diff options
| author | James Almer <jamrial@gmail.com> | 2014-05-26 03:49:35 -0300 |
|---|---|---|
| committer | Michael Niedermayer <michaelni@gmx.at> | 2014-05-27 05:55:11 +0200 |
| commit | e64e079ece7d037686c4c0f97eac9c62af6300b1 (patch) | |
| tree | 96227a1b784bbf67a6d3b6e9d4b9640b98ae6db6 /libavcodec/x86/dsputilenc.asm | |
| parent | a0c5cd3475fd93930604e4ec5ac1336f5732c04b (diff) | |
| download | ffmpeg-e64e079ece7d037686c4c0f97eac9c62af6300b1.tar.gz | |
x86/dsputilenc: implement SSE2 version of diff_pixels
Signed-off-by: James Almer <jamrial@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Diffstat (limited to 'libavcodec/x86/dsputilenc.asm')
| -rw-r--r-- | libavcodec/x86/dsputilenc.asm | 25 |
1 files changed, 25 insertions, 0 deletions
diff --git a/libavcodec/x86/dsputilenc.asm b/libavcodec/x86/dsputilenc.asm index 7426c01dbb..ba8a1773c2 100644 --- a/libavcodec/x86/dsputilenc.asm +++ b/libavcodec/x86/dsputilenc.asm @@ -419,6 +419,31 @@ cglobal diff_pixels, 4,5 jne .loop REP_RET +INIT_XMM sse2 +cglobal diff_pixels, 4, 5, 5 + movsxdifnidn r3, r3d + pxor m4, m4 + add r0, 128 + mov r4, -128 +.loop: + movh m0, [r1] + movh m2, [r2] + movh m1, [r1+r3] + movh m3, [r2+r3] + punpcklbw m0, m4 + punpcklbw m1, m4 + punpcklbw m2, m4 + punpcklbw m3, m4 + psubw m0, m2 + psubw m1, m3 + mova [r0+r4+0 ], m0 + mova [r0+r4+16], m1 + lea r1, [r1+r3*2] + lea r2, [r2+r3*2] + add r4, 32 + jne .loop + RET + INIT_MMX mmx ; int ff_pix_sum16_mmx(uint8_t *pix, int line_size) cglobal pix_sum16, 2, 3 |