]> git.sesse.net Git - x264/commit
projects / x264 / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 16d0372) | patch
x86: Don't use explicitly aligned versions of SAD on AVX CPUs
authorHenrik Gramner <henrik@gramner.com>
Fri, 3 May 2013 21:06:10 +0000 (23:06 +0200)
committerFiona Glaser <fiona@x264.com>
Fri, 17 May 2013 16:26:49 +0000 (09:26 -0700)
commit33c352673900bd1b362bb2fe0284e999fccd633d
tree529d0180619258552c1eb7d937ec6403d577404etree | snapshot
parent16d037211f1dd032288e25ab74d93a569fd93d6ccommit | diff
x86: Don't use explicitly aligned versions of SAD on AVX CPUs

On modern CPUs movdqu isn't slower than movdqa when used on aligned data and using the same code in both cases saves cache.

This was already done for the high bit-depth AVX2 implementation but the aligned version still exists as dead code so remove that.
common/pixel.c diff | blob | history
common/x86/pixel.h diff | blob | history
common/x86/sad16-a.asm diff | blob | history
Unnamed repository; edit this file 'description' to name the repository.