From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S264346AbTFPVfn (ORCPT ); Mon, 16 Jun 2003 17:35:43 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S264358AbTFPVfm (ORCPT ); Mon, 16 Jun 2003 17:35:42 -0400 Received: from e32.co.us.ibm.com ([32.97.110.130]:40351 "EHLO e32.co.us.ibm.com") by vger.kernel.org with ESMTP id S264346AbTFPVfi convert rfc822-to-8bit (ORCPT ); Mon, 16 Jun 2003 17:35:38 -0400 Content-Type: text/plain; charset=US-ASCII From: James Cleverdon Reply-To: jamesclv@us.ibm.com Organization: IBM xSeries Linux Solutions To: Mark Watts , linux-kernel@vger.kernel.org Subject: Re: Xeon processors &&Hyper-Threading Date: Mon, 16 Jun 2003 14:49:27 -0700 User-Agent: KMail/1.4.3 References: <3EE9FDFA.6020803@mindspring.com> <200306141450.11804.m.watts@mrw.demon.co.uk> In-Reply-To: <200306141450.11804.m.watts@mrw.demon.co.uk> MIME-Version: 1.0 Content-Transfer-Encoding: 7BIT Message-Id: <200306161449.27121.jamesclv@us.ibm.com> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Saturday 14 June 2003 06:50 am, Mark Watts wrote: > > You recompile the kernel for SMP as well as P4. If the motherboard > > hasn't disabled HT capabilities, you will take full advantage of > > the processor under Linux. Whatever "full advantage" means, is > > not absolute, but whatever it is, will be used to its fullest. > > Basically, if the code is I/O bound, you'll not see any difference. > > If the code is compute-intensive, you will. > > I discovered that you need the 'CPU Enumeration' part of ACPI to be enabled > otherwise the kernel only sees physical processors, not sibling HT > processors - shouldnt this be selected automatically when you select SMP ? > - Not if you have enabled full ACPI, or if you don't have any P4s. Some folks don't want any ACPI in their kernel at all. 8^) It's possible for 'CPU Enumeration Only' to be turned on auto-magically, but no one has written comprehensive enough logic into the config code, yet.... -- James Cleverdon IBM xSeries Linux Solutions {jamesclv(Unix, preferred), cleverdj(Notes)} at us dot ibm dot com