From: Vincent Guittot <vincent.guittot@linaro.org>
To: Sai Gurrappadi <sgurrappadi@nvidia.com>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>,
Linux PM <linux-pm@vger.kernel.org>,
Peter Zijlstra <peterz@infradead.org>,
LKML <linux-kernel@vger.kernel.org>,
Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>,
Viresh Kumar <viresh.kumar@linaro.org>,
Juri Lelli <juri.lelli@arm.com>,
Patrick Bellasi <patrick.bellasi@arm.com>,
Joel Fernandes <joelaf@google.com>,
Morten Rasmussen <morten.rasmussen@arm.com>,
Ingo Molnar <mingo@redhat.com>,
Thomas Gleixner <tglx@linutronix.de>,
Peter Boonstoppel <pboonstoppel@nvidia.com>
Subject: Re: [RFC][PATCH v3 2/2] cpufreq: schedutil: Avoid reducing frequency of busy CPUs prematurely
Date: Mon, 27 Mar 2017 09:04:15 +0200 [thread overview]
Message-ID: <CAKfTPtDAVZg--KchHN9S3hdgRB9OzQKJTR7ioCh8hTksRDDekg@mail.gmail.com> (raw)
In-Reply-To: <58D5C47A.4060305@nvidia.com>
On 25 March 2017 at 02:14, Sai Gurrappadi <sgurrappadi@nvidia.com> wrote:
> Hi Rafael,
>
> On 03/21/2017 04:08 PM, Rafael J. Wysocki wrote:
>> From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
>>
>> The way the schedutil governor uses the PELT metric causes it to
>> underestimate the CPU utilization in some cases.
>>
>> That can be easily demonstrated by running kernel compilation on
>> a Sandy Bridge Intel processor, running turbostat in parallel with
>> it and looking at the values written to the MSR_IA32_PERF_CTL
>> register. Namely, the expected result would be that when all CPUs
>> were 100% busy, all of them would be requested to run in the maximum
>> P-state, but observation shows that this clearly isn't the case.
>> The CPUs run in the maximum P-state for a while and then are
>> requested to run slower and go back to the maximum P-state after
>> a while again. That causes the actual frequency of the processor to
>> visibly oscillate below the sustainable maximum in a jittery fashion
>> which clearly is not desirable.
>>
>> That has been attributed to CPU utilization metric updates on task
>> migration that cause the total utilization value for the CPU to be
>> reduced by the utilization of the migrated task. If that happens,
>> the schedutil governor may see a CPU utilization reduction and will
>> attempt to reduce the CPU frequency accordingly right away. That
>> may be premature, though, for example if the system is generally
>> busy and there are other runnable tasks waiting to be run on that
>> CPU already.
>>
>
> Thinking out loud a bit, I wonder if what you really want to do is basically:
>
> schedutil_cpu_util(cpu) = max(cpu_rq(cpu)->cfs.util_avg, total_cpu_util_avg);
>
> Where total_cpu_util_avg tracks the average utilization of the CPU itself over time (% of time the CPU was busy) in the same PELT like manner. The difference here is that it doesn't change instantaneously as tasks migrate in/out but it decays/accumulates just like the per-entity util_avgs.
But we loose the interest of immediate decrease when tasks migrate.
Instead of total_cpu_util_avg we should better track RT utilization in
the same manner so with ongoing work for deadline we will have :
total_utilization = cfs.util_avg + rt's util_avg + deadline's util avg
and we still take advantage of task migration effect
>
> Over time, total_cpu_util_avg and cfs_rq(cpu)->util_avg will tend towards each other the lesser the amount of 'overlap' / overloading.
>
> Yes, the above metric would 'overestimate' in case all tasks have migrated away and we are left with an idle CPU. A fix for that could be to just use the PELT value like so:
>
> schedutil_cpu_util(cpu) = max(cpu_rq(cpu)->cfs.util_avg, idle_cpu(cpu) ? 0 : total_cpu_util_avg);
>
> Note that the problem described here in the commit message doesn't need fully runnable threads, it just needs two threads to execute in parallel on the same CPU for a period of time. I don't think looking at just idle_calls necessarily covers all cases.
>
> Thoughts?
>
> Thanks,
> -Sai
next prev parent reply other threads:[~2017-03-27 7:14 UTC|newest]
Thread overview: 71+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-03-19 13:21 [PATCH 0/2] cpufreq: schedutil: Fix and optimization Rafael J. Wysocki
2017-03-19 13:30 ` [PATCH 1/2] cpufreq: schedutil: Fix per-CPU structure initialization in sugov_start() Rafael J. Wysocki
2017-03-20 3:28 ` Viresh Kumar
2017-03-20 12:36 ` Rafael J. Wysocki
2017-03-19 13:34 ` [RFC][PATCH 2/2] cpufreq: schedutil: Force max frequency on busy CPUs Rafael J. Wysocki
2017-03-19 21:24 ` Rafael J. Wysocki
2017-03-19 21:42 ` Rafael J. Wysocki
2017-03-20 10:38 ` Peter Zijlstra
2017-03-20 12:31 ` Rafael J. Wysocki
2017-03-20 3:57 ` Viresh Kumar
2017-03-20 8:26 ` Vincent Guittot
2017-03-20 12:34 ` Patrick Bellasi
2017-03-22 23:56 ` Joel Fernandes
2017-03-23 22:08 ` Vincent Guittot
2017-03-25 3:48 ` Joel Fernandes
2017-03-27 6:59 ` Vincent Guittot
2017-03-20 12:59 ` Rafael J. Wysocki
2017-03-20 13:20 ` Vincent Guittot
2017-03-20 12:48 ` Rafael J. Wysocki
2017-03-20 10:36 ` Peter Zijlstra
2017-03-20 12:35 ` Rafael J. Wysocki
2017-03-20 12:50 ` Peter Zijlstra
2017-03-20 13:04 ` Rafael J. Wysocki
2017-03-20 13:06 ` Patrick Bellasi
2017-03-20 13:05 ` Rafael J. Wysocki
2017-03-20 14:13 ` Patrick Bellasi
2017-03-20 21:46 ` [RFC][PATCH v2 2/2] cpufreq: schedutil: Avoid decreasing frequency of " Rafael J. Wysocki
2017-03-21 6:40 ` Viresh Kumar
2017-03-21 12:30 ` Rafael J. Wysocki
2017-03-21 8:50 ` Vincent Guittot
2017-03-21 11:56 ` Patrick Bellasi
2017-03-21 13:22 ` Peter Zijlstra
2017-03-21 13:37 ` Vincent Guittot
2017-03-21 14:03 ` Peter Zijlstra
2017-03-21 14:18 ` Vincent Guittot
2017-03-21 14:25 ` Patrick Bellasi
[not found] ` <CAKfTPtALorn7HNpz4LOfWWSc3u+9y5iHB5byzfTHGQXDA+tVJQ@mail.gmail.com>
2017-03-21 14:58 ` Peter Zijlstra
2017-03-21 17:00 ` Vincent Guittot
2017-03-21 17:01 ` Vincent Guittot
2017-03-21 14:26 ` Rafael J. Wysocki
2017-03-21 14:38 ` Patrick Bellasi
2017-03-21 14:46 ` Rafael J. Wysocki
2017-03-21 14:50 ` Rafael J. Wysocki
2017-03-21 15:04 ` Peter Zijlstra
2017-03-21 15:18 ` Rafael J. Wysocki
2017-03-21 17:00 ` Peter Zijlstra
2017-03-21 17:17 ` Rafael J. Wysocki
2017-03-21 15:08 ` Patrick Bellasi
2017-03-21 15:18 ` Peter Zijlstra
2017-03-21 19:28 ` Patrick Bellasi
2017-03-21 15:02 ` Peter Zijlstra
2017-03-21 11:50 ` Patrick Bellasi
2017-03-21 23:08 ` [RFC][PATCH v3 2/2] cpufreq: schedutil: Avoid reducing frequency of busy CPUs prematurely Rafael J. Wysocki
2017-03-22 9:26 ` Peter Zijlstra
2017-03-22 9:54 ` Viresh Kumar
2017-03-23 1:04 ` Joel Fernandes
2017-03-23 19:26 ` Sai Gurrappadi
2017-03-23 20:48 ` Sai Gurrappadi
2017-03-24 1:39 ` Rafael J. Wysocki
2017-03-24 19:08 ` Sai Gurrappadi
2017-03-25 1:14 ` Sai Gurrappadi
2017-03-25 1:39 ` Rafael J. Wysocki
2017-03-27 7:04 ` Vincent Guittot [this message]
2017-03-27 21:01 ` Sai Gurrappadi
2017-03-27 21:11 ` Rafael J. Wysocki
2017-05-08 3:49 ` Wanpeng Li
2017-05-08 4:01 ` Viresh Kumar
2017-05-08 5:15 ` Wanpeng Li
2017-05-08 22:16 ` Rafael J. Wysocki
2017-05-08 22:36 ` Wanpeng Li
2017-05-08 23:01 ` Rafael J. Wysocki
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=CAKfTPtDAVZg--KchHN9S3hdgRB9OzQKJTR7ioCh8hTksRDDekg@mail.gmail.com \
--to=vincent.guittot@linaro.org \
--cc=joelaf@google.com \
--cc=juri.lelli@arm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-pm@vger.kernel.org \
--cc=mingo@redhat.com \
--cc=morten.rasmussen@arm.com \
--cc=patrick.bellasi@arm.com \
--cc=pboonstoppel@nvidia.com \
--cc=peterz@infradead.org \
--cc=rjw@rjwysocki.net \
--cc=sgurrappadi@nvidia.com \
--cc=srinivas.pandruvada@linux.intel.com \
--cc=tglx@linutronix.de \
--cc=viresh.kumar@linaro.org \
/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;
as well as URLs for NNTP newsgroup(s).