From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1750998AbWF3WtG (ORCPT ); Fri, 30 Jun 2006 18:49:06 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751207AbWF3WtF (ORCPT ); Fri, 30 Jun 2006 18:49:05 -0400 Received: from [212.33.163.222] ([212.33.163.222]:6411 "EHLO raad.intranet") by vger.kernel.org with ESMTP id S1751549AbWF3WtE (ORCPT ); Fri, 30 Jun 2006 18:49:04 -0400 From: Al Boldi To: linux-kernel@vger.kernel.org Subject: Re: [PATCH -mm] ide_end_drive_cmd(): avoid instruction pipeline stall Date: Sat, 1 Jul 2006 01:49:50 +0300 User-Agent: KMail/1.5 MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Message-Id: <200607010149.50573.a1426z@gawab.com> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Alan Cox wrote: > Ar Gwe, 2006-06-30 am 18:13 +0200, ysgrifennodd Andreas Mohr: > > Use an independently-formatted "unsigned int" for data instead of a > > restrictive "u16" to avoid instruction fetch pipeline stalls > > probably caused by the byte calculations later. I'm hitting this all the time on 2.6 more so than on 2.4, especially for CPU bound procs. > 1. This is a gcc problem Maybe, but on 2.6, echo 0 > /proc/sys/kernel/randomize_va_space seems to improve the situation most of the time, although not always. > 2. Not everyone is using an intel x86-32 box which has such problems Specifically, 3/4/586 don't show this problem, whereas 686+ show the problem to varying degrees with slowdown upto 333% on P4. Can you confirm this? Thanks! -- Al