x86/hevc_res_add: refactor ff_hevc_transform_add{16,32}_8
authorJames Almer <jamrial@gmail.com>
Wed, 20 Aug 2014 22:36:29 +0000 (19:36 -0300)
committerJames Almer <jamrial@gmail.com>
Thu, 21 Aug 2014 18:01:33 +0000 (15:01 -0300)
commit54ca4dd43bdc8658b7304d9309cdb096c8e8a394
tree087d3f24d4603982f1fc13c67162a89db1ebdc45
parent4a5cc34b46a8bf8d47ec907383be83b6153b9f69
x86/hevc_res_add: refactor ff_hevc_transform_add{16,32}_8

* Reduced xmm register count to 7 (As such they are now enabled for x86_32).
* Removed four movdqa (affects the sse2 version only).
* pxor is now used to clear m0 only once.

~5% faster.

Reviewed-by: Christophe Gisquet <christophe.gisquet@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>
libavcodec/x86/hevc_res_add.asm
libavcodec/x86/hevcdsp_init.c