Skip to content

Commit

Permalink
crypto: cast5-avx - tune assembler code for more performance
Browse files Browse the repository at this point in the history
Patch replaces 'movb' instructions with 'movzbl' to break false register
dependencies, interleaves instructions better for out-of-order scheduling
and merges constant 16-bit rotation with round-key variable rotation.

tcrypt ECB results (128bit key):

Intel Core i5-2450M:

size    old-vs-new      new-vs-generic  old-vs-generic
        enc     dec     enc     dec     enc     dec
256     1.18x   1.18x   2.45x   2.47x   2.08x   2.10x
1k      1.20x   1.20x   2.73x   2.73x   2.28x   2.28x
8k      1.20x   1.19x   2.73x   2.73x   2.28x   2.29x

[v2]
 - Do instruction interleaving another way to avoid adding new FPU<=>CPU
   register moves as these cause performance drop on Bulldozer.
 - Improvements to round-key variable rotation handling.
 - Further interleaving improvements for better out-of-order scheduling.

Cc: Johannes Goetzfried <Johannes.Goetzfried@informatik.stud.uni-erlangen.de>
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
  • Loading branch information
Jussi Kivilinna authored and Herbert Xu committed Sep 6, 2012
1 parent f94a73f commit ddaea78
Showing 1 changed file with 160 additions and 106 deletions.
Loading

0 comments on commit ddaea78

Please sign in to comment.