RepoMirrors
/
ffmpeg
mirror of https://git.ffmpeg.org/ffmpeg.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

vp9lpf/x86: save one register in loopfilter surface coverage. 

Signed-off-by: Anton Khirnov <anton@khirnov.net>
This commit is contained in:
Ronald S. Bultje 2014-12-19 21:44:57 -05:00 committed by Anton Khirnov
parent 0ed21bdc9e
commit a6e288d624
1 changed files with 28 additions and 28 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

54
libavcodec/x86/vp9lpf.asm
View File

@ -279,22 +279,22 @@ SECTION .text
%endmacro %endmacro
%macro DEFINE_REAL_P7_TO_Q7 0-1 0 %macro DEFINE_REAL_P7_TO_Q7 0-1 0
%define P7 dst1q + 2*mstrideq + %1 %define P7 dstq + 4*mstrideq + %1
%define P6 dst1q + mstrideq + %1 %define P6 dstq + mstride3q + %1
%define P5 dst1q + %1 %define P5 dstq + 2*mstrideq + %1
%define P4 dst1q + strideq + %1 %define P4 dstq + mstrideq + %1
%define P3 dstq + 4*mstrideq + %1 %define P3 dstq + %1
%define P2 dstq + mstride3q + %1 %define P2 dstq + strideq + %1
%define P1 dstq + 2*mstrideq + %1 %define P1 dstq + 2* strideq + %1
%define P0 dstq + mstrideq + %1 %define P0 dstq + stride3q + %1
%define Q0 dstq + %1 %define Q0 dstq + 4* strideq + %1
%define Q1 dstq + strideq + %1 %define Q1 dst2q + mstride3q + %1
%define Q2 dstq + 2*strideq + %1 %define Q2 dst2q + 2*mstrideq + %1
%define Q3 dstq + stride3q + %1 %define Q3 dst2q + mstrideq + %1
%define Q4 dstq + 4*strideq + %1 %define Q4 dst2q + %1
%define Q5 dst2q + mstrideq + %1 %define Q5 dst2q + strideq + %1
%define Q6 dst2q + %1 %define Q6 dst2q + 2* strideq + %1
%define Q7 dst2q + strideq + %1 %define Q7 dst2q + stride3q + %1
%endmacro %endmacro
; ..............AB -> AAAAAAAABBBBBBBB ; ..............AB -> AAAAAAAABBBBBBBB
@ -309,26 +309,26 @@ SECTION .text
%endmacro %endmacro
%macro LOOPFILTER 2 ; %1=v/h %2=size1 %macro LOOPFILTER 2 ; %1=v/h %2=size1
lea mstrideq, [strideq] mov mstrideq, strideq
neg mstrideq neg mstrideq
lea stride3q, [strideq+2*strideq] lea stride3q, [strideq*3]
mov mstride3q, stride3q lea mstride3q, [mstrideq*3]
neg mstride3q
%ifidn %1, h %ifidn %1, h
%if %2 > 16 %if %2 > 16
%define movx movh %define movx movh
lea dstq, [dstq + 8*strideq - 4] lea dstq, [dstq + 4*strideq - 4]
%else %else
%define movx movu %define movx movu
lea dstq, [dstq + 8*strideq - 8] ; go from top center (h pos) to center left (v pos) lea dstq, [dstq + 4*strideq - 8] ; go from top center (h pos) to center left (v pos)
%endif %endif
lea dst2q, [dstq + 8*strideq]
%else
lea dstq, [dstq + 4*mstrideq]
lea dst2q, [dstq + 8*strideq]
%endif %endif
lea dst1q, [dstq + 2*mstride3q] ; dst1q = &dst[stride * -6]
lea dst2q, [dstq + 2* stride3q] ; dst2q = &dst[stride * +6]
DEFINE_REAL_P7_TO_Q7 DEFINE_REAL_P7_TO_Q7
%ifidn %1, h %ifidn %1, h
@ -796,9 +796,9 @@ SECTION .text
%macro LPF_16_VH 2 %macro LPF_16_VH 2
INIT_XMM %2 INIT_XMM %2
cglobal vp9_loop_filter_v_%1_16, 5,10,16, dst, stride, E, I, H, mstride, dst1, dst2, stride3, mstride3 cglobal vp9_loop_filter_v_%1_16, 5,10,16, dst, stride, E, I, H, mstride, dst2, stride3, mstride3
LOOPFILTER v, %1 LOOPFILTER v, %1
cglobal vp9_loop_filter_h_%1_16, 5,10,16, 256, dst, stride, E, I, H, mstride, dst1, dst2, stride3, mstride3 cglobal vp9_loop_filter_h_%1_16, 5,10,16, 256, dst, stride, E, I, H, mstride, dst2, stride3, mstride3
LOOPFILTER h, %1 LOOPFILTER h, %1
%endmacro %endmacro