Date: Mon, 5 Dec 2016 13:35:46 +0000
From: Matt Fleming <matt@codeblueprint.co.uk>
To: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>,
        Peter Zijlstra <peterz@infradead.org>, Ingo Molnar <mingo@kernel.org>,
        LKML <linux-kernel@vger.kernel.org>, Morten.Rasmussen@arm.com,
        dietmar.eggemann@arm.com, kernellwp@gmail.com, yuyang.du@intel.com,
        umgwanakikbuti@gmail.com, Mel Gorman <mgorman@techsingularity.net>
Subject: Re: [PATCH 2/2 v2] sched: use load_avg for selecting idlest group
Message-ID: <20161205133546.GN20785@codeblueprint.co.uk>
References: <1480088073-11642-1-git-send-email-vincent.guittot@linaro.org>
 <1480088073-11642-3-git-send-email-vincent.guittot@linaro.org>
 <CAE40pde4VRH8LfRWMX3Vfq5pRoysUB6UuHTzN+VrcKMdSNW0uA@mail.gmail.com>
 <20161203214707.GI20785@codeblueprint.co.uk>
 <20161205092735.GA9161@linaro.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20161205092735.GA9161@linaro.org>
User-Agent: Mutt/1.5.24+41 (02bc14ed1569) (2015-08-30)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1084
Lines: 31

On Mon, 05 Dec, at 10:27:36AM, Vincent Guittot wrote:
> 
> Hi Matt,
> 
> Thanks for the results.
> 
> During the review, it has been pointed out by Morten that the test condition
> (100*this_avg_load < imbalance_scale*min_avg_load) makes more sense than
> (100*min_avg_load > imbalance_scale*this_avg_load). But i see lower
> performances with this change. Coud you run tests with the change below on
> top of the patchset ?
> 
> ---
>  kernel/sched/fair.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index e8d1ae7..0129fbb 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -5514,7 +5514,7 @@ find_idlest_group(struct sched_domain *sd, struct task_struct *p,
>  	if (!idlest ||
>  	    (min_runnable_load > (this_runnable_load + imbalance)) ||
>  	    ((this_runnable_load < (min_runnable_load + imbalance)) &&
> -			(100*min_avg_load > imbalance_scale*this_avg_load)))
> +			(100*this_avg_load < imbalance_scale*min_avg_load)))
>  		return NULL;
>  	return idlest;
>  }

Queued for testing.