From owner-freebsd-stable Wed Dec 22 14:53:19 1999 Delivered-To: freebsd-stable@freebsd.org Received: from dorifer.heim3.tu-clausthal.de (dorifer.heim3.tu-clausthal.de [139.174.243.252]) by hub.freebsd.org (Postfix) with ESMTP id B877614CEF for ; Wed, 22 Dec 1999 14:53:16 -0800 (PST) (envelope-from olli@dorifer.heim3.tu-clausthal.de) Received: (from olli@localhost) by dorifer.heim3.tu-clausthal.de (8.8.8/8.8.8) id XAA16064 for freebsd-stable@FreeBSD.ORG; Wed, 22 Dec 1999 23:53:13 +0100 (CET) (envelope-from olli) Date: Wed, 22 Dec 1999 23:53:13 +0100 (CET) From: Oliver Fromme Message-Id: <199912222253.XAA16064@dorifer.heim3.tu-clausthal.de> To: freebsd-stable@FreeBSD.ORG Subject: Re: AMD 3DNow instructions on FreeBSD Organization: Administration TU Clausthal Reply-To: freebsd-stable@FreeBSD.ORG MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit X-Newsreader: TIN [version 1.2 RZTUC(3) PL2] Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Sean-Paul Rees wrote in list.freebsd-stable: > Is there a MMX enabled setiathome for us with non-3DNow! enhanced > processors? No, MMX is completely useless for this kind of stuff, because MMX does not support floating point operations. Sorry. However, I modified the FFT algorithm to optimize memory throughput, at the cost of memory footprint (it needs about 2 Mbyte more RAM). The speed increase is not as impressive as with the 3DNow version, but it's still significant. Here are some numbers: Dual Celeron-466, Gigabyte 6BXD mainboard: original code: 18.0 - 18.5 hours per CPU generic optimized code: 10.5 - 11.0 hours per CPU speed increase: ~ 40% Single Celeron-333, Asus P2B mainboard: original code: 14.0 - 14.5 hours generic optimized code: 10.5 - 11.0 hours speed increase: ~ 25% Single Pentium-II-450, Asus P2B-LS mainboard: original code: 9.0 - 9.5 hours generic optimized code: 7.5 - 8.0 hours speed increase: ~ 16% As you can see, the Dual Celeron benefits the most, because it has a very weak RAM bandwidth, but fast (though small) L2 cache. AMD K7-500, MSI MS6167 mainboard: original code: 10.0 - 10.5 hours Athlon optimized code: 4.0 - 4.5 hours speed increase: ~ 60% These numbers speak for themselves. 3DNow rules. :) I'm still waiting for a response from the Seti@home folks. They have all the code, but they seem to ignore it. Regards Oliver -- Oliver Fromme, Leibnizstr. 18/61, 38678 Clausthal, Germany (Info: finger userinfo:olli@dorifer.heim3.tu-clausthal.de) "In jedem Stück Kohle wartet ein Diamant auf seine Geburt" (Terry Pratchett) To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message