---
libavcodec/x86/vp8dsp.asm | 141 +++++++++++++++++----------------------------
1 files changed, 53 insertions(+), 88 deletions(-)
diff --git a/libavcodec/x86/vp8dsp.asm b/libavcodec/x86/vp8dsp.asm
index e0041ec..9d1a9fd 100644
--- a/libavcodec/x86/vp8dsp.asm
+++ b/libavcodec/x86/vp8dsp.asm
@@ -678,9 +678,28 @@ FILTER_V 4
INIT_XMM sse2
FILTER_V 8
-%macro FILTER_BILINEAR 1
-cglobal put_vp8_bilinear%1_v, 7, 7, 7, dst, dststride, src, srcstride, height,
picreg, my
+%macro FILTER_BILINEAR 3
+cglobal put_vp8_bilinear%1_v, 7, 7, %2, dst, dststride, src, srcstride,
height, picreg, my
shl myd, 4
+%if cpuflag(ssse3)
+%ifdef PIC
+ lea picregq, [bilinear_filter_vb_m]
+%endif
+ pxor m4, m4
+ mova m3, [bilinear_filter_vb+myq-16]
+.nextrow
+ movh m0, [srcq+srcstrideq*0]
+ movh m1, [srcq+srcstrideq*1]
+ movh m2, [srcq+srcstrideq*2]
+ punpcklbw m0, m1
+ punpcklbw m1, m2
+ pmaddubsw m0, m3
+ pmaddubsw m1, m3
+ psraw m0, 2
+ psraw m1, 2
+ pavgw m0, m4
+ pavgw m1, m4
+%else ; cpuflag(ssse3)
%ifdef PIC
lea picregq, [bilinear_filter_vw_m]
%endif
@@ -706,13 +725,14 @@ cglobal put_vp8_bilinear%1_v, 7, 7, 7, dst, dststride,
src, srcstride, height, p
psraw m2, 2
pavgw m0, m6
pavgw m2, m6
+%endif ; cpuflag(ssse3)
%if mmsize == 8
packuswb m0, m0
- packuswb m2, m2
+ packuswb %3, %3
movh [dstq+dststrideq*0], m0
- movh [dstq+dststrideq*1], m2
+ movh [dstq+dststrideq*1], %3
%else
- packuswb m0, m2
+ packuswb m0, %3
movh [dstq+dststrideq*0], m0
movhps [dstq+dststrideq*1], m0
%endif
@@ -723,8 +743,27 @@ cglobal put_vp8_bilinear%1_v, 7, 7, 7, dst, dststride,
src, srcstride, height, p
jg .nextrow
REP_RET
-cglobal put_vp8_bilinear%1_h, 6, 6 + npicregs, 7, dst, dststride, src,
srcstride, height, mx, picreg
+cglobal put_vp8_bilinear%1_h, 6, 6 + npicregs, %2, dst, dststride, src,
srcstride, height, mx, picreg
shl mxd, 4
+%if cpuflag(ssse3)
+%ifdef PIC
+ lea picregq, [bilinear_filter_vb_m]
+%endif
+ pxor m4, m4
+ mova m2, [filter_h2_shuf]
+ mova m3, [bilinear_filter_vb+mxq-16]
+.nextrow
+ movu m0, [srcq+srcstrideq*0]
+ movu m1, [srcq+srcstrideq*1]
+ pshufb m0, m2
+ pshufb m1, m2
+ pmaddubsw m0, m3
+ pmaddubsw m1, m3
+ psraw m0, 2
+ psraw m1, 2
+ pavgw m0, m4
+ pavgw m1, m4
+%else ; cpuflag(ssse3)
%ifdef PIC
lea picregq, [bilinear_filter_vw_m]
%endif
@@ -751,13 +790,14 @@ cglobal put_vp8_bilinear%1_h, 6, 6 + npicregs, 7, dst,
dststride, src, srcstride
psraw m2, 2
pavgw m0, m6
pavgw m2, m6
+%endif ; cpuflag(ssse3)
%if mmsize == 8
packuswb m0, m0
- packuswb m2, m2
+ packuswb %3, %3
movh [dstq+dststrideq*0], m0
- movh [dstq+dststrideq*1], m2
+ movh [dstq+dststrideq*1], %3
%else
- packuswb m0, m2
+ packuswb m0, %3
movh [dstq+dststrideq*0], m0
movhps [dstq+dststrideq*1], m0
%endif
@@ -770,88 +810,13 @@ cglobal put_vp8_bilinear%1_h, 6, 6 + npicregs, 7, dst,
dststride, src, srcstride
%endmacro
INIT_MMX mmxext
-FILTER_BILINEAR 4
+FILTER_BILINEAR 4, 7, m2
INIT_XMM sse2
-FILTER_BILINEAR 8
-
-%macro FILTER_BILINEAR_SSSE3 1
-cglobal put_vp8_bilinear%1_v, 7, 7, 5, dst, dststride, src, srcstride, height,
picreg, my
- shl myd, 4
-%ifdef PIC
- lea picregq, [bilinear_filter_vb_m]
-%endif
- pxor m4, m4
- mova m3, [bilinear_filter_vb+myq-16]
-.nextrow
- movh m0, [srcq+srcstrideq*0]
- movh m1, [srcq+srcstrideq*1]
- movh m2, [srcq+srcstrideq*2]
- punpcklbw m0, m1
- punpcklbw m1, m2
- pmaddubsw m0, m3
- pmaddubsw m1, m3
- psraw m0, 2
- psraw m1, 2
- pavgw m0, m4
- pavgw m1, m4
-%if mmsize==8
- packuswb m0, m0
- packuswb m1, m1
- movh [dstq+dststrideq*0], m0
- movh [dstq+dststrideq*1], m1
-%else
- packuswb m0, m1
- movh [dstq+dststrideq*0], m0
- movhps [dstq+dststrideq*1], m0
-%endif
-
- lea dstq, [dstq+dststrideq*2]
- lea srcq, [srcq+srcstrideq*2]
- sub heightd, 2
- jg .nextrow
- REP_RET
-
-cglobal put_vp8_bilinear%1_h, 6, 6 + npicregs, 5, dst, dststride, src,
srcstride, height, mx, picreg
- shl mxd, 4
-%ifdef PIC
- lea picregq, [bilinear_filter_vb_m]
-%endif
- pxor m4, m4
- mova m2, [filter_h2_shuf]
- mova m3, [bilinear_filter_vb+mxq-16]
-.nextrow
- movu m0, [srcq+srcstrideq*0]
- movu m1, [srcq+srcstrideq*1]
- pshufb m0, m2
- pshufb m1, m2
- pmaddubsw m0, m3
- pmaddubsw m1, m3
- psraw m0, 2
- psraw m1, 2
- pavgw m0, m4
- pavgw m1, m4
-%if mmsize==8
- packuswb m0, m0
- packuswb m1, m1
- movh [dstq+dststrideq*0], m0
- movh [dstq+dststrideq*1], m1
-%else
- packuswb m0, m1
- movh [dstq+dststrideq*0], m0
- movhps [dstq+dststrideq*1], m0
-%endif
-
- lea dstq, [dstq+dststrideq*2]
- lea srcq, [srcq+srcstrideq*2]
- sub heightd, 2
- jg .nextrow
- REP_RET
-%endmacro
-
+FILTER_BILINEAR 8, 7, m2
INIT_MMX ssse3
-FILTER_BILINEAR_SSSE3 4
+FILTER_BILINEAR 4, 7, m2
INIT_XMM ssse3
-FILTER_BILINEAR_SSSE3 8
+FILTER_BILINEAR 8, 5, m1
INIT_MMX mmx
cglobal put_vp8_pixels8, 5, 5, 0, dst, dststride, src, srcstride, height
--
1.7.2.5
_______________________________________________
libav-devel mailing list
libav-devel@libav.org
https://lists.libav.org/mailman/listinfo/libav-devel
