Matrix Multiplication Optimization Results
The graph shows the performance of the class's square matrix multiplication kernels.
For comparison, it includes the speed of Sun's
Performance Library, and older version of ATLAS, and older version of
PHiPAC, and Strazdins's
UltraSparc BLAS. The latter is the fastest UltraSparc-I BLAS I
know. Labelle, Jiang, and Yi's kernel is
competitive, and it clearly beats the vendor library. They also
handily beat last year's results.
Descriptions of the code for each group:
-
Bonachea,
Chapman, and Putnam
-
Horner, Kramer, and
Kundu
-
Bindel, Lou, and Yihan
-
Eaton and Scott
-
Labelle, Jiang, and Yi
-
Chien, Deng, and Subramanian
Main CS267
page,
the this assignment,
and the
TA's CS267 page
E. Jason Riedy
ejr@cs.berkeley.edu
