]> granicus.if.org Git - libx264/commit
projects / libx264 / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 5ebbcd8) | patch
x86inc: AVX symmetry optimization
authorLoren Merritt <pengvado@akuvian.org>
Sun, 23 Oct 2011 23:15:11 +0000 (23:15 +0000)
committerFiona Glaser <fiona@x264.com>
Thu, 1 Dec 2011 20:27:40 +0000 (12:27 -0800)
commitf3a7517cb9b06a580623cbea0f140be534b99877
tree7bfe7fcb47a3577c3a105b5c2faae41c9698d64ctree | snapshot
parent5ebbcd8748ae8d8b184db5a8f9b46a9ad865f0aecommit | diff
x86inc: AVX symmetry optimization
3-arg AVX ops with a memory arg can only have it in src2,
whereas SSE emulation of 3-arg prefers to have it in src1 (i.e. the move).
So, if the op is symmetric and the wrong one is memory, swap them.
Eliminates redundant moves in some cases when using 3-operand without AVX with memory arguments.
Also fix movss and movsd in some cases, and flag shufps correctly as float.
common/x86/x86inc.asm diff | blob | history
Unnamed repository; edit this file 'description' to name the repository.