From owner-freebsd-amd64@FreeBSD.ORG Thu Jul 28 16:46:31 2005 Return-Path: X-Original-To: freebsd-amd64@freebsd.org Delivered-To: freebsd-amd64@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 097AE16A41F for ; Thu, 28 Jul 2005 16:46:31 +0000 (GMT) (envelope-from ray@redshift.com) Received: from outgoing.redshift.com (outgoing.redshift.com [207.177.231.8]) by mx1.FreeBSD.org (Postfix) with ESMTP id 94FE843D45 for ; Thu, 28 Jul 2005 16:46:30 +0000 (GMT) (envelope-from ray@redshift.com) Received: from workstation (216-228-19-21.dsl.redshift.com [216.228.19.21]) by outgoing.redshift.com (Postfix) with SMTP id 597D89791F; Thu, 28 Jul 2005 09:46:27 -0700 (PDT) Message-Id: <3.0.1.32.20050728094629.00a50b10@pop.redshift.com> X-Mailer: na X-Sender: redshift.com Date: Thu, 28 Jul 2005 09:46:29 -0700 To: Steve Kargl From: ray@redshift.com In-Reply-To: <20050728163602.GC64153@troutmask.apl.washington.edu> References: <3.0.1.32.20050728092338.01207628@pop.redshift.com> <3.0.1.32.20050728080526.00aa4098@pop.redshift.com> <86mzo7yvpe.fsf@xps.des.no> <3.0.1.32.20050728013152.00a4d188@pop.redshift.com> <86mzo7yvpe.fsf@xps.des.no> <3.0.1.32.20050728080526.00aa4098@pop.redshift.com> <3.0.1.32.20050728092338.01207628@pop.redshift.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Cc: freebsd-amd64@freebsd.org Subject: Re: Benchmarks: AMD64 vs i386 on Dual 246 Opteron 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: Thu, 28 Jul 2005 16:46:31 -0000 At 09:36 AM 7/28/2005 -0700, Steve Kargl wrote: | On Thu, Jul 28, 2005 at 09:23:38AM -0700, ray@redshift.com wrote: | > | | > | Drop 8 GB of memory into the box and see how the 32-bit | > | FreeBSD performs in comparison to the 64-bit FreeBSD | > | when your process consumes greater than 4GB of memory. | > | > True, once you go over the 4GB limit it's a different ball of wax. | > However, until that time, it would be nice to get to the bottom of | > why 64 bit code is running half the speed of 32 bit code on the exact | > same machine - don't you think? | | Well, I have 12 GB of memory and run numerical intensive codes | that easily can grab 4+ GB, so I've never explored i386 FreeBSD on | an amd64 system. | | As mentioned elsewhere, I would look for optimizations within | the software packages that target i386. Additionally, the | instruction schedulers in gcc/gas have had many more years | of development in comparison to the amd64 schedulers. | | An interesting test would be to build math/atlas on 32-bit | and 64-bit FreeBSD and then run some linpack benchmarks. All good points. Maybe part of it does boil down to the fact that AMD64 is new and hasn't had the mileage and hammering that i386 has had all right - that makes a lot of sense certainly. My main question relating to all of this is whether the 64 bit code should run as fast as the 32 bit code. If not, that's fine. But if it should, then - at least in my testing - there is something resulting in quite a delay when it comes to the AMD64 branch vs. the i386 branch, at least in relation to something like Apache and MySQL. Let me put it this way... I know if I was in your shoes, running apps that could only run on the 64 bit version of FreeBSD, I would want to know for sure that I wasn't giving up half my potential speed just in order to gain access to system memory above 4GB's. That's really what my concerns are - even though currently I do not have any servers that need require more than a couple of GB's of RAM. Ray