mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Dietmar Eggemann <dietmar.eggemann@arm.com>
To: Qais Yousef <qyousef@layalina.io>
Cc: Vincent Guittot <vincent.guittot@linaro.org>,
	mingo@kernel.org, peterz@infradead.org, rafael@kernel.org,
	viresh.kumar@linaro.org, vschneid@redhat.com,
	linux-pm@vger.kernel.org, linux-kernel@vger.kernel.org,
	lukasz.luba@arm.com, wvw@google.com, xuewen.yan94@gmail.com,
	han.lin@mediatek.com, Jonathan.JMChen@mediatek.com
Subject: Re: [PATCH v3] sched/fair: unlink misfit task from cpu overutilized
Date: Wed, 18 Jan 2023 16:17:31 +0000	[thread overview]
Message-ID: <69457060-ede6-c805-af1d-a6e2b05fd55e@arm.com> (raw)
In-Reply-To: <20230117163841.d5jv6ysqf5kmvvmh@airbuntu>

On 17/01/2023 16:38, Qais Yousef wrote:
> On 01/16/23 09:07, Dietmar Eggemann wrote:
> 
> [...]
> 
>> Not sure if people get what `performance requirements` mean here? I know
>> we want to use `performance` rather `bandwidth hint` to describe what
>> uclamp is. So shouldn't we use `utilization but also uclamp`?
> 
> We do have the uclamp doc now which explains this, no? I'm not keen on
> utilization because it's an overloaded term. In the context of uclamp

And `performance` isn't ? ;-) True, the doc refers to uclamp as a
`performance requirements`.

> - utilization _signal_ in the scheduler is used to indicate performance
> requirements of a workload, no?

I was referring to:

 4569 static inline int task_fits_cpu(struct task_struct *p, int cpu)
 4570 {
 4571   unsigned long uclamp_min = uclamp_eff_value(p, UCLAMP_MIN);
 4572   unsigned long uclamp_max = uclamp_eff_value(p, UCLAMP_MAX);
 4573   unsigned long util = task_util_est(p);
 4574   /*
 4575   * Return true only if the cpu fully fits the task requirements,
 4576   * which include the utilization but also the performance.
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 4577   */
 4578   return (util_fits_cpu(util, uclamp_min, uclamp_max, cpu) > 0);
 4579 }

So here we explicitly talk about `utilization` (util_avg/util_est)
versus `uclamp (max/min)` and the latter is referred as `performance`.
You're right, we shouldn't refer to `uclamp (min/max)` as `utilization`
either.

In other places we use:

select_idle_capacity()

/* This CPU fits with all capacity and performance requirements */
                          ^^^^^^^^     ^^^^^^^^^^^^^^^^^^^^^^^^

`capacity` is probably equal `utilization`? and `performance
requirements` stand for `uclamp (min/max)`.

/* Only the min performance (i.e. uclamp_min) doesn't fit */
            ^^^^^^^^^^^^^^^
here we link `min performance` explicitly to `uclamp_min`.

/* Look for the CPU with highest performance capacity.
                                 ^^^^^^^^^^^^^^^^^^^^
I guess this stands for `cap_orig - thermal_load_avg()`

feec()

/* Both don't fit performance (i.e. uclamp_min) but best energy cpu has
                  ^^^^^^^^^^^  ^^^^^^^^^^^^^^^
better performance. */
^^^^^^ ^^^^^^^^^^^

Here I assume `better performance` stands for higher `cap_orig -
thermal_pressure', not for `uclamp min or max`?

---

IMHO, referring to `uclamp (min/max)` as `performance (min/max)
hint/(requirement)` is fine as long as it's done consistently in
comments and the alias is not used for other items.

> 
> Using 'uclamp hint' if you found it really confusing, is fine by me. But I'd
> rather steer away from 'bandwidth' or 'utilization' when describing uclamp and
> its intention.
> 
> I like using performance requirements because it enforces what this hint
> actually means.


      reply	other threads:[~2023-01-18 16:21 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-01-13 13:40 Vincent Guittot
2023-01-15  0:19 ` Qais Yousef
2023-01-16 11:08   ` Vincent Guittot
2023-01-17 16:20     ` Qais Yousef
2023-01-18 13:49       ` Vincent Guittot
2023-01-16  9:07 ` Dietmar Eggemann
2023-01-16 11:23   ` Vincent Guittot
2023-01-16 14:56     ` Dietmar Eggemann
2023-01-17 20:20       ` Qais Yousef
2023-01-18  8:15       ` Vincent Guittot
2023-01-18 16:48         ` Qais Yousef
2023-01-19 10:08           ` Vincent Guittot
2023-01-20 18:53             ` Dietmar Eggemann
2023-01-17 16:38   ` Qais Yousef
2023-01-18 16:17     ` Dietmar Eggemann [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=69457060-ede6-c805-af1d-a6e2b05fd55e@arm.com \
    --to=dietmar.eggemann@arm.com \
    --cc=Jonathan.JMChen@mediatek.com \
    --cc=han.lin@mediatek.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-pm@vger.kernel.org \
    --cc=lukasz.luba@arm.com \
    --cc=mingo@kernel.org \
    --cc=peterz@infradead.org \
    --cc=qyousef@layalina.io \
    --cc=rafael@kernel.org \
    --cc=vincent.guittot@linaro.org \
    --cc=viresh.kumar@linaro.org \
    --cc=vschneid@redhat.com \
    --cc=wvw@google.com \
    --cc=xuewen.yan94@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®