x86/me_cmp: port mmxext and sse2 sad functions to yasm
authorJames Almer <jamrial@gmail.com>
Wed, 17 Sep 2014 00:41:47 +0000 (21:41 -0300)
committerMichael Niedermayer <michaelni@gmx.at>
Wed, 17 Sep 2014 09:12:50 +0000 (11:12 +0200)
commit0456d169c469a79e305813d14c873fe698c8c572
tree42982e394a09a6ebb90bf1aa57a2554476afe5fe
parentf1a8943ddd15223483cd342a322d02d97e2116c2
x86/me_cmp: port mmxext and sse2 sad functions to yasm

Also add a missing c->pix_abs[0][0] initialization, and sse2 versions of
sad16_x2, sad16_y2 and sad16_xy2 (%15 to %20 faster than mmxext).
Since the _xy2 versions are not bitexact, they are accordingly marked as
approximate.

Signed-off-by: James Almer <jamrial@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
libavcodec/x86/me_cmp.asm
libavcodec/x86/me_cmp_init.c