From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1751839AbcERGls (ORCPT <rfc822;w@1wt.eu>);
	Wed, 18 May 2016 02:41:48 -0400
Received: from mail-wm0-f68.google.com ([74.125.82.68]:32989 "EHLO
	mail-wm0-f68.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1751257AbcERGlr (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Wed, 18 May 2016 02:41:47 -0400
Message-ID: <1463553703.4012.29.camel@gmail.com>
Subject: Re: sched: tweak select_idle_sibling to look for idle threads
From: Mike Galbraith <umgwanakikbuti@gmail.com>
To: Yuyang Du <yuyang.du@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>, Chris Mason <clm@fb.com>,
        Ingo Molnar <mingo@kernel.org>,
        Matt Fleming <matt@codeblueprint.co.uk>, linux-kernel@vger.kernel.org
Date: Wed, 18 May 2016 08:41:43 +0200
In-Reply-To: <20160510191646.GA4870@intel.com>
References: <1462694935.4155.83.camel@suse.de>
	 <20160508185747.GL16093@intel.com> <1462765540.3803.44.camel@suse.de>
	 <20160508202201.GM16093@intel.com> <1462779853.3803.128.camel@suse.de>
	 <20160509011311.GQ16093@intel.com> <1462786745.3803.181.camel@suse.de>
	 <20160509232623.GR16093@intel.com> <1462866562.3702.33.camel@suse.de>
	 <1462893965.3702.56.camel@gmail.com> <20160510191646.GA4870@intel.com>
Content-Type: text/plain; charset="UTF-8"
X-Mailer: Evolution 3.16.5 
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On Wed, 2016-05-11 at 03:16 +0800, Yuyang Du wrote:

> > --- a/kernel/sched/fair.c
> > +++ b/kernel/sched/fair.c
> > @@ -3027,6 +3027,9 @@ void remove_entity_load_avg(struct sched
> >  
> >  static inline unsigned long cfs_rq_runnable_load_avg(struct cfs_rq *cfs_rq)
> >  {
> > +> > 	> > if (sched_feat(LB_TIP_AVG_HIGH) && cfs_rq->load.weight > cfs_rq->runnable_load_avg*2)
> > +> > 	> > 	> > return cfs_rq->runnable_load_avg + min_t(unsigned long, NICE_0_LOAD,
> > +> > 	> > 	> > 	> > 	> > 	> > 	> > 	> >  cfs_rq->load.weight/2);
> >  > > 	> > return cfs_rq->runnable_load_avg;
> >  }
>   
> cfs_rq->runnable_load_avg is for sure no greater than (in this case much less
> than, maybe 1/2 of) load.weight, whereas load_avg is not necessarily a rock
> in gearbox that only impedes speed up, but also speed down.

BTW, the reason hack helped is that the long (30ms) sleep/run cycle of
the benchmark's default settings causes large amplitude sawtooth of
load numbers (~300 - ~700 range), dinging up load delta resolvability.

	-Mike