From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S264839AbUEYJvh (ORCPT ); Tue, 25 May 2004 05:51:37 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S264833AbUEYJvh (ORCPT ); Tue, 25 May 2004 05:51:37 -0400 Received: from mx1.elte.hu ([157.181.1.137]:17301 "EHLO mx1.elte.hu") by vger.kernel.org with ESMTP id S264827AbUEYJvf (ORCPT ); Tue, 25 May 2004 05:51:35 -0400 Date: Tue, 25 May 2004 13:42:58 +0200 From: Ingo Molnar To: Andrew Morton Cc: linux-kernel@vger.kernel.org Subject: Re: Help understanding slow down Message-ID: <20040525114258.GA6674@elte.hu> References: <20040524062754.GO1833@holomorphy.com> <20040524063959.5107.qmail@web90007.mail.scd.yahoo.com> <20040524005331.71465614.akpm@osdl.org> <20040525103238.GA4212@elte.hu> <20040525022941.62ab4cc4.akpm@osdl.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20040525022941.62ab4cc4.akpm@osdl.org> User-Agent: Mutt/1.4.1i X-ELTE-SpamVersion: MailScanner 4.26.8-itk2 (ELTE 1.1) SpamAssassin 2.63 ClamAV 0.65 X-ELTE-VirusStatus: clean X-ELTE-SpamCheck: no X-ELTE-SpamCheck-Details: score=-4.9, required 5.9, autolearn=not spam, BAYES_00 -4.90 X-ELTE-SpamLevel: X-ELTE-SpamScore: -4 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org * Andrew Morton wrote: > Ingo Molnar wrote: > > > > with the patch below we will print a big fat warning. (I did not want to > > deny idle=poll altogether - future HT implementations might work fine > > with polling idle.) > > idle=poll is handy when profiling the kernel with oprofile > clock-unhalted events. Because if you use the normal halt-based idle > loop no profile "ticks" are accounted to idle time at all and the > results are really hard to understand. it makes it a bit more plausible, but kernel profiling based on ticks in a HT environment is still quite unreliable, even with idle=poll. The HT cores will yield to each other on various occasions - like spinlock loops. This disproportionatly increases the hits of various looping functions, creating false impressions of lock contention where there's only little contention. Plus idle=poll is a constant ~20% performance drain on the non-idle HT core, further distorting the profile. HT makes profiling really hard, no matter what. but ... we agree on the warning printk, right? Ingo