Fix matmul performance on gcc 4.9
authorAndrew Waterman <waterman@cs.berkeley.edu>
Tue, 27 Jan 2015 08:33:47 +0000 (00:33 -0800)
committerAndrew Waterman <waterman@cs.berkeley.edu>
Tue, 27 Jan 2015 08:33:47 +0000 (00:33 -0800)
commit94537321a456c65c4a14e005de03fced33e0a43b
tree4f9d4f3281339a88f3d10ba7216bd5fa9c2b0222
parent160bdaa323bc8f8e651f9f546822336cf17d92f5
Fix matmul performance on gcc 4.9

It's just loop interchange in the register blocking loop.
benchmarks/mm/gen.scala
benchmarks/mm/rb.h