From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S263555AbUE2E7Y (ORCPT ); Sat, 29 May 2004 00:59:24 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S263609AbUE2E7W (ORCPT ); Sat, 29 May 2004 00:59:22 -0400 Received: from palrel13.hp.com ([156.153.255.238]:11422 "EHLO palrel13.hp.com") by vger.kernel.org with ESMTP id S263555AbUE2E7T (ORCPT ); Sat, 29 May 2004 00:59:19 -0400 From: David Mosberger MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <16568.6306.306924.154136@napali.hpl.hp.com> Date: Fri, 28 May 2004 21:59:14 -0700 To: nickpiggin@yahoo.com.au, akpm@osdl.org Cc: linux-kernel@vger.kernel.org Subject: big bw_pipe drop due to sched-domain patch? X-Mailer: VM 7.18 under Emacs 21.3.1 Reply-To: davidm@hpl.hp.com X-URL: http://www.hpl.hp.com/personal/David_Mosberger/ Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Today I noticed that going from 2.6.6 to 2.6.7-rc1 caused a big drop in LMbench2 bw_pipe throughput on a dual-CPU machine (2.6.7-rc1 shows 10 times lower bandwidth than 2.6.6). It appears that the drop is due to the two tasks being distributed across the two CPUs, rather than being kept on the same CPU. If I force the tasks to be pinned on a single CPU, e.g., like so: $ taskset 1 ./bw_pipe then performance is "only" about 10% worse than with 2.6.6. Is this kind of drop expected? --david