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;
MIME-Version: 1.0
References: <20211008000825.1364224-1-joshdon@google.com> <20211009155435.GW174703@worktop.programming.kicks-ass.net>
 <CABk29Nu6F4__ryF5p0En--Ze6CCev1Jy81W=LkTYaacf-YLkFg@mail.gmail.com> <YWV/HNDJaIAOLdrt@hirez.programming.kicks-ass.net>
In-Reply-To: <YWV/HNDJaIAOLdrt@hirez.programming.kicks-ass.net>
From:   Josh Don <joshdon@google.com>
Date:   Tue, 12 Oct 2021 12:45:28 -0700
Message-ID: <CABk29NucE__6r3P64Ts3Nbf4sUy5Zkw1sbNNnab9KZ=68ydy=w@mail.gmail.com>
Subject: Re: [PATCH] sched/core: forced idle accounting
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     Ingo Molnar <mingo@redhat.com>, Juri Lelli <juri.lelli@redhat.com>,
        Vincent Guittot <vincent.guittot@linaro.org>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        Ben Segall <bsegall@google.com>, Mel Gorman <mgorman@suse.de>,
        Daniel Bristot de Oliveira <bristot@redhat.com>,
        Joel Fernandes <joel@joelfernandes.org>,
        Vineeth Pillai <vineethrp@gmail.com>,
        Hao Luo <haoluo@google.com>,
        linux-kernel <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, Oct 12, 2021 at 5:27 AM Peter Zijlstra <peterz@infradead.org> wrote:
>
> On Mon, Oct 11, 2021 at 05:12:43PM -0700, Josh Don wrote:
>
> > > > +             if (WARN_ON_ONCE(!nr_running)) {
> > > > +                     /* can't be forced idle without a running task */
> > > > +             } else {
> > > > +                     delta *= nr_forced_idle;
> > > > +                     delta /= nr_running;
> > > > +             }
> > >
> > > Now the comment sayeth:
> > >
> > > > +     /*
> > > > +      * For larger SMT configurations, we need to scale the charged
> > > > +      * forced idle amount since there can be more than one forced idle
> > > > +      * sibling and more than one running cookied task.
> > > > +      */
> > >
> > > But why?
> >
> > We scale by the number of cpus actually forced idle, since we don't
> > want to falsely over or under charge forced idle time (defined
> > strictly as time where we have a runnable task but idle the cpu). The
> > more important scaling here though is the division over the number of
> > running entities. This is done so that the aggregate amount of forced
> > idle over some group of threads makes sense. Ie if we have a cpu with
> > SMT8, and a group of 7 threads sharing a cookie, we don't want to
> > accrue 7 units of forced idle time per unit time while the 8th SMT is
> > forced idle.
>
> So why not simply compute the strict per-cpu force-idle time and let
> userspace sort out the rest?

Do you mean to compute force idle solely as a per-cpu value? I think
that would be fine in addition to the per-thread field, but a
desirable property here is proper attribution to the cause of the
force idle. That lets system management understand which jobs are the
most antagonistic from a coresched perspective, and is a signal
(albeit noisy, due to system state and load balancing decisions) for
scaling their capacity requirements.