From owner-freebsd-performance@FreeBSD.ORG Tue Aug 30 22:21:03 2011 Return-Path: Delivered-To: freebsd-performance@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 4D8201065672; Tue, 30 Aug 2011 22:21:03 +0000 (UTC) (envelope-from ohartman@zedat.fu-berlin.de) Received: from outpost1.zedat.fu-berlin.de (outpost1.zedat.fu-berlin.de [130.133.4.66]) by mx1.freebsd.org (Postfix) with ESMTP id BD4078FC14; Tue, 30 Aug 2011 22:21:02 +0000 (UTC) Received: from inpost2.zedat.fu-berlin.de ([130.133.4.69]) by outpost1.zedat.fu-berlin.de (Exim 4.69) with esmtp (envelope-from ) id <1QyWg1-00011V-PI>; Wed, 31 Aug 2011 00:21:01 +0200 Received: from e178029245.adsl.alicedsl.de ([85.178.29.245] helo=thor.walstatt.dyndns.org) by inpost2.zedat.fu-berlin.de (Exim 4.69) with esmtpsa (envelope-from ) id <1QyWg1-0007Pz-KH>; Wed, 31 Aug 2011 00:21:01 +0200 Message-ID: <4E5D624D.4050109@zedat.fu-berlin.de> Date: Wed, 31 Aug 2011 00:21:01 +0200 From: "Hartmann, O." User-Agent: Mozilla/5.0 (X11; FreeBSD amd64; rv:6.0) Gecko/20110825 Thunderbird/6.0 MIME-Version: 1.0 To: "K. Macy" References: <1314733660.22981.YahooMailClassic@web113503.mail.gq1.yahoo.com> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 8bit X-Originating-IP: 85.178.29.245 Cc: "freebsd-performance@freebsd.org" , giffunip@tutopia.com, "b. f." , freebsd-current@freebsd.org Subject: Re: PHORONIX: OpenCL, GLSL Back-End For LLVM May Soon Open Up X-BeenThere: freebsd-performance@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Performance/tuning List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 30 Aug 2011 22:21:03 -0000 On 08/30/11 22:14, K. Macy wrote: > On Tue, Aug 30, 2011 at 9:47 PM, Pedro F. Giffuni wrote: >> FWIW; >> >> Christopher Bergström and Pathscale delivered the EKOPath >> Compiler Suite, but no one followed up: >> >> (From the WantedPorts Wiki) >> https://github.com/pathscale/path64-suite >> >> There has been very low interest in the FreeBSD port, >> and unfortunately this is a bad signal that we give to >> companies that want to contribute :(. >> > The problem I have with that is that they only support the high-end > computing variant of the card which I doubt any of us has. Without the > documentation to extend the work to ordinary cards, e.g. my GTX460, it > isn't that useful. > > Thanks > _______________________________________________ > freebsd-current@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-current > To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org" Well, C. Bergström wrote that the major focus is on the TESLA card, but there may be chance that it also is working on the GTX570 or GTX580 (or even the GTX 560Ti, which is also spread around our department for GPGPU stuff). But I never had a chance to test this. I'm no developer and my software is also "copied" and home-brewn, but I think it was a chance to do a real life test. Since April we run a server with a TESLA M2050, the card the Enzo claims to support. But my time is up by the end of this year and I need to get finnished my thesis so we decided to install Linux for CUDA usage. I left two spare partitions for FreeBSD ... but I never got a call ... Well, if they offer support for expensive card in the first place and we would show and assure that there is potential for the consumer cards, i guess then there will be a motivation to do the development that direction. In this case we, the customer, have to be active! Fact is, that most departments I know (mainly nuclear physics, theoretical meteorologists, astrophysicists and planetologists) do not have high-end GPU computing cards, all of them do have a lot of smaller consumer cards and they run CUDA with incredible results in performance! Well, a TESLA M2050 is much faster, but is is four to five times more the money I have to spend for the low end cards. I think this fact is also known by PatScale as well as it is obvious to me. But if the department realizes that there is a softare/compiler, than their willing to spend more money on maybe a professional card is less "frictional". Well, as the advert of PathScales EKOpath compiler promised, there would also be a CUDA support. Could you imagine what kind of Mekka this could be, simply compile CUDA stuff on a FreeBSD box and also having the ability to code the same stuff with HMPP in a next, optimizing step? Well, now I learn to code OpenCL and CUDA and maybe I'm back soon for HMPP. But it seems quite sure that none of the GPGPU stuff will soon be usable on a FreeBSD box. So why avoiding further Linux .... :-(