[<prev] [next>] [day] [month] [year] [list]
Message-ID: <20220916145100.GC29395@ranerica-svr.sc.intel.com>
Date: Fri, 16 Sep 2022 07:51:01 -0700
From: Ricardo Neri <ricardo.neri-calderon@...ux.intel.com>
To: Hillf Danton <hdanton@...a.com>
Cc: Peter Zijlstra <peterz@...radead.org>,
Mel Gorman <mgorman@...e.de>, linux-kernel@...r.kernel.org,
Vincent Guittot <vincent.guittot@...aro.org>
Subject: Re: [RFC PATCH 06/23] sched/core: Update the classification of the
current task
On Sat, Sep 10, 2022 at 03:21:20PM +0800, Hillf Danton wrote:
> On 9 Sep 2022 16:11:48 -0700 Ricardo Neri <ricardo.neri-calderon@...ux.intel.com> wrote:
> > @@ -5445,6 +5445,9 @@ void scheduler_tick(bool user_tick)
> > unsigned long thermal_pressure;
> > u64 resched_latency;
> >
> > + if (sched_task_classes_enabled() && user_tick)
> > + arch_update_task_class(curr, is_core_idle(cpu));
> > +
> > arch_scale_freq_tick();
> > sched_clock_tick();
>
Thank you very much for your feedback Hillf!
> Given user_tick == true, core is not idle regardless of SMT.
> IOW I doubt is_core_idle() helps here.
Perhaps is_core_idle() is a bad name? The second argument of arch_update_
task_class() is smt_siblings_idle. is_core_idle() gives us the answer
we want.
is_core_idle() only checks the siblings of @cpu. It explicitly skips itself
from the checks of idle state. We are only interested in the idle state of
the siblings.
Am I missing anything?
Thanks and BR,
Ricardo
Powered by blists - more mailing lists