From owner-freebsd-amd64@FreeBSD.ORG Wed Oct 25 12:13:42 2006 Return-Path: X-Original-To: freebsd-amd64@freebsd.org Delivered-To: freebsd-amd64@freebsd.org Received: from localhost.my.domain (localhost [127.0.0.1]) by hub.freebsd.org (Postfix) with ESMTP id 13CD116A494; Wed, 25 Oct 2006 12:13:42 +0000 (UTC) (envelope-from davidxu@freebsd.org) From: David Xu To: freebsd-amd64@freebsd.org Date: Wed, 25 Oct 2006 20:13:38 +0800 User-Agent: KMail/1.8.2 References: <102420062150.6756.453E8A9D000980D700001A6422007613940E999D0A07960B02019D@comcast.net> In-Reply-To: <102420062150.6756.453E8A9D000980D700001A6422007613940E999D0A07960B02019D@comcast.net> MIME-Version: 1.0 Content-Type: text/plain; charset="gb2312" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200610252013.38119.davidxu@freebsd.org> Cc: rondzierwa@comcast.net Subject: Re: Hardware topology X-BeenThere: freebsd-amd64@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Porting FreeBSD to the AMD64 platform List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 25 Oct 2006 12:13:42 -0000 On Wednesday 25 October 2006 05:50, rondzierwa@comcast.net wrote: > Does the amd64 kernel know anything about which memory is attached to > which processor, and which i/o bus is attached to which hypertransport > link? Can it use this information to do things like allocate pages to a > process from the memory that is physically attached to the cpu upon which > the process is running? Along these lines, is there any way to set > affinity between a process and a cpu (or set of cpu's in the case of > multicore)? > > Likewise with i/o devices, if a process or device driver wants to operate a > particular device, can it be set to run on the cpu that owns the > hypertransport connection upon which the device is connected? > > thanks, > ron. There is no unique hardware topology structure in kernel, but scheduler has some APIs can bind thread to a specific CPU, though there is no any syscall can let you do it. David Xu