diff options
| author | James Almer <jamrial@gmail.com> | 2014-08-04 01:18:46 -0300 |
|---|---|---|
| committer | Michael Niedermayer <michaelni@gmx.at> | 2014-08-04 14:47:15 +0200 |
| commit | b7863c972c937712eb0210177c47e4b51112290b (patch) | |
| tree | d0dc5472c35429452bf3dc7db45c90046886ab75 /libavcodec/x86/hevc_mc.asm | |
| parent | b1a44e6bf5e2c708d98d3eeeb2068c4b82987819 (diff) | |
| download | ffmpeg-b7863c972c937712eb0210177c47e4b51112290b.tar.gz | |
x86/hevc_mc: use fewer instructions in hevc_put_hevc_{uni, bi}_w[24]_{8, 10, 12}
Signed-off-by: James Almer <jamrial@gmail.com>
Reviewed-by: Christophe Gisquet <christophe.gisquet@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Diffstat (limited to 'libavcodec/x86/hevc_mc.asm')
| -rw-r--r-- | libavcodec/x86/hevc_mc.asm | 35 |
1 files changed, 34 insertions, 1 deletions
diff --git a/libavcodec/x86/hevc_mc.asm b/libavcodec/x86/hevc_mc.asm index a16b0ab532..c525078143 100644 --- a/libavcodec/x86/hevc_mc.asm +++ b/libavcodec/x86/hevc_mc.asm @@ -1157,9 +1157,16 @@ cglobal hevc_put_hevc_uni_w%1_%2, 6, 6, 7, dst, dststride, src, srcstride, heigh %define SHIFT denomd %endif lea SHIFT, [SHIFT+14-%2] ; shift = 14 - bitd + denom +%if %1 <= 4 + pxor m1, m1 +%endif movd m2, wxm ; WX movd m4, SHIFT ; shift +%if %1 <= 4 + punpcklwd m2, m1 +%else punpcklwd m2, m2 +%endif dec SHIFT movdqu m5, [one_per_32] movd m6, SHIFT @@ -1176,6 +1183,13 @@ cglobal hevc_put_hevc_uni_w%1_%2, 6, 6, 7, dst, dststride, src, srcstride, heigh %endif .loop SIMPLE_LOAD %1, 10, srcq, m0 +%if %1 <= 4 + punpcklwd m0, m1 + pmaddwd m0, m2 + paddd m0, m5 + psrad m0, m4 + paddd m0, m3 +%else pmulhw m6, m0, m2 pmullw m0, m2 punpckhwd m1, m0, m6 @@ -1186,6 +1200,7 @@ cglobal hevc_put_hevc_uni_w%1_%2, 6, 6, 7, dst, dststride, src, srcstride, heigh psrad m1, m4 paddd m0, m3 paddd m1, m3 +%endif packusdw m0, m1 %if %2 == 8 packuswb m0, m0 @@ -1201,13 +1216,21 @@ cglobal hevc_put_hevc_uni_w%1_%2, 6, 6, 7, dst, dststride, src, srcstride, heigh cglobal hevc_put_hevc_bi_w%1_%2, 6, 7, 10, dst, dststride, src, srcstride, src2, src2stride, height, denom, wx0, wx1, ox0, ox1 mov r6d, denomm +%if %1 <= 4 + pxor m1, m1 +%endif movd m2, wx0m ; WX0 lea r6d, [r6d+14-%2] ; shift = 14 - bitd + denom movd m3, wx1m ; WX1 movd m0, r6d ; shift +%if %1 <= 4 + punpcklwd m2, m1 + punpcklwd m3, m1 +%else punpcklwd m2, m2 - inc r6d punpcklwd m3, m3 +%endif + inc r6d movd m5, r6d ; shift+1 pshufd m2, m2, 0 mov r6d, ox0m @@ -1225,6 +1248,15 @@ cglobal hevc_put_hevc_bi_w%1_%2, 6, 7, 10, dst, dststride, src, srcstride, src2, .loop SIMPLE_LOAD %1, 10, srcq, m0 SIMPLE_LOAD %1, 10, src2q, m8 +%if %1 <= 4 + punpcklwd m0, m1 + punpcklwd m8, m1 + pmaddwd m0, m3 + pmaddwd m8, m2 + paddd m0, m4 + paddd m0, m8 + psrad m0, m5 +%else pmulhw m6, m0, m3 pmullw m0, m3 pmulhw m7, m8, m2 @@ -1239,6 +1271,7 @@ cglobal hevc_put_hevc_bi_w%1_%2, 6, 7, 10, dst, dststride, src, srcstride, src2, paddd m1, m4 psrad m0, m5 psrad m1, m5 +%endif packusdw m0, m1 %if %2 == 8 packuswb m0, m0 |