Re: [PATCH 2/2] thermal: cpufreq_cooling: Reuse effective_cpu_util()

From: "Rafael J. Wysocki" <rafael@kernel.org>
To: Viresh Kumar <viresh.kumar@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>,
	Ingo Molnar <mingo@redhat.com>,
	Vincent Guittot <vincent.guittot@linaro.org>,
	Zhang Rui <rui.zhang@intel.com>,
	Daniel Lezcano <daniel.lezcano@linaro.org>,
	Amit Daniel Kachhap <amit.kachhap@gmail.com>,
	Javi Merino <javi.merino@kernel.org>,
	Amit Kucheria <amit.kucheria@verdurent.com>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Quentin Perret <qperret@google.com>,
	Rafael Wysocki <rjw@rjwysocki.net>,
	Linux PM <linux-pm@vger.kernel.org>,
	Lukasz Luba <lukasz.luba@arm.com>
Subject: Re: [PATCH 2/2] thermal: cpufreq_cooling: Reuse effective_cpu_util()
Date: Thu, 22 Oct 2020 13:30:01 +0200	[thread overview]
Message-ID: <CAJZ5v0jZC=UwW9L+KB3pugsTL9P1tZmvQ-sVMV-udn7+L_gEeA@mail.gmail.com> (raw)
In-Reply-To: <20201022110656.gaphjv2tzhj4f5y6@vireshk-i7>

On Thu, Oct 22, 2020 at 1:07 PM Viresh Kumar <viresh.kumar@linaro.org> wrote:
>
> On 22-10-20, 11:05, Peter Zijlstra wrote:
> > On Thu, Oct 22, 2020 at 02:02:55PM +0530, Viresh Kumar wrote:
> > > One of the issues I see with this is that schedutil may not be
> > > available in all configurations and it is still absolutely fine to
> > > using the suggested helper to get the energy numbers in such cases, so
> > > we shouldn't really make it scheutil dependent.
> >
> > The only constraint on schedutil is SMP I think; aside from that it
> > should/could always be available.
> >
> > Given the trainwreck here:
> >
> >   20201022071145.GM2628@hirez.programming.kicks-ass.net
> >
> > (you're on Cc), I'm starting to lean more and more towards making it
> > unconditionally available (when SMP).
> >
> > Anybody forcing it off either sets performance (in which case we don't
> > care about energy usage anyway)
>
> I agree.

That's not really the case, though.

Many people use intel_pstate in the active mode with HWP enabled too.

Arguably, that doesn't need to compute the effective utilization, so I
guess it is not relevant for the discussion here, but it is not
negligible in general.

> > or they select one of the old (broken)
> > ondemand/conservative things and I don't give a crap.
>
> The other kernel layers, for example cpufreq-cooling in question here,
> don't really need to bother with the governor in use and should be
> able to get the energy numbers anyway. So for me, the energy number
> that the cpufreq-cooling stuff gets should be same irrespective of the
> governor in use, schedutil or ondemand.
>
> Having said that, schedutil really doesn't need to install the
> fallback (which you suggested earlier), rather the scheduler core can
> do that directly with cpufreq core and schedutil can also use the same
> fallback mechanism maybe ? And so we can avoid the exporting of stuff
> that way.

I like this better than the fallback idea TBH.