From h.rzepa@ic.ac.uk Sat Jan 6 06:46:05 1996 Received: from punch.ic.ac.uk for h.rzepa@ic.ac.uk by www.ccl.net (8.6.10/950822.1) id GAA26137; Sat, 6 Jan 1996 06:38:05 -0500 Received: from judy.ic.ac.uk by punch.ic.ac.uk with SMTP (PP); Sat, 6 Jan 1996 11:37:45 +0000 Received: from cscmgb.cc.ic.ac.uk (sg1.cc.ic.ac.uk [155.198.63.8]) by judy.ic.ac.uk (8.7.3/V1.4) with SMTP id LAA16861 for ; Sat, 6 Jan 1996 11:37:28 GMT Received: from [155.198.63.19] by cscmgb.cc.ic.ac.uk (940816.SGI.8.6.9/4.0) id LAA05237; Sat, 6 Jan 1996 11:37:24 GMT X-Sender: rzepa@155.198.63.8 Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: Sat, 6 Jan 1996 11:43:42 +0000 To: CHEMISTRY@www.ccl.net From: h.rzepa@ic.ac.uk (Rzepa,Henry) Subject: Re: CCL:POWERMAC clones for compchem >To follow-up on the recent comments of John Nash (1-5-96): > >>It would be foolish, I think, to ignore the Power Macintosh >>as such a system based on long-held OS biases as to what a "real PC" is. I have used LS Fortran intermittently for some time, and it is certainly a good compiler. However, I would suggest that the PowerMac needs to get its pre-emptive multitasking/ protected memory OS (Copland and Gershwin) before it can be taken totally seriously. Also, the internal bus structure has been substantially improved with the PCI (which not all Macs have), although I/O is still not up to the Unix boys standards. On a different theme, I have been experimenting with QuickDraw 3D, which addresses the very serious limitations on 3D graphics that Macs (and PCs) used to have. The QD3D accelerator card ($500?) certainly starts to give e.g. SGI entry level machines a run for their money, and presumably we can expect the card to be built into the mother board of higher end Macs soon. All we need now is software! From martinh@paston.co.uk Sat Jan 6 11:01:10 1996 Received: from red.paston.co.uk for martinh@paston.co.uk by www.ccl.net (8.6.10/950822.1) id LAA26934; Sat, 6 Jan 1996 11:00:44 -0500 Received: from martinh.paston.co.uk ([194.129.188.112]) by red.paston.co.uk with SMTP id <22727-1>; Sat, 6 Jan 1996 15:59:43 -0000 X-Sender: martinh@mail.paston.co.uk X-Mailer: Windows Eudora Version 1.4.3 Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" To: toni@athe.wustl.edu (Toni Kazic) From: martinh@paston.co.uk (Martin Hargreaves) Subject: Re: CCL:PowerMac as computational workstation Cc: CHEMISTRY@www.ccl.net Message-Id: <96Jan6.155943-0000gmt.22727-1+25@red.paston.co.uk> Date: Sat, 6 Jan 1996 15:59:39 -0000 toni@athe.wustl.edu (Toni Kazic) writes: >Perhaps I missed this in the current thread, but do the Unix FORTRAN >compilers run on PCs running Linux, and does this combination provide the >performance expected from the raw numbers? f2c (a FORTRAN to C converter) and gcc do a pretty seamless job of compiling a lot of FORTRAN code. I've built a few packages from source on Linux (PSI88 and some MD code), I've also used MOPAC and GAMESS binaries on Linux which seem to work fine. I haven't looked at the numbers but from memory MOPAC on Linux was _much_ faster than MOPAC for DOS (I got both from the OSC FTP site). I'd highly recommend the OS/2 version of MOPAC ported at Cracow. It appeared to run slightly faster than the Linux version and had a nicer interface/better documentation. I never had the disk space to do anything serious with GAMESS... :-( Regards, Martin. ################################################################### # Martin Hargreaves martinh@paston.co.uk # # Unix systems/security consultant # # Computational Chemist http://www.chem.surrey.ac.uk/~ch11mh # ################################################################### From ross@cgl.ucsf.EDU Sat Jan 6 17:31:14 1996 Received: from socrates.ucsf.EDU for ross@cgl.ucsf.EDU by www.ccl.net (8.6.10/950822.1) id RAA01167; Sat, 6 Jan 1996 17:22:02 -0500 From: Received: (from ross@localhost) by socrates.ucsf.EDU (8.7.3/GSC4.24) id OAA00693 for chemistry@www.ccl.net; Sat, 6 Jan 1996 14:22:02 -0800 (PST) Date: Sat, 6 Jan 1996 14:22:02 -0800 (PST) Message-Id: <199601062222.OAA00693@socrates.ucsf.EDU> To: chemistry@www.ccl.net Subject: linux/g77/etc toni@athe.wustl.edu (Toni Kazic) writes: >>Perhaps I missed this in the current thread, but do the Unix FORTRAN >>compilers run on PCs running Linux, and does this combination provide the >>performance expected from the raw numbers? >f2c (a FORTRAN to C converter) and gcc do a pretty seamless job of compiling >a lot of FORTRAN code. The new g77 seems a little faster, at least on molecular mechanical codes - see the benchmarks under http://www.amber.ucsf.edu/amber/. I haven't evaluated how they compare to 'raw numbers', but this comparison could be done w/ those benchmarks if one had those numbers in hand (SPECfp #s?). Bill Ross