Skip to content

F
ffmpeg.wasm-core
Commit d20f133e authored by Diego Biurrun's avatar Diego Biurrun
Browse files
Options

x86: h264_intrapred: port to cpuflag macros

parent 2696789c
Hide whitespace changes
Inline Side-by-side
Showing with 168 additions and 146 deletions
libavcodec/x86/h264_intrapred.asm
View file @ d20f133e
...@@ -87,23 +87,23 @@ cglobal pred16x16_vertical_sse, 2,3 ...@@ -87,23 +87,23 @@ cglobal pred16x16_vertical_sse, 2,3
; void pred16x16_horizontal(uint8_t *src, int stride) ; void pred16x16_horizontal(uint8_t *src, int stride)
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
%macro PRED16x16_H 1 %macro PRED16x16_H 0
cglobal pred16x16_horizontal_%1, 2,3 cglobal pred16x16_horizontal, 2,3
mov r2, 8 mov r2, 8
%ifidn %1, ssse3 %if cpuflag(ssse3)
mova m2, [pb_3] mova m2, [pb_3]
%endif %endif
.loop: .loop:
movd m0, [r0+r1*0-4] movd m0, [r0+r1*0-4]
movd m1, [r0+r1*1-4] movd m1, [r0+r1*1-4]
%ifidn %1, ssse3 %if cpuflag(ssse3)
pshufb m0, m2 pshufb m0, m2
pshufb m1, m2 pshufb m1, m2
%else %else
punpcklbw m0, m0 punpcklbw m0, m0
punpcklbw m1, m1 punpcklbw m1, m1
%ifidn %1, mmxext %if cpuflag(mmx2)
pshufw m0, m0, 0xff pshufw m0, m0, 0xff
pshufw m1, m1, 0xff pshufw m1, m1, 0xff
%else %else
...@@ -124,18 +124,20 @@ cglobal pred16x16_horizontal_%1, 2,3 ...@@ -124,18 +124,20 @@ cglobal pred16x16_horizontal_%1, 2,3
REP_RET REP_RET
%endmacro %endmacro
INIT_MMX INIT_MMX mmx
PRED16x16_H mmx PRED16x16_H
PRED16x16_H mmxext INIT_MMX mmx2
PRED16x16_H
INIT_XMM ssse3
PRED16x16_H
INIT_XMM INIT_XMM
PRED16x16_H ssse3
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
; void pred16x16_dc(uint8_t *src, int stride) ; void pred16x16_dc(uint8_t *src, int stride)
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
%macro PRED16x16_DC 1 %macro PRED16x16_DC 0
cglobal pred16x16_dc_%1, 2,7 cglobal pred16x16_dc, 2,7
mov r4, r0 mov r4, r0
sub r0, r1 sub r0, r1
pxor mm0, mm0 pxor mm0, mm0
...@@ -158,19 +160,19 @@ cglobal pred16x16_dc_%1, 2,7 ...@@ -158,19 +160,19 @@ cglobal pred16x16_dc_%1, 2,7
add r5d, r6d add r5d, r6d
lea r2d, [r2+r5+16] lea r2d, [r2+r5+16]
shr r2d, 5 shr r2d, 5
%ifidn %1, mmxext %if cpuflag(ssse3)
pxor m1, m1
movd m0, r2d movd m0, r2d
punpcklbw m0, m0 pshufb m0, m1
pshufw m0, m0, 0 %elif cpuflag(sse2)
%elifidn %1, sse2
movd m0, r2d movd m0, r2d
punpcklbw m0, m0 punpcklbw m0, m0
pshuflw m0, m0, 0 pshuflw m0, m0, 0
punpcklqdq m0, m0 punpcklqdq m0, m0
%elifidn %1, ssse3 %elif cpuflag(mmx2)
pxor m1, m1
movd m0, r2d movd m0, r2d
pshufb m0, m1 punpcklbw m0, m0
pshufw m0, m0, 0
%endif %endif
%if mmsize==8 %if mmsize==8
...@@ -195,18 +197,20 @@ cglobal pred16x16_dc_%1, 2,7 ...@@ -195,18 +197,20 @@ cglobal pred16x16_dc_%1, 2,7
REP_RET REP_RET
%endmacro %endmacro
INIT_MMX INIT_MMX mmx2
PRED16x16_DC mmxext PRED16x16_DC
INIT_XMM sse2
PRED16x16_DC
INIT_XMM ssse3
PRED16x16_DC
INIT_XMM INIT_XMM
PRED16x16_DC sse2
PRED16x16_DC ssse3
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
; void pred16x16_tm_vp8(uint8_t *src, int stride) ; void pred16x16_tm_vp8(uint8_t *src, int stride)
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
%macro PRED16x16_TM_MMX 1 %macro PRED16x16_TM_MMX 0
cglobal pred16x16_tm_vp8_%1, 2,5 cglobal pred16x16_tm_vp8, 2,5
sub r0, r1 sub r0, r1
pxor mm7, mm7 pxor mm7, mm7
movq mm0, [r0+0] movq mm0, [r0+0]
...@@ -223,11 +227,11 @@ cglobal pred16x16_tm_vp8_%1, 2,5 ...@@ -223,11 +227,11 @@ cglobal pred16x16_tm_vp8_%1, 2,5
movzx r2d, byte [r0+r1-1] movzx r2d, byte [r0+r1-1]
sub r2d, r3d sub r2d, r3d
movd mm4, r2d movd mm4, r2d
%ifidn %1, mmx %if cpuflag(mmx2)
pshufw mm4, mm4, 0
%else
punpcklwd mm4, mm4 punpcklwd mm4, mm4
punpckldq mm4, mm4 punpckldq mm4, mm4
%else
pshufw mm4, mm4, 0
%endif %endif
movq mm5, mm4 movq mm5, mm4
movq mm6, mm4 movq mm6, mm4
...@@ -246,8 +250,11 @@ cglobal pred16x16_tm_vp8_%1, 2,5 ...@@ -246,8 +250,11 @@ cglobal pred16x16_tm_vp8_%1, 2,5
REP_RET REP_RET
%endmacro %endmacro
PRED16x16_TM_MMX mmx INIT_MMX mmx
PRED16x16_TM_MMX mmxext PRED16x16_TM_MMX
INIT_MMX mmx2
PRED16x16_TM_MMX
INIT_MMX
cglobal pred16x16_tm_vp8_sse2, 2,6,6 cglobal pred16x16_tm_vp8_sse2, 2,6,6
sub r0, r1 sub r0, r1
...@@ -288,8 +295,8 @@ cglobal pred16x16_tm_vp8_sse2, 2,6,6 ...@@ -288,8 +295,8 @@ cglobal pred16x16_tm_vp8_sse2, 2,6,6
; void pred16x16_plane(uint8_t *src, int stride) ; void pred16x16_plane(uint8_t *src, int stride)
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
%macro H264_PRED16x16_PLANE 3 %macro H264_PRED16x16_PLANE 1
cglobal pred16x16_plane_%3_%1, 2, 9, %2 cglobal pred16x16_plane_%1, 2,9,7
mov r2, r1 ; +stride mov r2, r1 ; +stride
neg r1 ; -stride neg r1 ; -stride
...@@ -310,7 +317,10 @@ cglobal pred16x16_plane_%3_%1, 2, 9, %2 ...@@ -310,7 +317,10 @@ cglobal pred16x16_plane_%3_%1, 2, 9, %2
paddw m0, m2 paddw m0, m2
paddw m1, m3 paddw m1, m3
%else ; mmsize == 16 %else ; mmsize == 16
%ifidn %1, sse2 %if cpuflag(ssse3)
movhps m0, [r0+r1 +8]
pmaddubsw m0, [plane_shuf] ; H coefficients
%else ; sse2
pxor m2, m2 pxor m2, m2
movh m1, [r0+r1 +8] movh m1, [r0+r1 +8]
punpcklbw m0, m2 punpcklbw m0, m2
...@@ -318,29 +328,26 @@ cglobal pred16x16_plane_%3_%1, 2, 9, %2 ...@@ -318,29 +328,26 @@ cglobal pred16x16_plane_%3_%1, 2, 9, %2
pmullw m0, [pw_m8tom1] pmullw m0, [pw_m8tom1]
pmullw m1, [pw_1to8] pmullw m1, [pw_1to8]
paddw m0, m1 paddw m0, m1
%else ; ssse3
movhps m0, [r0+r1 +8]
pmaddubsw m0, [plane_shuf] ; H coefficients
%endif %endif
movhlps m1, m0 movhlps m1, m0
%endif %endif
paddw m0, m1 paddw m0, m1
%ifidn %1, mmx %if cpuflag(sse2)
pshuflw m1, m0, 0xE
%elif cpuflag(mmx2)
pshufw m1, m0, 0xE
%elif cpuflag(mmx)
mova m1, m0 mova m1, m0
psrlq m1, 32 psrlq m1, 32
%elifidn %1, mmx2
pshufw m1, m0, 0xE
%else ; mmsize == 16
pshuflw m1, m0, 0xE
%endif %endif
paddw m0, m1 paddw m0, m1
%ifidn %1, mmx %if cpuflag(sse2)
pshuflw m1, m0, 0x1
%elif cpuflag(mmx2)
pshufw m1, m0, 0x1
%elif cpuflag(mmx)
mova m1, m0 mova m1, m0
psrlq m1, 16 psrlq m1, 16
%elifidn %1, mmx2
pshufw m1, m0, 0x1
%else
pshuflw m1, m0, 0x1
%endif %endif
paddw m0, m1 ; sum of H coefficients paddw m0, m1 ; sum of H coefficients
...@@ -424,13 +431,13 @@ cglobal pred16x16_plane_%3_%1, 2, 9, %2 ...@@ -424,13 +431,13 @@ cglobal pred16x16_plane_%3_%1, 2, 9, %2
mov r0, r0m mov r0, r0m
%endif %endif
%ifidn %3, h264 %ifidn %1, h264
lea r5, [r5*5+32] lea r5, [r5*5+32]
sar r5, 6 sar r5, 6
%elifidn %3, rv40 %elifidn %1, rv40
lea r5, [r5*5] lea r5, [r5*5]
sar r5, 6 sar r5, 6
%elifidn %3, svq3 %elifidn %1, svq3
test r5, r5 test r5, r5
lea r6, [r5+3] lea r6, [r5+3]
cmovs r5, r6 cmovs r5, r6
...@@ -449,8 +456,8 @@ cglobal pred16x16_plane_%3_%1, 2, 9, %2 ...@@ -449,8 +456,8 @@ cglobal pred16x16_plane_%3_%1, 2, 9, %2
movd r1d, m0 movd r1d, m0
movsx r1d, r1w movsx r1d, r1w
%ifnidn %3, svq3 %ifnidn %1, svq3
%ifidn %3, h264 %ifidn %1, h264
lea r1d, [r1d*5+32] lea r1d, [r1d*5+32]
%else ; rv40 %else ; rv40
lea r1d, [r1d*5] lea r1d, [r1d*5]
...@@ -476,26 +483,26 @@ cglobal pred16x16_plane_%3_%1, 2, 9, %2 ...@@ -476,26 +483,26 @@ cglobal pred16x16_plane_%3_%1, 2, 9, %2
movd m1, r5d movd m1, r5d
movd m3, r3d movd m3, r3d
%ifidn %1, mmx %if cpuflag(sse2)
punpcklwd m0, m0
punpcklwd m1, m1
punpcklwd m3, m3
punpckldq m0, m0
punpckldq m1, m1
punpckldq m3, m3
%elifidn %1, mmx2
pshufw m0, m0, 0x0
pshufw m1, m1, 0x0
pshufw m3, m3, 0x0
%else
pshuflw m0, m0, 0x0 pshuflw m0, m0, 0x0
pshuflw m1, m1, 0x0 pshuflw m1, m1, 0x0
pshuflw m3, m3, 0x0 pshuflw m3, m3, 0x0
punpcklqdq m0, m0 ; splat H (words) punpcklqdq m0, m0 ; splat H (words)
punpcklqdq m1, m1 ; splat V (words) punpcklqdq m1, m1 ; splat V (words)
punpcklqdq m3, m3 ; splat a (words) punpcklqdq m3, m3 ; splat a (words)
%elif cpuflag(mmx2)
pshufw m0, m0, 0x0
pshufw m1, m1, 0x0
pshufw m3, m3, 0x0
%elif cpuflag(mmx)
punpcklwd m0, m0
punpcklwd m1, m1
punpcklwd m3, m3
punpckldq m0, m0
punpckldq m1, m1
punpckldq m3, m3
%endif %endif
%ifidn %3, svq3 %ifidn %1, svq3
SWAP 0, 1 SWAP 0, 1
%endif %endif
mova m2, m0 mova m2, m0
...@@ -568,27 +575,30 @@ cglobal pred16x16_plane_%3_%1, 2, 9, %2 ...@@ -568,27 +575,30 @@ cglobal pred16x16_plane_%3_%1, 2, 9, %2
REP_RET REP_RET
%endmacro %endmacro
INIT_MMX INIT_MMX mmx
H264_PRED16x16_PLANE mmx, 0, h264 H264_PRED16x16_PLANE h264
H264_PRED16x16_PLANE mmx, 0, rv40 H264_PRED16x16_PLANE rv40
H264_PRED16x16_PLANE mmx, 0, svq3 H264_PRED16x16_PLANE svq3
H264_PRED16x16_PLANE mmx2, 0, h264 INIT_MMX mmx2
H264_PRED16x16_PLANE mmx2, 0, rv40 H264_PRED16x16_PLANE h264
H264_PRED16x16_PLANE mmx2, 0, svq3 H264_PRED16x16_PLANE rv40
H264_PRED16x16_PLANE svq3
INIT_XMM sse2
H264_PRED16x16_PLANE h264
H264_PRED16x16_PLANE rv40
H264_PRED16x16_PLANE svq3
INIT_XMM ssse3
H264_PRED16x16_PLANE h264
H264_PRED16x16_PLANE rv40
H264_PRED16x16_PLANE svq3
INIT_XMM INIT_XMM
H264_PRED16x16_PLANE sse2, 8, h264
H264_PRED16x16_PLANE sse2, 8, rv40
H264_PRED16x16_PLANE sse2, 8, svq3
H264_PRED16x16_PLANE ssse3, 8, h264
H264_PRED16x16_PLANE ssse3, 8, rv40
H264_PRED16x16_PLANE ssse3, 8, svq3
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
; void pred8x8_plane(uint8_t *src, int stride) ; void pred8x8_plane(uint8_t *src, int stride)
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
%macro H264_PRED8x8_PLANE 2 %macro H264_PRED8x8_PLANE 0
cglobal pred8x8_plane_%1, 2, 9, %2 cglobal pred8x8_plane, 2,9,7
mov r2, r1 ; +stride mov r2, r1 ; +stride
neg r1 ; -stride neg r1 ; -stride
...@@ -601,39 +611,39 @@ cglobal pred8x8_plane_%1, 2, 9, %2 ...@@ -601,39 +611,39 @@ cglobal pred8x8_plane_%1, 2, 9, %2
pmullw m0, [pw_m4to4] pmullw m0, [pw_m4to4]
pmullw m1, [pw_m4to4+8] pmullw m1, [pw_m4to4+8]
%else ; mmsize == 16 %else ; mmsize == 16
%ifidn %1, sse2 %if cpuflag(ssse3)
movhps m0, [r0+r1 +4] ; this reads 4 bytes more than necessary
pmaddubsw m0, [plane8_shuf] ; H coefficients
%else ; sse2
pxor m2, m2 pxor m2, m2
movd m1, [r0+r1 +4] movd m1, [r0+r1 +4]
punpckldq m0, m1 punpckldq m0, m1
punpcklbw m0, m2 punpcklbw m0, m2
pmullw m0, [pw_m4to4] pmullw m0, [pw_m4to4]
%else ; ssse3
movhps m0, [r0+r1 +4] ; this reads 4 bytes more than necessary
pmaddubsw m0, [plane8_shuf] ; H coefficients
%endif %endif
movhlps m1, m0 movhlps m1, m0
%endif %endif
paddw m0, m1 paddw m0, m1
%ifnidn %1, ssse3 %if notcpuflag(ssse3)
%ifidn %1, mmx %if cpuflag(sse2) ; mmsize == 16
pshuflw m1, m0, 0xE
%elif cpuflag(mmx2)
pshufw m1, m0, 0xE
%elif cpuflag(mmx)
mova m1, m0 mova m1, m0
psrlq m1, 32 psrlq m1, 32
%elifidn %1, mmx2
pshufw m1, m0, 0xE
%else ; mmsize == 16
pshuflw m1, m0, 0xE
%endif %endif
paddw m0, m1 paddw m0, m1
%endif ; !ssse3 %endif ; !ssse3
%ifidn %1, mmx %if cpuflag(sse2)
pshuflw m1, m0, 0x1
%elif cpuflag(mmx2)
pshufw m1, m0, 0x1
%elif cpuflag(mmx)
mova m1, m0 mova m1, m0
psrlq m1, 16 psrlq m1, 16
%elifidn %1, mmx2
pshufw m1, m0, 0x1
%else
pshuflw m1, m0, 0x1
%endif %endif
paddw m0, m1 ; sum of H coefficients paddw m0, m1 ; sum of H coefficients
...@@ -701,24 +711,24 @@ cglobal pred8x8_plane_%1, 2, 9, %2 ...@@ -701,24 +711,24 @@ cglobal pred8x8_plane_%1, 2, 9, %2
movd m1, r5d movd m1, r5d
movd m3, r3d movd m3, r3d
%ifidn %1, mmx %if cpuflag(sse2)
punpcklwd m0, m0
punpcklwd m1, m1
punpcklwd m3, m3
punpckldq m0, m0
punpckldq m1, m1
punpckldq m3, m3
%elifidn %1, mmx2
pshufw m0, m0, 0x0
pshufw m1, m1, 0x0
pshufw m3, m3, 0x0
%else
pshuflw m0, m0, 0x0 pshuflw m0, m0, 0x0
pshuflw m1, m1, 0x0 pshuflw m1, m1, 0x0
pshuflw m3, m3, 0x0 pshuflw m3, m3, 0x0
punpcklqdq m0, m0 ; splat H (words) punpcklqdq m0, m0 ; splat H (words)
punpcklqdq m1, m1 ; splat V (words) punpcklqdq m1, m1 ; splat V (words)
punpcklqdq m3, m3 ; splat a (words) punpcklqdq m3, m3 ; splat a (words)
%elif cpuflag(mmx2)
pshufw m0, m0, 0x0
pshufw m1, m1, 0x0
pshufw m3, m3, 0x0
%elif cpuflag(mmx)
punpcklwd m0, m0
punpcklwd m1, m1
punpcklwd m3, m3
punpckldq m0, m0
punpckldq m1, m1
punpckldq m3, m3
%endif %endif
%if mmsize == 8 %if mmsize == 8
mova m2, m0 mova m2, m0
...@@ -768,12 +778,15 @@ ALIGN 16 ...@@ -768,12 +778,15 @@ ALIGN 16
REP_RET REP_RET
%endmacro %endmacro
INIT_MMX INIT_MMX mmx
H264_PRED8x8_PLANE mmx, 0 H264_PRED8x8_PLANE
H264_PRED8x8_PLANE mmx2, 0 INIT_MMX mmx2
H264_PRED8x8_PLANE
INIT_XMM sse2
H264_PRED8x8_PLANE
INIT_XMM ssse3
H264_PRED8x8_PLANE
INIT_XMM INIT_XMM
H264_PRED8x8_PLANE sse2, 8
H264_PRED8x8_PLANE ssse3, 8
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
; void pred8x8_vertical(uint8_t *src, int stride) ; void pred8x8_vertical(uint8_t *src, int stride)
...@@ -795,22 +808,22 @@ cglobal pred8x8_vertical_mmx, 2,2 ...@@ -795,22 +808,22 @@ cglobal pred8x8_vertical_mmx, 2,2
; void pred8x8_horizontal(uint8_t *src, int stride) ; void pred8x8_horizontal(uint8_t *src, int stride)
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
%macro PRED8x8_H 1 %macro PRED8x8_H 0
cglobal pred8x8_horizontal_%1, 2,3 cglobal pred8x8_horizontal, 2,3
mov r2, 4 mov r2, 4
%ifidn %1, ssse3 %if cpuflag(ssse3)
mova m2, [pb_3] mova m2, [pb_3]
%endif %endif
.loop: .loop:
movd m0, [r0+r1*0-4] movd m0, [r0+r1*0-4]
movd m1, [r0+r1*1-4] movd m1, [r0+r1*1-4]
%ifidn %1, ssse3 %if cpuflag(ssse3)
pshufb m0, m2 pshufb m0, m2
pshufb m1, m2 pshufb m1, m2
%else %else
punpcklbw m0, m0 punpcklbw m0, m0
punpcklbw m1, m1 punpcklbw m1, m1
%ifidn %1, mmxext %if cpuflag(mmx2)
pshufw m0, m0, 0xff pshufw m0, m0, 0xff
pshufw m1, m1, 0xff pshufw m1, m1, 0xff
%else %else
...@@ -828,10 +841,13 @@ cglobal pred8x8_horizontal_%1, 2,3 ...@@ -828,10 +841,13 @@ cglobal pred8x8_horizontal_%1, 2,3
REP_RET REP_RET
%endmacro %endmacro
INIT_MMX mmx
PRED8x8_H
INIT_MMX mmx2
PRED8x8_H
INIT_MMX ssse3
PRED8x8_H
INIT_MMX INIT_MMX
PRED8x8_H mmx
PRED8x8_H mmxext
PRED8x8_H ssse3
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
; void pred8x8_top_dc_mmxext(uint8_t *src, int stride) ; void pred8x8_top_dc_mmxext(uint8_t *src, int stride)
...@@ -967,8 +983,8 @@ cglobal pred8x8_dc_rv40_mmxext, 2,7 ...@@ -967,8 +983,8 @@ cglobal pred8x8_dc_rv40_mmxext, 2,7
; void pred8x8_tm_vp8(uint8_t *src, int stride) ; void pred8x8_tm_vp8(uint8_t *src, int stride)
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
%macro PRED8x8_TM_MMX 1 %macro PRED8x8_TM_MMX 0
cglobal pred8x8_tm_vp8_%1, 2,6 cglobal pred8x8_tm_vp8, 2,6
sub r0, r1 sub r0, r1
pxor mm7, mm7 pxor mm7, mm7
movq mm0, [r0] movq mm0, [r0]
...@@ -984,14 +1000,14 @@ cglobal pred8x8_tm_vp8_%1, 2,6 ...@@ -984,14 +1000,14 @@ cglobal pred8x8_tm_vp8_%1, 2,6
sub r3d, r4d sub r3d, r4d
movd mm2, r2d movd mm2, r2d
movd mm4, r3d movd mm4, r3d
%ifidn %1, mmx %if cpuflag(mmx2)
pshufw mm2, mm2, 0
pshufw mm4, mm4, 0
%else
punpcklwd mm2, mm2 punpcklwd mm2, mm2
punpcklwd mm4, mm4 punpcklwd mm4, mm4
punpckldq mm2, mm2 punpckldq mm2, mm2
punpckldq mm4, mm4 punpckldq mm4, mm4
%else
pshufw mm2, mm2, 0
pshufw mm4, mm4, 0
%endif %endif
movq mm3, mm2 movq mm3, mm2
movq mm5, mm4 movq mm5, mm4
...@@ -1009,8 +1025,11 @@ cglobal pred8x8_tm_vp8_%1, 2,6 ...@@ -1009,8 +1025,11 @@ cglobal pred8x8_tm_vp8_%1, 2,6
REP_RET REP_RET
%endmacro %endmacro
PRED8x8_TM_MMX mmx INIT_MMX mmx
PRED8x8_TM_MMX mmxext PRED8x8_TM_MMX
INIT_MMX mmx2
PRED8x8_TM_MMX
INIT_MMX
cglobal pred8x8_tm_vp8_sse2, 2,6,4 cglobal pred8x8_tm_vp8_sse2, 2,6,4
sub r0, r1 sub r0, r1
...@@ -2510,8 +2529,8 @@ cglobal pred4x4_dc_mmxext, 3,5 ...@@ -2510,8 +2529,8 @@ cglobal pred4x4_dc_mmxext, 3,5
; void pred4x4_tm_vp8_mmxext(uint8_t *src, const uint8_t *topright, int stride) ; void pred4x4_tm_vp8_mmxext(uint8_t *src, const uint8_t *topright, int stride)
;----------------------------------------------------------------------------- ;-----------------------------------------------------------------------------
%macro PRED4x4_TM_MMX 1 %macro PRED4x4_TM_MMX 0
cglobal pred4x4_tm_vp8_%1, 3,6 cglobal pred4x4_tm_vp8, 3,6
sub r0, r2 sub r0, r2
pxor mm7, mm7 pxor mm7, mm7
movd mm0, [r0] movd mm0, [r0]
...@@ -2525,14 +2544,14 @@ cglobal pred4x4_tm_vp8_%1, 3,6 ...@@ -2525,14 +2544,14 @@ cglobal pred4x4_tm_vp8_%1, 3,6
sub r3d, r4d sub r3d, r4d
movd mm2, r1d movd mm2, r1d
movd mm4, r3d movd mm4, r3d
%ifidn %1, mmx %if cpuflag(mmx2)
pshufw mm2, mm2, 0
pshufw mm4, mm4, 0
%else
punpcklwd mm2, mm2 punpcklwd mm2, mm2
punpcklwd mm4, mm4 punpcklwd mm4, mm4
punpckldq mm2, mm2 punpckldq mm2, mm2
punpckldq mm4, mm4 punpckldq mm4, mm4
%else
pshufw mm2, mm2, 0
pshufw mm4, mm4, 0
%endif %endif
paddw mm2, mm0 paddw mm2, mm0
paddw mm4, mm0 paddw mm4, mm0
...@@ -2546,8 +2565,11 @@ cglobal pred4x4_tm_vp8_%1, 3,6 ...@@ -2546,8 +2565,11 @@ cglobal pred4x4_tm_vp8_%1, 3,6
REP_RET REP_RET
%endmacro %endmacro
PRED4x4_TM_MMX mmx INIT_MMX mmx
PRED4x4_TM_MMX mmxext PRED4x4_TM_MMX
INIT_MMX mmx2
PRED4x4_TM_MMX
INIT_MMX
cglobal pred4x4_tm_vp8_ssse3, 3,3 cglobal pred4x4_tm_vp8_ssse3, 3,3
sub r0, r2 sub r0, r2
......
libavcodec/x86/h264_intrapred_init.c
View file @ d20f133e
...@@ -96,9 +96,9 @@ PRED16x16(horizontal, 10, sse2) ...@@ -96,9 +96,9 @@ PRED16x16(horizontal, 10, sse2)
void ff_pred16x16_vertical_mmx (uint8_t *src, int stride); void ff_pred16x16_vertical_mmx (uint8_t *src, int stride);
void ff_pred16x16_vertical_sse (uint8_t *src, int stride); void ff_pred16x16_vertical_sse (uint8_t *src, int stride);
void ff_pred16x16_horizontal_mmx (uint8_t *src, int stride); void ff_pred16x16_horizontal_mmx (uint8_t *src, int stride);
void ff_pred16x16_horizontal_mmxext(uint8_t *src, int stride); void ff_pred16x16_horizontal_mmx2 (uint8_t *src, int stride);
void ff_pred16x16_horizontal_ssse3 (uint8_t *src, int stride); void ff_pred16x16_horizontal_ssse3 (uint8_t *src, int stride);
void ff_pred16x16_dc_mmxext (uint8_t *src, int stride); void ff_pred16x16_dc_mmx2 (uint8_t *src, int stride);
void ff_pred16x16_dc_sse2 (uint8_t *src, int stride); void ff_pred16x16_dc_sse2 (uint8_t *src, int stride);
void ff_pred16x16_dc_ssse3 (uint8_t *src, int stride); void ff_pred16x16_dc_ssse3 (uint8_t *src, int stride);
void ff_pred16x16_plane_h264_mmx (uint8_t *src, int stride); void ff_pred16x16_plane_h264_mmx (uint8_t *src, int stride);
...@@ -114,21 +114,21 @@ void ff_pred16x16_plane_svq3_mmx2 (uint8_t *src, int stride); ...@@ -114,21 +114,21 @@ void ff_pred16x16_plane_svq3_mmx2 (uint8_t *src, int stride);
void ff_pred16x16_plane_svq3_sse2 (uint8_t *src, int stride); void ff_pred16x16_plane_svq3_sse2 (uint8_t *src, int stride);
void ff_pred16x16_plane_svq3_ssse3 (uint8_t *src, int stride); void ff_pred16x16_plane_svq3_ssse3 (uint8_t *src, int stride);
void ff_pred16x16_tm_vp8_mmx (uint8_t *src, int stride); void ff_pred16x16_tm_vp8_mmx (uint8_t *src, int stride);
void ff_pred16x16_tm_vp8_mmxext (uint8_t *src, int stride); void ff_pred16x16_tm_vp8_mmx2 (uint8_t *src, int stride);
void ff_pred16x16_tm_vp8_sse2 (uint8_t *src, int stride); void ff_pred16x16_tm_vp8_sse2 (uint8_t *src, int stride);
void ff_pred8x8_top_dc_mmxext (uint8_t *src, int stride); void ff_pred8x8_top_dc_mmxext (uint8_t *src, int stride);
void ff_pred8x8_dc_rv40_mmxext (uint8_t *src, int stride); void ff_pred8x8_dc_rv40_mmxext (uint8_t *src, int stride);
void ff_pred8x8_dc_mmxext (uint8_t *src, int stride); void ff_pred8x8_dc_mmxext (uint8_t *src, int stride);
void ff_pred8x8_vertical_mmx (uint8_t *src, int stride); void ff_pred8x8_vertical_mmx (uint8_t *src, int stride);
void ff_pred8x8_horizontal_mmx (uint8_t *src, int stride); void ff_pred8x8_horizontal_mmx (uint8_t *src, int stride);
void ff_pred8x8_horizontal_mmxext (uint8_t *src, int stride); void ff_pred8x8_horizontal_mmx2 (uint8_t *src, int stride);
void ff_pred8x8_horizontal_ssse3 (uint8_t *src, int stride); void ff_pred8x8_horizontal_ssse3 (uint8_t *src, int stride);
void ff_pred8x8_plane_mmx (uint8_t *src, int stride); void ff_pred8x8_plane_mmx (uint8_t *src, int stride);
void ff_pred8x8_plane_mmx2 (uint8_t *src, int stride); void ff_pred8x8_plane_mmx2 (uint8_t *src, int stride);
void ff_pred8x8_plane_sse2 (uint8_t *src, int stride); void ff_pred8x8_plane_sse2 (uint8_t *src, int stride);
void ff_pred8x8_plane_ssse3 (uint8_t *src, int stride); void ff_pred8x8_plane_ssse3 (uint8_t *src, int stride);
void ff_pred8x8_tm_vp8_mmx (uint8_t *src, int stride); void ff_pred8x8_tm_vp8_mmx (uint8_t *src, int stride);
void ff_pred8x8_tm_vp8_mmxext (uint8_t *src, int stride); void ff_pred8x8_tm_vp8_mmx2 (uint8_t *src, int stride);
void ff_pred8x8_tm_vp8_sse2 (uint8_t *src, int stride); void ff_pred8x8_tm_vp8_sse2 (uint8_t *src, int stride);
void ff_pred8x8_tm_vp8_ssse3 (uint8_t *src, int stride); void ff_pred8x8_tm_vp8_ssse3 (uint8_t *src, int stride);
void ff_pred8x8l_top_dc_mmxext (uint8_t *src, int has_topleft, int has_topright, int stride); void ff_pred8x8l_top_dc_mmxext (uint8_t *src, int has_topleft, int has_topright, int stride);
...@@ -163,7 +163,7 @@ void ff_pred4x4_vertical_right_mmxext(uint8_t *src, const uint8_t *topright, int ...@@ -163,7 +163,7 @@ void ff_pred4x4_vertical_right_mmxext(uint8_t *src, const uint8_t *topright, int
void ff_pred4x4_horizontal_up_mmxext(uint8_t *src, const uint8_t *topright, int stride); void ff_pred4x4_horizontal_up_mmxext(uint8_t *src, const uint8_t *topright, int stride);
void ff_pred4x4_horizontal_down_mmxext(uint8_t *src, const uint8_t *topright, int stride); void ff_pred4x4_horizontal_down_mmxext(uint8_t *src, const uint8_t *topright, int stride);
void ff_pred4x4_tm_vp8_mmx (uint8_t *src, const uint8_t *topright, int stride); void ff_pred4x4_tm_vp8_mmx (uint8_t *src, const uint8_t *topright, int stride);
void ff_pred4x4_tm_vp8_mmxext (uint8_t *src, const uint8_t *topright, int stride); void ff_pred4x4_tm_vp8_mmx2 (uint8_t *src, const uint8_t *topright, int stride);
void ff_pred4x4_tm_vp8_ssse3 (uint8_t *src, const uint8_t *topright, int stride); void ff_pred4x4_tm_vp8_ssse3 (uint8_t *src, const uint8_t *topright, int stride);
void ff_pred4x4_vertical_vp8_mmxext(uint8_t *src, const uint8_t *topright, int stride); void ff_pred4x4_vertical_vp8_mmxext(uint8_t *src, const uint8_t *topright, int stride);
...@@ -199,10 +199,10 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth ...@@ -199,10 +199,10 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
} }
if (mm_flags & AV_CPU_FLAG_MMX2) { if (mm_flags & AV_CPU_FLAG_MMX2) {
h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_mmxext; h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_mmx2;
h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_mmxext; h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_mmx2;
if (chroma_format_idc == 1) if (chroma_format_idc == 1)
h->pred8x8[HOR_PRED8x8 ] = ff_pred8x8_horizontal_mmxext; h->pred8x8[HOR_PRED8x8 ] = ff_pred8x8_horizontal_mmx2;
h->pred8x8l [TOP_DC_PRED ] = ff_pred8x8l_top_dc_mmxext; h->pred8x8l [TOP_DC_PRED ] = ff_pred8x8l_top_dc_mmxext;
h->pred8x8l [DC_PRED ] = ff_pred8x8l_dc_mmxext; h->pred8x8l [DC_PRED ] = ff_pred8x8l_dc_mmxext;
h->pred8x8l [HOR_PRED ] = ff_pred8x8l_horizontal_mmxext; h->pred8x8l [HOR_PRED ] = ff_pred8x8l_horizontal_mmxext;
...@@ -232,10 +232,10 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth ...@@ -232,10 +232,10 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
} }
} }
if (codec_id == CODEC_ID_VP8) { if (codec_id == CODEC_ID_VP8) {
h->pred16x16[PLANE_PRED8x8 ] = ff_pred16x16_tm_vp8_mmxext; h->pred16x16[PLANE_PRED8x8 ] = ff_pred16x16_tm_vp8_mmx2;
h->pred8x8 [DC_PRED8x8 ] = ff_pred8x8_dc_rv40_mmxext; h->pred8x8 [DC_PRED8x8 ] = ff_pred8x8_dc_rv40_mmxext;
h->pred8x8 [PLANE_PRED8x8 ] = ff_pred8x8_tm_vp8_mmxext; h->pred8x8 [PLANE_PRED8x8 ] = ff_pred8x8_tm_vp8_mmx2;
h->pred4x4 [TM_VP8_PRED ] = ff_pred4x4_tm_vp8_mmxext; h->pred4x4 [TM_VP8_PRED ] = ff_pred4x4_tm_vp8_mmx2;
h->pred4x4 [VERT_PRED ] = ff_pred4x4_vertical_vp8_mmxext; h->pred4x4 [VERT_PRED ] = ff_pred4x4_vertical_vp8_mmxext;
} else { } else {
if (chroma_format_idc == 1) if (chroma_format_idc == 1)
......
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