Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Fri, 18 Dec 2020 16:32:31 +0100
From:   Peter Zijlstra <peterz@infradead.org>
To:     Doug Smythies <dsmythies@telus.net>
Cc:     "'Rafael J. Wysocki'" <rafael@kernel.org>,
        'Giovanni Gherdovich' <ggherdovich@suse.com>,
        "'Rafael J. Wysocki'" <rjw@rjwysocki.net>,
        'Linux PM' <linux-pm@vger.kernel.org>,
        'LKML' <linux-kernel@vger.kernel.org>,
        'Viresh Kumar' <viresh.kumar@linaro.org>,
        'Srinivas Pandruvada' <srinivas.pandruvada@linux.intel.com>,
        efault@gmx.de, Mel Gorman <mgorman@suse.de>,
        Vincent Guittot <vincent.guittot@linaro.org>,
        Ingo Molnar <mingo@kernel.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Juri Lelli <juri.lelli@gmail.com>
Subject: Re: [PATCH v1 0/4] cpufreq: Allow drivers to receive more
 information from the governor
Message-ID: <20201218153231.GB2414@hirez.programming.kicks-ass.net>
References: <20360841.iInq7taT2Z@kreacher>
 <1607445035.2673.64.camel@suse.com>
 <CAJZ5v0j2u7MrO82+ubx01kvyhDUKo11mfyofF-TAqdSLx_i3Ng@mail.gmail.com>
 <001d01d6cd96$601c8a80$20559f80$@net>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <001d01d6cd96$601c8a80$20559f80$@net>
Precedence: bulk

On Tue, Dec 08, 2020 at 11:14:36AM -0800, Doug Smythies wrote:
> At least on my system, it is most evident for some of the pipe type tests,
> where the schedutil governor has never really known what to do. This patch
> set seems to add enough of a downward bias that this version of the schedutil
> governor now behaves much like the other versions

Yeah, pipe relies on task-task interaction, where one task waits on
another, and by boosting the producer the consumer can start earlier and
we get more cycles done etc.. Rather similar to IO-wait, where by
boosting the producer we gain throughput.

schedutil doesn't track anything useful here, but it is a semi common
pattern and it would be really good if we could somehow fix this.

We obviously have access to the task A wakes task B information, but I'm
not sure what to do with it, we're tried some things like this in the
past (although for slightly different reasons) and they've always ended
up being a mess :/