diff options
| author | Paul B Mahol <onemda@gmail.com> | 2015-10-07 10:12:26 +0200 |
|---|---|---|
| committer | Paul B Mahol <onemda@gmail.com> | 2015-10-07 22:50:15 +0200 |
| commit | 0948ba320496d02ad185487c18b249610de1a184 (patch) | |
| tree | 11694f840cbe4272e79c8895e46ad10f2b50d295 /libavfilter/x86/vf_blend.asm | |
| parent | 8a9fa46e87daf09fca14c903e6ee0c13fd453998 (diff) | |
| download | ffmpeg-streaming-0948ba320496d02ad185487c18b249610de1a184.zip ffmpeg-streaming-0948ba320496d02ad185487c18b249610de1a184.tar.gz | |
avfilter/x86/vf_blend.asm: add hardmix and phoenix sse2 SIMD
Signed-off-by: Paul B Mahol <onemda@gmail.com>
Diffstat (limited to 'libavfilter/x86/vf_blend.asm')
| -rw-r--r-- | libavfilter/x86/vf_blend.asm | 64 |
1 files changed, 64 insertions, 0 deletions
diff --git a/libavfilter/x86/vf_blend.asm b/libavfilter/x86/vf_blend.asm index 167e72b..54b5430 100644 --- a/libavfilter/x86/vf_blend.asm +++ b/libavfilter/x86/vf_blend.asm @@ -27,6 +27,8 @@ SECTION_RODATA pw_128: times 8 dw 128 pw_255: times 8 dw 255 +pb_128: times 16 db 128 +pb_255: times 16 db 255 SECTION .text @@ -273,6 +275,37 @@ cglobal blend_darken, 9, 10, 2, 0, top, top_linesize, bottom, bottom_linesize, d jg .nextrow REP_RET +cglobal blend_hardmix, 9, 10, 4, 0, top, top_linesize, bottom, bottom_linesize, dst, dst_linesize, width, start, end + add topq, widthq + add bottomq, widthq + add dstq, widthq + sub endq, startq + mova m2, [pb_255] + mova m3, [pb_128] + neg widthq +.nextrow: + mov r10q, widthq + %define x r10q + + .loop: + movu m0, [topq + x] + movu m1, [bottomq + x] + pxor m1, m2 + pxor m0, m3 + pxor m1, m3 + pcmpgtb m1, m0 + pxor m1, m2 + mova [dstq + x], m1 + add r10q, mmsize + jl .loop + + add topq, top_linesizeq + add bottomq, bottom_linesizeq + add dstq, dst_linesizeq + sub endd, 1 + jg .nextrow +REP_RET + cglobal blend_lighten, 9, 10, 2, 0, top, top_linesize, bottom, bottom_linesize, dst, dst_linesize, width, start, end add topq, widthq add bottomq, widthq @@ -298,6 +331,37 @@ cglobal blend_lighten, 9, 10, 2, 0, top, top_linesize, bottom, bottom_linesize, jg .nextrow REP_RET +cglobal blend_phoenix, 9, 10, 4, 0, top, top_linesize, bottom, bottom_linesize, dst, dst_linesize, width, start, end + add topq, widthq + add bottomq, widthq + add dstq, widthq + sub endq, startq + mova m3, [pb_255] + neg widthq +.nextrow: + mov r10q, widthq + %define x r10q + + .loop: + movu m0, [topq + x] + movu m1, [bottomq + x] + mova m2, m0 + pminub m0, m1 + pmaxub m1, m2 + mova m2, m3 + psubusb m2, m1 + paddusb m2, m0 + mova [dstq + x], m2 + add r10q, mmsize + jl .loop + + add topq, top_linesizeq + add bottomq, bottom_linesizeq + add dstq, dst_linesizeq + sub endd, 1 + jg .nextrow +REP_RET + INIT_XMM ssse3 cglobal blend_difference, 9, 10, 3, 0, top, top_linesize, bottom, bottom_linesize, dst, dst_linesize, width, start, end add topq, widthq |
