On 17 Feb, Marcelo E. Magallon wrote: > On 16 Feb 1998, Ben Pfaff wrote: > >> Can you point me to the source code for the benchmark? I can run it >> on my PII/233 for comparison if you want. > > and Alex Yukhimets: > >> Same here, only with PII/300. > > You can find the source code here: > > http://www.efis.ucr.ac.cr/~mmagallo/flops_p.c > > This is extracted from the flops program written by Al Burto > ([EMAIL PROTECTED]). This is just module 1, using 7 sums, 6 mults, 1 div. > I think I got this from http://www.hensa.ac.uk/parallel/ or netlib. > > Compiling with gcc -O2 gives a bit more than 39 on the PII. Using egcc > (1.0.1-0.3) with -mpentiumpro gives almost 40.9 > > Many thanks in advance. > > Marcelo
All right, if I was your professor, I'd ask for my money back. I have a dual PPro 200, and I get roughly 67, from gcc 2.7.2.3 and gcc 2.8.0. gcc 2.8 didn't help nearly as much as I was hoping it would. I also have access to a PII/266, and get roughly 81 from that. Both of these machines were running the distributed.net client in the background, in case that matters. For the person who was asking about SGI's, we have an SGI Origin here (one of the little, 4x180 Mhz processor ones), and get 72 from that, using just -O2 for optimizations. All of these numbers are per processor, as I've only tested one process at a time. I've spent $2800 on the dual PPro, $3500 if you count the peripherals I brought over from my old 486. I have no idea what kind of obscene amounts of money the SGI Origin cost. You may draw your own conclusions about cost effectiveness :-) -- Stephen Ryan Debian GNU/Linux Mathematics graduate student, Dartmouth College -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED] .