From mboxrd@z Thu Jan 1 00:00:00 1970 From: Peter Zijlstra Subject: Re: [PATCH v3 4/6] add a new scheduler hook for context switch Date: Wed, 30 May 2012 13:40:51 +0200 Message-ID: <1338378051.26856.267.camel@twins> References: <1338371317-5980-1-git-send-email-glommer@parallels.com> <1338371317-5980-5-git-send-email-glommer@parallels.com> <1338376851.26856.260.camel@twins> Mime-Version: 1.0 Content-Transfer-Encoding: 7BIT Return-path: In-Reply-To: <1338376851.26856.260.camel@twins> Sender: cgroups-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-ID: Content-Type: text/plain; charset="us-ascii" To: Glauber Costa Cc: linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, devel-GEFAQzZX7r8dnm+yROfE0A@public.gmane.org, Paul Turner , Tejun Heo , "Eric W. Biederman" , handai.szj-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org, Andrew.Phillips-xheW4WVAX9Y@public.gmane.org, Serge Hallyn On Wed, 2012-05-30 at 13:20 +0200, Peter Zijlstra wrote: > > $ perf stat -e cycles --repeat 100 perf bench sched messaging -p -g 100 > > 48,826,146,710 cycles # 2.470 GHz ( +- 0.17% ) > 2.149005270 seconds time elapsed ( +- 0.12% ) A repeat count of 250 dropped it down to: 2.144582157 seconds time elapsed ( +- 0.08% ) Also, if you're poking at the context switch path, something like: $ taskset 1 perf stat -e cycles --repeat 10 perf bench sched pipe -l 100000 gives a good number and is usually more stable than hackbench.