From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751703Ab1B0LyH (ORCPT ); Sun, 27 Feb 2011 06:54:07 -0500 Received: from mx3.mail.elte.hu ([157.181.1.138]:44683 "EHLO mx3.mail.elte.hu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750955Ab1B0LyE (ORCPT ); Sun, 27 Feb 2011 06:54:04 -0500 Date: Sun, 27 Feb 2011 12:53:50 +0100 From: Ingo Molnar To: Tejun Heo Cc: Yinghai Lu , Ingo Molnar , Thomas Gleixner , "H. Peter Anvin" , x86@kernel.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH 2/2] x86,mm,64bit: Round up memory boundary for init_memory_mapping_high() Message-ID: <20110227115350.GB16453@elte.hu> References: <4D656D1A.7030006@kernel.org> <20110223204656.GA27738@atj.dyndns.org> <4D657359.5060901@kernel.org> <20110223210326.GB27738@atj.dyndns.org> <20110224091557.GD7840@htj.dyndns.org> <4D674A33.8000809@kernel.org> <20110225111606.GG24828@htj.dyndns.org> <4D680EA4.6050402@kernel.org> <20110226085737.GT24828@htj.dyndns.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20110226085737.GT24828@htj.dyndns.org> User-Agent: Mutt/1.5.20 (2009-08-17) X-ELTE-SpamScore: -2.0 X-ELTE-SpamLevel: X-ELTE-SpamCheck: no X-ELTE-SpamVersion: ELTE 2.0 X-ELTE-SpamCheck-Details: score=-2.0 required=5.9 tests=BAYES_00 autolearn=no SpamAssassin version=3.2.5 -2.0 BAYES_00 BODY: Bayesian spam probability is 0 to 1% [score: 0.0000] Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org * Tejun Heo wrote: > > will have problem with cross node conf. like 0-4g, 8-12g on node0, 4g-8g, > > 12g-16g on node1. > > And how common are they? This whole cruft is basically meaningless if 1GiB > mapping is supported, IOW, basically on all AMD 64s and all post-nehalem intels. > Why not just cite the limitation in the comment and stick to something simple? Such complexity should be justified via very careful "perf stat --repeat" measurements. I.e. showing 'before patch' and 'after patch' instruction, TLB miss and cycle counts, showing that a positive effect that goes beyond the noise of the measurement exists. 1GB mappings should be assumed as the common case - anything else probably does not matter from a future performance/scalability POV. For vmalloc() it might make sense - but even for those precise measurements should be done about the positive effect. Thanks, Ingo