[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Altivec matmul kernel (attachment)



Hi folks,

Attached is my Altivec l1 matmul kernel.  I think I did the scases file 
correctly.  This is really only a preliminary version, but it may turn 
out that I can't do any better than this!  Using this kernel, I get 1620 
Mflops peak SGEMM on my 533 G4 vs. 670 Mflops for the scalar SGEMM.  
These results are for Mac OS X (I have linux but don't have the 
Altivec-enabled compiler so I didn't try it there).


-Nick

atlas_altivec.tar.gz



--
Nicholas Coult, Ph.D.,  web: http://melby.augsburg.edu/~coult
Assistant Professor, Department of Mathematics, Augsburg College
coult@augsburg.edu, phone:  (612) 330-1064 office: Science Hall 137B