On Tue, 9 Sep 2003, Gregor Stößer wrote:

On Tue, Sep 09, 2003 at 09:00:32AM -0600, James Eppich wrote:

...

We bought a dual opteron and run SuSE 8.2 beta 9. We have been trying to run software like Gaussian and scientific number crunching software like

I think Gaussian03 should have a makefile for the Opteron using the Portland Group compiler (I don't know for sure, I don't like Gaussian because it's so slow and nasty..)

The Portland 64bit compiler will not run on SuSE 8.2 beta. The previous release (5.0.1) simply crashed. Portland's solution was 5.0.2 which tests which Linux you're running and tells you to go away if it's not SLES 8. Which is why I've now ordered SLES8... Bjørn -- Bjørn Tore Sund Phone: (+47) 555-84894 Stupidity is like a System administrator Fax: (+47) 555-89672 fractal; universal and Math. Department Mobile: (+47) 918 68075 infinitely repetitive. University of Bergen VIP: 81724 teknisk@mi.uib.no Email: bjornts@mi.uib.no http://www.mi.uib.no/

G> do you have any ideas how good the performance of the Portland compiler G> compared with Intel's compiler ifc (IA32) is? Please submit your results G> here, I'm very interested. G> Cheers try this http://www.digit-life.com/articles2/insidespeccpu/insidespeccpu2000-opteron2... also there is part 1, but only on russian: http://www.ixbt.com/cpu/insidespeccpu2000-opteron1.shtml __________ www.newmail.ru -- Новая Почта: все по-новому.

On Wed, Sep 10, 2003 at 01:20:42PM +0200, Bjorn Tore Sund wrote:

We haven't used the Intel compiler for anything interesting here. I know from discussions on the Beowulf mailing lists that it's the compiler of choice for Intel 32bit architectures both for precision and speed, but no personal experience. We're very close to being an Athlon-only system.

Same for us, we're nearly solely buying AMD CPU's, but we use the Intel Compilers because they are even for the Athlon noticable faster than PGI and they are free of charge for academic use (there occur only problems when you want to do calculations over nfs, don't know why). Even AMD uses for the SPEC benchmarks Intel Compilers and not PGI.

I obviously haven't tested 64bit PGI yet, I need to wait until I have a version of Linux it'll run on. We have been using PGI for the last three years now. People generally seem to prefer the GNU C compilers but the PGI Fortran compilers if looking only at speed, though we see better precision with the GNU compilers than with the Portland compilers. Whether this will stay true with the 64bit compilers remain to be seen.

Bjørn --

Gregor -- -*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*- Gregor Stößer email: gregor.stoesser@chemie.uni-karlsruhe.de Institut für Anorganische Chemie Universität Karlsruhe Tel: 0721/608 2988 Fax: 0721/608 4854 *-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-

Well like I mentioned, I'm not a programer or linux person, or even the primary person to use the computer. I'm more the one that told my boss we need to buy a faster computer than a 500 MHz to do number crunching... But, I feel I am the one who need to make sure it works, so please be patient with my ignorance. 1) I honestly dont know what is not working about. The 2 people I have asked just tell me they have not been able to get it to work yet. - We are at the Univ. of Utah and have a site license to Gaussian 03 and gaussview. so if anyone could send me a tar ball, I would be sure to put you in contact with the computer licensing people here at the chem department. (this would be the easist way to help me since, they wont give me the source code) 2) supposedly gauss view is on the machine and running. I have loaded the program, but when I try to load .com or .log files generated (I'm guessing by g98 on another machine) it gives an error message. 3) we don't have the portland compilers, but the High Performance Computer Center here just bought 1000 opterons. I'm guessing they will buy the software eventually. But they don't want to support us until they have their cluster running... I have gotten LAPACK compiled (I think) and gOpenMol. they run... if that is at all useful to know. any suggestions on where to go from here, or what information to relay to the people I have asked to compile this software. James At 05:19 PM 9/9/2003 +0200, you wrote:

James Eppich writes:

...

We bought a dual opteron and run SuSE 8.2 beta 9. We have been trying to run software like Gaussian and scientific number crunching software like that. I'm not really a linux person but I have been working with other people on trying to get this software compiled. The other people have since left me hanging, saying they either dont have time or are just having trouble getting it to work. Has anyone had success on getting programs like these to work in x86-64 mode? (not IA-32/64 compiled) thanks,

What problems exactly are you facing? We've ported our whole distribution and do know a bit about those problems. So, if you have specific questions, ask...

Andreas -- Andreas Jaeger, aj@suse.de, http://www.suse.de/~aj SuSE Linux AG, Deutschherrnstr. 15-19, 90429 Nürnberg, Germany GPG fingerprint = 93A3 365E CE47 B889 DF7F FED1 389A 563C C272 A126