From chemistry-request@www.ccl.net Wed Nov 25 04:34:55 1998 Received: from romeo.ic.ac.uk (romeo.ic.ac.uk [155.198.5.9]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with SMTP id EAA10934 Wed, 25 Nov 1998 04:34:54 -0500 (EST) Received: from oban.cc.ic.ac.uk [155.198.5.28] by romeo.ic.ac.uk with smtp (Exim 1.62 #1) id 0zibLb-000615-00; Wed, 25 Nov 1998 09:34:55 +0000 Received: from sg1.cc.ic.ac.uk ([155.198.63.8] helo=cscmgb.cc.ic.ac.uk) by oban.cc.ic.ac.uk with esmtp (Exim 1.90 #1) for CHEMISTRY@www.ccl.net id 0zibLN-00071l-00; Wed, 25 Nov 1998 09:34:41 +0000 Received: from [155.198.224.86] by cscmgb.cc.ic.ac.uk (980427.SGI.8.8.8/4.0) id JAA03719; Wed, 25 Nov 1998 09:34:40 GMT Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" ; format="flowed" X-Sender: rzepa@155.198.63.8 Message-Id: In-Reply-To: Date: Wed, 25 Nov 1998 09:35:35 +0000 To: CHEMISTRY@www.ccl.net From: "Rzepa, Henry" Subject: Re: CCL:M:MOPAC97 >Hi again, > > Can anybody tell me where I can get the source code for MOPAC97 or >MOPAC95? QCPE only goes upto MOPAC93. > > I believe that MOPAC97 will only be distributed directly by Fujitsu, and not by QCPE. You should contact a Fujitsu office or reseller. Dr Henry Rzepa, Dept. Chemistry, Imperial College, LONDON SW7 2AY; mailto:rzepa@ic.ac.uk; Tel (44) 171 594 5774; Fax: (44) 171 594 5804. URL: http://www.ch.ic.ac.uk/rzepa/ If my digital email signature is invalid, download a new root at http://www.belsign.be/en/services/receive/install-ca.html From chemistry-request@www.ccl.net Wed Nov 25 07:20:28 1998 Received: from chemistry.leeds.ac.uk (chmsun1.leeds.ac.uk [129.11.12.1]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id HAA11271 Wed, 25 Nov 1998 07:20:23 -0500 (EST) Received: from [129.11.12.138] (chmmac37.leeds.ac.uk [129.11.12.138]) by chemistry.leeds.ac.uk (8.8.8/8.8.8) with ESMTP id MAA28400 for ; Wed, 25 Nov 1998 12:20:00 GMT Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Sender: darrena@chem.leeds.ac.uk Message-Id: Date: Wed, 25 Nov 1998 12:19:43 +0000 To: chemistry@www.ccl.net From: Darren Andrews Subject: 3D Graphics Accelerators as computational tools Has anyone thought of using the awesome floating point performance of 3D graphics accelerators in computational chemistry? A decent 3D Voodoo II card is theoretically capable of about 9Gflops/sec, two of them together give 18Gflops/sec, and with an AGP interface can talk to the cpu at over 500M/sec. Is it possible to harness this power, and would it even be worthwhile trying? Darren Andrews. PostGraduate Student, School of Chemistry, University of Leeds, Leeds. LS2 9JT. England. Darrena@chem.leeds.ac.uk Tel: 0113 233 6594. Fax: 0113 233 6565. From chemistry-request@www.ccl.net Wed Nov 25 09:03:01 1998 Received: from ausmail1.austin.ibm.com (ausmail1.austin.ibm.com [192.35.232.12]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id JAA11989 Wed, 25 Nov 1998 09:03:00 -0500 (EST) Received: from netmail1.austin.ibm.com (netmail1.austin.ibm.com [9.53.250.96]) by ausmail1.austin.ibm.com (8.9.1/8.8.5) with ESMTP id IAA49906; Wed, 25 Nov 1998 08:03:31 -0600 Received: from mjodalfr.austin.ibm.com (mjodalfr.austin.ibm.com [9.53.67.212]) by netmail1.austin.ibm.com (8.8.5/8.8.5) with SMTP id IAA24622; Wed, 25 Nov 1998 08:02:46 -0600 Received: by mjodalfr.austin.ibm.com (AIX 4.1/UCB 5.64/4.03-client-2.6) for chemistry@www.ccl.net at austin.ibm.com; id AA30914; Wed, 25 Nov 1998 08:02:44 -0600 Date: Wed, 25 Nov 1998 08:02:42 -0600 (CST) From: "Kenneth W. Egan" To: Darren Andrews Cc: chemistry@www.ccl.net Subject: Re: CCL:3D Graphics Accelerators as computational tools In-Reply-To: Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII > Has anyone thought of using the awesome floating point performance of 3D > graphics accelerators in computational chemistry? A decent 3D Voodoo II > card is theoretically capable of about 9Gflops/sec, two of them together > give 18Gflops/sec, and with an AGP interface can talk to the cpu at over > 500M/sec. Is it possible to harness this power, and would it even be > worthwhile trying? This brings up several issues.... 1). Do the chips actual implement IEEE floating point, or is it proprietary...like 24.8 or 22.10 ( fixed ) 2.) Practical only in Geometry processors. However, to use each FP multiplier, as that, requires making sure none of the other pipeline units are operative, or at least those stages are not interactive.... World scaling, vertex transformation, lighting, clipping....etc. The matrices are usually hard coded, so initialization may require significant setup. 3.) End result----> must do setup, AGP write to adapter then AGP read....max theoritical bandwidth is thus 250 MFLOPS regardless of the number of FP multipliers in the chip. The IBM/motorola Power 630 is capable of 1 FP op/cycle, the newest chips can achieve 450 MFLOPS. I think the new IMAC has a 630...not sure. Not a bad line of reasoning though. Ken Egan IBM 3D VLSI design Austin