Commit 74f8d9aa authored by Timothy Gu's avatar Timothy Gu

x86/vf_blend: Add SSE2 optimization for screen

10x faster than C.
Reviewed-by: 's avatarPaul B Mahol <onemda@gmail.com>
parent c8b1612a
......@@ -111,6 +111,13 @@ BLEND_END
psrlw %1, 8 ; 00xx00xx a * b / 255
%endmacro
%macro SCREEN 4 ; a, b, pw_1, pw_255
pxor %1, %4 ; 00xx00xx 255 - a
pxor %2, %4
MULTIPLY %1, %2, %3
pxor %1, %4 ; 00xx00xx 255 - x / 255
%endmacro
BLEND_INIT multiply, 4
pxor m2, m2
mova m3, [pw_1]
......@@ -134,6 +141,28 @@ BLEND_INIT multiply, 4
jl .loop
BLEND_END
BLEND_INIT screen, 5
pxor m2, m2
mova m3, [pw_1]
mova m4, [pw_255]
.nextrow:
mov xq, widthq
.loop:
movh m0, [topq + xq] ; 0000xxxx
movh m1, [bottomq + xq]
punpcklbw m0, m2 ; 00xx00xx
punpcklbw m1, m2
SCREEN m0, m1, m3, m4
packuswb m0, m0 ; 0000xxxx
movh [dstq + xq], m0
add xq, mmsize / 2
jl .loop
BLEND_END
BLEND_INIT average, 3
pxor m2, m2
.nextrow:
......
......@@ -37,6 +37,7 @@ BLEND_FUNC(and, sse2)
BLEND_FUNC(darken, sse2)
BLEND_FUNC(difference128, sse2)
BLEND_FUNC(multiply, sse2)
BLEND_FUNC(screen, sse2)
BLEND_FUNC(hardmix, sse2)
BLEND_FUNC(lighten, sse2)
BLEND_FUNC(or, sse2)
......@@ -65,6 +66,7 @@ av_cold void ff_blend_init_x86(FilterParams *param, int is_16bit)
case BLEND_MULTIPLY: param->blend = ff_blend_multiply_sse2; break;
case BLEND_OR: param->blend = ff_blend_or_sse2; break;
case BLEND_PHOENIX: param->blend = ff_blend_phoenix_sse2; break;
case BLEND_SCREEN: param->blend = ff_blend_screen_sse2; break;
case BLEND_SUBTRACT: param->blend = ff_blend_subtract_sse2; break;
case BLEND_XOR: param->blend = ff_blend_xor_sse2; break;
case BLEND_DIFFERENCE: param->blend = ff_blend_difference_sse2; break;
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment