Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220217154403.6497-1-wuyun.abel@bytedance.com>
 <20220224164711.GA4363@suse.de> <99805673-50a1-78cd-b710-0f00ce7de809@bytedance.com>
 <20220225101625.GB4363@suse.de> <6737ab1a-a794-7e06-fbb1-8f154c4e901b@bytedance.com>
In-Reply-To: <6737ab1a-a794-7e06-fbb1-8f154c4e901b@bytedance.com>
From:   Josh Don <joshdon@google.com>
Date:   Tue, 1 Mar 2022 16:41:57 -0800
Message-ID: <CABk29NsXc6J_zM+WaZxdjLtx_NC6+86VpMrtXrGZZ4WoQ_PPvg@mail.gmail.com>
Subject: Re: [RFC PATCH 0/5] introduce sched-idle balancing
To:     Abel Wu <wuyun.abel@bytedance.com>
Cc:     Mel Gorman <mgorman@suse.de>, Ben Segall <bsegall@google.com>,
        Daniel Bristot de Oliveira <bristot@redhat.com>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        Ingo Molnar <mingo@redhat.com>,
        Juri Lelli <juri.lelli@redhat.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Steven Rostedt <rostedt@goodmis.org>,
        Vincent Guittot <vincent.guittot@linaro.org>,
        linux-kernel <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, Feb 25, 2022 at 5:36 AM Abel Wu <wuyun.abel@bytedance.com> wrote:
[snip]
> > Also place the filter first and do any measurements of any change to
> > balancing versus the filter. I'm suggesting placing the filter first
> > because it's less controversial than a new balancer. Just be aware that
> > the filter alone is not a guarantee of merging as there have been a few
> > approaches to filtering and so far all of them had downsides on either cost
>
> Yes, understood. I will adjust the patches as you suggested and send v2
> together with more tests next week.

+1 to trying the filter rather than introducing a new balance path.

We've found the sched_idle_cpu() checks in the wakeup path to be
adequate in allowing non-idle tasks to fully consume cpu resources
(but that of course relies on wakeup balancing, and not periodic
balancing).

Please cc me on the next series.

Thanks,
Josh