Conference eps::benchmark

256.0. "MFLOPS on 4100 running DU" by AXPBIZ::GUDE () Fri May 30 1997 18:32

This question came to us from one of our partners. 

Kevin McManus of the University of Greenwich (UK) has been running his own 
benchmark for floating point performance (a f77 SAXPY program) comparing 
different machines (including Alpha/Sparc/Ultra). 

He claims to a get a maximal MFLOP on a 4100 of 231.4. Since it is a 466Mhz CPU
and can perform 2 fp's per clock, he expects to see close to 3 times that many
MFLOPs (close to 466*2). Is he correct in his assumptions ?

Anyway,  our partner would like some explanation for the 4100's 'poor' results
in McManus's test. 

McManus's code can be found at http://www.gre.ac.uk/~k.mcmanus/saxpy.f 
 
Thanks,

Dev

(This note has also been submitted to the SPEC conference)