baseline.
The series were stacked on top of each others. Results from this run should be
compared to remove-margins[1] tables too.
>
> static inline unsigned long map_util_perf(unsigned long util)
> {
> return util + (util >> 2);
> }
>
> on baseline?
This is not baseline. See above.
>
> By patch you refer to the whole patch-set + [1]?
>
> And I assume 'patch + 1.25 headroom' is 'response_time_ms' tuned to
> reach 1.25 ?
Yes. Which is done by multiplying the response_time_ms with 0.8.
Cheers
--
Qais Yousef
Powered by blists - more mailing lists