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

Re: sgemm questions

Hi Clint!

OK, at


I have updated gemm kernels and new cleanup kernels for PIII and Athlon.

The performance Peter has been getting for the Athlon sounds
wonderful!  Congratulations!  I doubt if these Athlon kernels will
measure up, but I include them anyway.  I only get about 2.0*Mhz for
the Athlon, which is better that the PIII mostly because of the pfacc
instruction, and the convenience of doing just one 64bit write for 2
elements from one register when using a 2x2xkb strategy.

The cleanups can probably be improved.  Also could not check the
variable m and n, but should work.  KB still compile time, but not
required to be a multiple of 4.

Take care,

Camm Maguire			     			camm@enhanced.com
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah