Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <1535548752-4434-1-git-send-email-vincent.guittot@linaro.org>
 <1535548752-4434-4-git-send-email-vincent.guittot@linaro.org>
 <20180904082424.GA2090@linux.vnet.ibm.com> <20180904093626.GA23936@linaro.org>
 <20180904103742.GC61288@linux.vnet.ibm.com>
In-Reply-To: <20180904103742.GC61288@linux.vnet.ibm.com>
From:   Vincent Guittot <vincent.guittot@linaro.org>
Date:   Wed, 5 Sep 2018 09:36:42 +0200
Message-ID: <CAKfTPtBVLWgtRJpP0nTOVpkwJse6BwrGa28o4qat0T3kOvEAfw@mail.gmail.com>
Subject: Re: [RFC PATCH 3/4] sched/topology: remove smt_gain
To:     Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc:     Peter Zijlstra <peterz@infradead.org>,
        Ingo Molnar <mingo@redhat.com>,
        linux-kernel <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, 4 Sep 2018 at 12:37, Srikar Dronamraju
<srikar@linux.vnet.ibm.com> wrote:
>
> * Vincent Guittot <vincent.guittot@linaro.org> [2018-09-04 11:36:26]:
>
> > Hi Srikar,
> >
> > Le Tuesday 04 Sep 2018 =C3=A0 01:24:24 (-0700), Srikar Dronamraju a =C3=
=A9crit :
> > > However after this change, capacity_orig of each SMT thread would be
> > > 1024. For example SMT 8 core capacity_orig would now be 8192.
> > >
> > > smt_gain was suppose to make a multi threaded core was slightly more
> > > powerful than a single threaded core. I suspect if that sometimes hur=
t
> >
> > Is there system with both single threaded and multi threaded core ?
> > That was the main open point for me (and for Qais too)
> >
>
> I dont know of any systems that have come with single threaded and
> multithreaded. However some user can still offline few threads in a core
> while leaving other cores untouched. I dont really know why somebody
> would want to do it.  For example, some customer was toying with SMT 3
> mode in a SMT 8 power8 box.

In this case, it means that we have the same core capacity whatever
the number of CPUs
and a core with SMT 3 will be set with the same compute capacity as
the core with SMT 8.
Does it still make sense ?

>
> >
> > > us when doing load balance between 2 cores i.e at MC or DIE sched
> > > domain. Even with 2 threads running on a core, the core might look
> > > lightly loaded 2048/8192. Hence might dissuade movement to a idle cor=
e.
> >
> > Then, there is the sibling flag at SMT level that normally ensures 1 ta=
sk per
> > core for such UC
> >
>
> Agree.
>
> > >
> > > I always wonder why arch_scale_cpu_capacity() is called with NULL
> > > sched_domain, in scale_rt_capacity(). This way capacity might actuall=
y
> >
> > Probably because until this v4.19-rcxx version, the rt scaling was done
> > relatively to local cpu capacity:
> >   capacity  =3D arch_scale_cpu() * scale_rt_capacity / SCHED_CAPACITY_S=
CALE
> >
> > Whereas now, it directly returns the remaining capacity
> >
> > > be more than the capacity_orig. I am always under an impression that
> > > capacity_orig > capacity.  Or am I misunderstanding that?
> >
> > You are right, there is a bug for SMT and the patch below should fix it=
.
> > Nevertheless, we still have the problem in some other places in the cod=
e.
> >
> > Subject: [PATCH] sched/fair: fix scale_rt_capacity() for SMT
> >
> > Since commit:
> >   commit 523e979d3164 ("sched/core: Use PELT for scale_rt_capacity()")
> > scale_rt_capacity() returns the remaining capacity and not a scale fact=
or
> > to apply on cpu_capacity_orig. arch_scale_cpu() is directly called by
> > scale_rt_capacity() so we must take the sched_domain argument
> >
> > Fixes: 523e979d3164 ("sched/core: Use PELT for scale_rt_capacity()")
> > Reported-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
> > Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
>
> Reviewed-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
>
> --
> Thanks and Regards
> Srikar Dronamraju
>