From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261193AbTIONvC (ORCPT ); Mon, 15 Sep 2003 09:51:02 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261336AbTIONvB (ORCPT ); Mon, 15 Sep 2003 09:51:01 -0400 Received: from pc1-cwma1-5-cust4.swan.cable.ntl.com ([80.5.120.4]:62114 "EHLO dhcp23.swansea.linux.org.uk") by vger.kernel.org with ESMTP id S261193AbTIONum (ORCPT ); Mon, 15 Sep 2003 09:50:42 -0400 Subject: Re: [PATCH] 2.6 workaround for Athlon/Opteron prefetch errata From: Alan Cox To: Bill Davidsen Cc: Zwane Mwaikambo , Linux Kernel Mailing List In-Reply-To: References: Content-Type: text/plain Content-Transfer-Encoding: 7bit Message-Id: <1063633696.2674.20.camel@dhcp23.swansea.linux.org.uk> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.4 (1.4.4-6) Date: Mon, 15 Sep 2003 14:48:17 +0100 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Llu, 2003-09-15 at 13:11, Bill Davidsen wrote: > The code to disable prefetch on Athlon is 300 bytes and hurts your PIV? > Really? I'll dig back through the code, but I recall it as adding or > deleting an entry in a table to enable prefetch. If it's affecting PIV the > code to use prefetch is seriously broken. Big time. Its over 300 bytes long because its embedded in each inline prefetch and it causes a branch misprediction almost every time. Amazing what you find when you actually measure stuff 8) So right now, its faster on PIV to delete the prefetch than use the current hack, and adding the Athlon fix makes Athlon and PIV faster and total memory size lower.