home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.sgi
- Path: sparky!uunet!haven.umd.edu!darwin.sura.net!sgiblab!sgigate!odin!fido!zola!zuni!anchor!olson
- From: olson@anchor.esd.sgi.com (Dave Olson)
- Subject: Re: duplicating crimson benchmarks?
- Message-ID: <ptavrag@zuni.esd.sgi.com>
- Keywords: benchmarks
- Sender: news@zuni.esd.sgi.com (Net News)
- Organization: Silicon Graphics, Inc. Mountain View, CA
- References: <9846@fs3.cam.nist.gov>
- Date: Wed, 16 Sep 92 01:44:22 GMT
- Lines: 26
-
- In <9846@fs3.cam.nist.gov> gpfrn@cam.nist.gov (Glenn_P._Forney_x2313) writes:
-
- | it is my understanding that the floating point
- | performance of the crimson (or equivalently the
- | r4000 indigo) is about 15mflops where the mflop's
- | are measured using dongarra's double precision
- | linpack benchmark . is this correct?
- |
- | in any case,
- | i obtained this benchmark from netlib, compiled it using
- | the -O2 option. I also used the assembly coded blas
- | via the -lblas compiler option. i am getting
- | about 11 mflops instead of 15 .
- |
- | i would like some more definitive
- | info on how the 15 mflop number was arrived at, like
- | the compiler options or a make file.
- | we are running system 4.0.5 and fortran 3.4.1.
-
- That is probably the problem. Use the 4.1 (MIPS 3.10)
- compilers, and use the -mips2 option. I don't think
- we did much else differently.
- --
- Let no one tell me that silence gives consent, | Dave Olson
- because whoever is silent dissents. | Silicon Graphics, Inc.
- Maria Isabel Barreno | olson@sgi.com
-