From mboxrd@z Thu Jan 1 00:00:00 1970 From: Catalin Marinas Subject: Re: [RFC][PATCH v5 00/14] sched: packing tasks Date: Tue, 12 Nov 2013 23:14:05 +0000 Message-ID: <42638CC1-ACC7-4330-A4F4-D78C88BE8155@arm.com> References: <1382097147-30088-1-git-send-email-vincent.guittot@linaro.org> <20131111163630.GD26898@twins.programming.kicks-ass.net> <52810851.4090907@linux.intel.com> <20131111181805.GE29572@arm.com> <52825BE9.2080605@linux.intel.com> Mime-Version: 1.0 (Mac OS X Mail 7.0 \(1822\)) Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Received: from mail-wi0-f172.google.com ([209.85.212.172]:56379 "EHLO mail-wi0-f172.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757316Ab3KLXOI convert rfc822-to-8bit (ORCPT ); Tue, 12 Nov 2013 18:14:08 -0500 In-Reply-To: <52825BE9.2080605@linux.intel.com> Sender: linux-pm-owner@vger.kernel.org List-Id: linux-pm@vger.kernel.org To: Arjan van de Ven Cc: Peter Zijlstra , Vincent Guittot , linux-kernel , Ingo Molnar , Paul Turner , Morten Rasmussen , Chris Metcalf , Tony Luck , "alex.shi@intel.com" , Preeti U Murthy , linaro-kernel , "len.brown@intel.com" , "l.majewski@samsung.com" , Jonathan Corbet , "Rafael J. Wysocki" , Paul McKenney , "linux-pm@vger.kernel.org" On 12 Nov 2013, at 16:48, Arjan van de Ven wrot= e: > On 11/11/2013 10:18 AM, Catalin Marinas wrote: >> The ordering is based on the actual C-state, so a simple way is to w= ake >> up the CPU in the shallowest C-state. With asymmetric configurations >> (big.LITTLE) we have different costs for the same C-state, so this w= ould >> come in handy. >=20 > btw I was considering something else; in practice CPUs will be in the= deepest state.. > ... at which point I was going to go with some other metrics of what = is best from a platform level I agree, other metrics are needed. The problem is that we currently only have (relatively, guessed from the target residency) the cost of transition from a C-state to a P-state (for the latter, not sure which)= =2E But we don=92t know what the power (saving) on that C-state is nor the = one=20 at a P-state (and vendors reluctant to provide such information). So th= e=20 best the scheduler can do is optimise the wake-up cost and blindly assu= me=20 that deeper C-state on a CPU is more efficient than lower P-states on t= wo=20 other CPUs (or the other way around). If we find a good use for such metrics in the scheduler, I think the vendors would be more open to providing at least some relative (rather than absolute) numbers. Catalin