From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261313AbVFAHYP (ORCPT ); Wed, 1 Jun 2005 03:24:15 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261316AbVFAHYP (ORCPT ); Wed, 1 Jun 2005 03:24:15 -0400 Received: from mo.optusnet.com.au ([203.10.68.101]:19121 "EHLO mo.optusnet.com.au") by vger.kernel.org with ESMTP id S261313AbVFAHYL (ORCPT ); Wed, 1 Jun 2005 03:24:11 -0400 To: Andi Kleen Cc: Denis Vlasenko , dean gaudet , Jeff Garzik , Benjamin LaHaise , linux-kernel@vger.kernel.org Subject: Re: [RFC] x86-64: Use SSE for copy_page and clear_page References: <20050530181626.GA10212@kvack.org> <20050530193225.GC25794@muc.de> <200505311137.00011.vda@ilport.com.ua> <200505311215.06495.vda@ilport.com.ua> <20050531092358.GA9372@muc.de> From: michael@optusnet.com.au Date: 01 Jun 2005 17:22:28 +1000 In-Reply-To: <20050531092358.GA9372@muc.de> Message-ID: User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.3 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Andi Kleen writes: > > Thus with "normal" page clear and "nt" page copy routines > > both clear and copy benchmarks run faster than with > > stock kernel, both with small and large working set. > > > > Am I wrong? > > fork is only a corner case. The main case is a process allocating > memory using brk/mmap and then using it. Key point: "using it". This normally involves writes to memory. Most applications don't commonly read memory that they haven't previously written to. (valgrind et al call that behaviour a "bug" :). Given that, I'd say you really don't want the page zero routines touching the cache. Michael.