From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754218AbYGSJS3 (ORCPT ); Sat, 19 Jul 2008 05:18:29 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752300AbYGSJSW (ORCPT ); Sat, 19 Jul 2008 05:18:22 -0400 Received: from bombadil.infradead.org ([18.85.46.34]:54872 "EHLO bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751814AbYGSJSV (ORCPT ); Sat, 19 Jul 2008 05:18:21 -0400 Subject: Re: Problem: Load average of 6, no I/O, 25% idle time. Why? From: Peter Zijlstra To: Timothy Normand Miller Cc: linux-kernel@vger.kernel.org In-Reply-To: <9871ee5f0807181512u231e1aaraee5622d0a814f97@mail.gmail.com> References: <9871ee5f0807180549l58e849d9ge6e7deee3d0c37fa@mail.gmail.com> <1216387716.28405.45.camel@twins> <9871ee5f0807181512u231e1aaraee5622d0a814f97@mail.gmail.com> Content-Type: text/plain Date: Sat, 19 Jul 2008 11:18:16 +0200 Message-Id: <1216459096.7257.2.camel@twins> Mime-Version: 1.0 X-Mailer: Evolution 2.22.3.1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, 2008-07-18 at 18:12 -0400, Timothy Normand Miller wrote: > On Fri, Jul 18, 2008 at 9:28 AM, Peter Zijlstra wrote: > > On Fri, 2008-07-18 at 08:49 -0400, Timothy Normand Miller wrote: > >> I have a four processor box, running three compute loads. The kernel > >> I'm running is 2.6.25-gentoo-r5. > >> > >> One process has four threads, and it's niced to 19. > >> Another process has one thread and is niced to -10 > >> The last process has one thread and is niced to 0 > >> > >> One processor is always idle. > >> > >> If I change either the 0 to a -10 or the -10 to a zero, then all four > >> cores are properly utilized. > > Interesting corollary: If the only thing running is the one nice=19, > four-thread process, then there's about 4% idle time. I have to raise > the priority to get the idle time to go away. Since when do active > processes yield to idle time? They shouldn't... I'm suspecting it might be due to some form of priority inversion or some such thing. > Two of my three loads are finished, so when this last one is done, > I'll try 2.6.26. Was something like this actively discussed and fixed > between 2.6.25 and 2.6.26? Not much, but it has seen drastic change post .26. So if you could also try Linus' latest -git that would be good.