From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752176AbeA3M7X (ORCPT ); Tue, 30 Jan 2018 07:59:23 -0500 Received: from outbound-smtp27.blacknight.com ([81.17.249.195]:44242 "EHLO outbound-smtp27.blacknight.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751577AbeA3M7W (ORCPT ); Tue, 30 Jan 2018 07:59:22 -0500 Date: Tue, 30 Jan 2018 12:59:20 +0000 From: Mel Gorman To: Peter Zijlstra Cc: Mike Galbraith , Matt Fleming , LKML Subject: Re: [PATCH 4/4] sched/fair: Use a recently used CPU as an idle candidate and the basis for SIS Message-ID: <20180130125920.oi6eyvx6mwahh2oh@techsingularity.net> References: <20180130104555.4125-1-mgorman@techsingularity.net> <20180130104555.4125-5-mgorman@techsingularity.net> <20180130115349.GB2269@hirez.programming.kicks-ass.net> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Disposition: inline In-Reply-To: <20180130115349.GB2269@hirez.programming.kicks-ass.net> User-Agent: NeoMutt/20170912 (1.9.0) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Jan 30, 2018 at 12:53:49PM +0100, Peter Zijlstra wrote: > On Tue, Jan 30, 2018 at 10:45:55AM +0000, Mel Gorman wrote: > > The results can be less dramatic on NUMA where automatic balancing interferes > > with the test. It's also known that network benchmarks running on localhost > > also benefit quite a bit from this patch (roughly 10% on netperf RR for UDP > > and TCP depending on the machine). Hackbench also seens small improvements > > (6-11% depending on machine and thread count). The facebook schbench was also > > tested but in most cases showed little or no different to wakeup latencies. > > What cpufreq setting were you using for these tests? Default powersave settings in general -- intel_pstate driver is the one used most often but HWP was not always available. By and large, no special tuning was applied. Two sets of tests were run, one with turbostat and perf keeping track of p-states and migrations and one without to make sure the measured increase was real. The irony when hitting cpufreq-related problems is that monitoring can mask the issue by increasing overall utilisation. -- Mel Gorman SUSE Labs