Re: sgemm questions

[R Clint Whaley <rwhaley@cs.utk.edu>]

Camm,

>Hello again!  Just looked at this stuff again today, and did a rather
>simple change which makes the kernel work for the Athlon, with a
>slightly higher percentage of peak than the Intel, it appears.  

Peter Soendergaard has been working on 3DNow! SGEMM; coincidentally,
he made a rather simple change to his 3DNow! code to run SSE the other
day, and got about the same performance as your SSE :)

Hopefully, Peter will reply more fully regarding what instructions he
used, and why.  If he used Athlon-specific ones, having a K6x version
would be nice as well, so long as it is trivial.  Last I knew, Peter was
getting about 2.4 Gflops on our 1Ghz Athlon.  Anyway, I obviously don't have
all the details you need . . .

Cheers,
Clint