From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753367Ab0H0PuR (ORCPT ); Fri, 27 Aug 2010 11:50:17 -0400 Received: from mail.openrapids.net ([64.15.138.104]:39398 "EHLO blackscsi.openrapids.net" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1751561Ab0H0PuP (ORCPT ); Fri, 27 Aug 2010 11:50:15 -0400 Date: Fri, 27 Aug 2010 11:50:13 -0400 From: Mathieu Desnoyers To: Thomas Gleixner Cc: Peter Zijlstra , LKML , Linus Torvalds , Andrew Morton , Ingo Molnar , Steven Rostedt , Tony Lindgren , Mike Galbraith Subject: Re: [RFC PATCH 00/11] sched: CFS low-latency features Message-ID: <20100827155013.GF14926@Krystal> References: <20100826180908.648103531@efficios.com> <1282849045.1975.1587.camel@laptop> <20100826230934.GA4194@Krystal> <20100826233651.GB28177@Krystal> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: X-Editor: vi X-Info: http://www.efficios.com X-Operating-System: Linux/2.6.26-2-686 (i686) X-Uptime: 11:44:33 up 216 days, 18:21, 5 users, load average: 0.11, 0.07, 0.03 User-Agent: Mutt/1.5.18 (2008-05-17) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org * Thomas Gleixner (tglx@linutronix.de) wrote: > On Thu, 26 Aug 2010, Mathieu Desnoyers wrote: [...] > > Hrm, thinking about it a little more, one of the "plus" sides of these > > SIGEV_THREAD timers is that a single timer can fork threads that will run on > > many cores on a multi-core system. If we go for preallocation of a single > > thread, we lose that. Maybe we could think of a way to preallocate a thread pool > > instead ? > > Why should a single timer fork many threads? Just because a previous > thread did not complete before the timer fires again? That's > braindamage as all threads call the same function which then needs to > be serialized anyway. We really do not need a function which creates > tons of threads which get all stuck on the same resource. It could make sense if the workload is mostly CPU-bound and there is only a very short critical section shared between the threads. But I agree that in many cases this will generate an utter contention mess. Thanks, Mathieu -- Mathieu Desnoyers Operating System Efficiency R&D Consultant EfficiOS Inc. http://www.efficios.com