linux-kernel - Re: [PATCH V4 3/3] thermal: cpufreq_cooling: Reuse sched_cpu

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 7 Dec 2020 17:47:04 +0530
From:   Viresh Kumar <viresh.kumar@...aro.org>
To:     Dietmar Eggemann <dietmar.eggemann@....com>
Cc:     Ingo Molnar <mingo@...hat.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Amit Daniel Kachhap <amit.kachhap@...il.com>,
        Daniel Lezcano <daniel.lezcano@...aro.org>,
        Javi Merino <javi.merino@...nel.org>,
        Zhang Rui <rui.zhang@...el.com>,
        Amit Kucheria <amitk@...nel.org>, linux-kernel@...r.kernel.org,
        Quentin Perret <qperret@...gle.com>,
        Lukasz Luba <lukasz.luba@....com>, linux-pm@...r.kernel.org
Subject: Re: [PATCH V4 3/3] thermal: cpufreq_cooling: Reuse sched_cpu_util()
 for SMP platforms

On 03-12-20, 12:54, Dietmar Eggemann wrote:
> On 24/11/2020 07:26, Viresh Kumar wrote:
> > Several parts of the kernel are already using the effective CPU
> > utilization (as seen by the scheduler) to get the current load on the
> > CPU, do the same here instead of depending on the idle time of the CPU,
> > which isn't that accurate comparatively.
> > 
> > This is also the right thing to do as it makes the cpufreq governor
> > (schedutil) align better with the cpufreq_cooling driver, as the power
> > requested by cpufreq_cooling governor will exactly match the next
> > frequency requested by the schedutil governor since they are both using
> > the same metric to calculate load.
> > 
> > This was tested on ARM Hikey6220 platform with hackbench, sysbench and
> > schbench. None of them showed any regression or significant
> > improvements. Schbench is the most important ones out of these as it
> > creates the scenario where the utilization numbers provide a better
> > estimate of the future.
> > 
> > Scenario 1: The CPUs were mostly idle in the previous polling window of
> > the IPA governor as the tasks were sleeping and here are the details
> > from traces (load is in %):
> > 
> >  Old: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=203 load={{0x35,0x1,0x0,0x31,0x0,0x0,0x64,0x0}} dynamic_power=1339
> >  New: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=600 load={{0x60,0x46,0x45,0x45,0x48,0x3b,0x61,0x44}} dynamic_power=3960
> 
> When I ran schbench (-t 16 -r 5) on hikey960 I get multiple (~50)
> instances of ~80ms task activity phase and then ~20ms idle phase on all
> CPUs.
> 
> So I assume that scenario 1 is at the beginning (but you mentioned the
> task were sleeping?)

I am not able to find the exact values I used, but I did something
like this to create a scenario where the old computations shall find
the CPU as idle in the last IPA window:

- schbench -m 2 -t 4 -s 25000 -c 20000 -r 60

- sampling rate of IPA to 10 ms

With this IPA wakes up many times and finds the CPUs to have been idle
in the last IPA window (i.e. 10ms).

> and scenario 2 is somewhere in the middle of the
> testrun?

This also happens all the time, as there will be cases when the IPA
runs and finds the CPUs to be always running in last 10 ms.

> IMHO, the util-based approach delivers really better results at the
> beginning and at the end of the entire testrun.
> During the testrun, the util-based and the idle-based approach deliver
> similar results.
> 
> It's a little bit tricky to compare test results since the IPA sampling
> rate is 100ms and the load values you get depend on how the workload
> pattern and the IPA sampling align.

Right.

> > Here, the "Old" line gives the load and requested_power (dynamic_power
> > here) numbers calculated using the idle time based implementation, while
> > "New" is based on the CPU utilization from scheduler.
> > 
> > As can be clearly seen, the load and requested_power numbers are simply
> > incorrect in the idle time based approach and the numbers collected from
> > CPU's utilization are much closer to the reality.
> 
> I assume the IPA sampling is done after ~50ms of the first task activity
> phase.
> 
> > Scenario 2: The CPUs were busy in the previous polling window of the IPA
> > governor:
> > 
> >  Old: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=800 load={{0x64,0x64,0x64,0x64,0x64,0x64,0x64,0x64}} dynamic_power=5280
> >  New: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=708 load={{0x4d,0x5c,0x5c,0x5b,0x5c,0x5c,0x51,0x5b}} dynamic_power=4672
> > 
> > As can be seen, the idle time based load is 100% for all the CPUs as it
> > took only the last window into account, but in reality the CPUs aren't
> > that loaded as shown by the utilization numbers.
> 
> Is this an IPA sampling at the end of the ~20ms idle phase?

This is during the phase where the CPUs were fully busy for the last
period.

-- 
viresh