From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from psmtp.com (na3sys010amx200.postini.com [74.125.245.200]) by kanga.kvack.org (Postfix) with SMTP id 47C026B005A for ; Fri, 29 Jun 2012 16:01:53 -0400 (EDT) Received: by lbjn8 with SMTP id n8so6538009lbj.14 for ; Fri, 29 Jun 2012 13:01:51 -0700 (PDT) MIME-Version: 1.0 In-Reply-To: <1340995986.28750.114.camel@twins> References: <1340888180-15355-1-git-send-email-aarcange@redhat.com> <1340888180-15355-14-git-send-email-aarcange@redhat.com> <1340895238.28750.49.camel@twins> <20120629125517.GD32637@gmail.com> <4FEDDD0C.60609@redhat.com> <1340995986.28750.114.camel@twins> Date: Sat, 30 Jun 2012 04:01:50 +0800 Message-ID: Subject: Re: [PATCH 13/40] autonuma: CPU follow memory algorithm From: Nai Xia Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Sender: owner-linux-mm@kvack.org List-ID: To: Peter Zijlstra Cc: dlaor@redhat.com, Ingo Molnar , Hillf Danton , Andrea Arcangeli , linux-kernel@vger.kernel.org, linux-mm@kvack.org, Dan Smith , Linus Torvalds , Andrew Morton , Thomas Gleixner , Ingo Molnar , Paul Turner , Suresh Siddha , Mike Galbraith , "Paul E. McKenney" , Lai Jiangshan , Bharata B Rao , Lee Schermerhorn , Rik van Riel , Johannes Weiner , Srivatsa Vaddagiri , Christoph Lameter , Alex Shi , Mauricio Faria de Oliveira , Konrad Rzeszutek Wilk , Don Morris , Benjamin Herrenschmidt On Sat, Jun 30, 2012 at 2:53 AM, Peter Zijlstra wr= ote: > On Fri, 2012-06-29 at 12:51 -0400, Dor Laor wrote: >> The previous comments were not shouts but the mother of all NAKs. > > I never said any such thing. I just said why should I bother reading > your stuff if you're ignoring most my feedback anyway. > > If you want to read that as a NAK, not my problem. Hey guys, Can I say NAK to these patches ? Now I aware that this sampling algorithm is completely broken, if we take a few seconds to see what it is trying to solve: We all know that LRU is try to solve the question of "what are the pages recently accessed?", so its engouth to use pte bits to approximate. However, the numa balancing problem is fundamentally like this: In some time unit, W =3D pages_accessed * average_page_access_frequence We are trying to move process to the node having max W, right? Andrea's patch can only approximate the pages_accessed number in a time unit(scan interval), I don't think it can catch even 1% of average_page_access_frequence on a busy workload. Blindly assuming that all the pages' average_page_access_frequence is the same is seemly broken to me. Sometimes, it's good to have a good view of your problem before spending a lot time coding. > > -- > To unsubscribe, send a message with 'unsubscribe linux-mm' in > the body to majordomo@kvack.org. =A0For more info on Linux MM, > see: http://www.linux-mm.org/ . > Don't email: email@kvack.org -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org