.src_unaligned:
ADD_HFYU_LEFT_LOOP 0, 0
-
-; float scalarproduct_float_sse(const float *v1, const float *v2, int len)
-INIT_XMM sse
-cglobal scalarproduct_float, 3,3,2, v1, v2, offset
- neg offsetq
- shl offsetq, 2
- sub v1q, offsetq
- sub v2q, offsetq
- xorps xmm0, xmm0
- .loop:
- movaps xmm1, [v1q+offsetq]
- mulps xmm1, [v2q+offsetq]
- addps xmm0, xmm1
- add offsetq, 16
- js .loop
- movhlps xmm1, xmm0
- addps xmm0, xmm1
- movss xmm1, xmm0
- shufps xmm0, xmm0, 1
- addss xmm0, xmm1
-%if ARCH_X86_64 == 0
- movss r0m, xmm0
- fld dword r0m
-%endif
- RET
-
;-----------------------------------------------------------------------------
; void ff_vector_clip_int32(int32_t *dst, const int32_t *src, int32_t min,
; int32_t max, unsigned int len)
VECTOR_CLIP_INT32 6, 1, 0, 0
%endif
-;-----------------------------------------------------------------------------
-; void vector_fmul_reverse(float *dst, const float *src0, const float *src1,
-; int len)
-;-----------------------------------------------------------------------------
-%macro VECTOR_FMUL_REVERSE 0
-cglobal vector_fmul_reverse, 4,4,2, dst, src0, src1, len
- lea lenq, [lend*4 - 2*mmsize]
-ALIGN 16
-.loop:
-%if cpuflag(avx)
- vmovaps xmm0, [src1q + 16]
- vinsertf128 m0, m0, [src1q], 1
- vshufps m0, m0, m0, q0123
- vmovaps xmm1, [src1q + mmsize + 16]
- vinsertf128 m1, m1, [src1q + mmsize], 1
- vshufps m1, m1, m1, q0123
-%else
- mova m0, [src1q]
- mova m1, [src1q + mmsize]
- shufps m0, m0, q0123
- shufps m1, m1, q0123
-%endif
- mulps m0, m0, [src0q + lenq + mmsize]
- mulps m1, m1, [src0q + lenq]
- mova [dstq + lenq + mmsize], m0
- mova [dstq + lenq], m1
- add src1q, 2*mmsize
- sub lenq, 2*mmsize
- jge .loop
- REP_RET
-%endmacro
-
-INIT_XMM sse
-VECTOR_FMUL_REVERSE
-INIT_YMM avx
-VECTOR_FMUL_REVERSE
-
-;-----------------------------------------------------------------------------
-; vector_fmul_add(float *dst, const float *src0, const float *src1,
-; const float *src2, int len)
-;-----------------------------------------------------------------------------
-%macro VECTOR_FMUL_ADD 0
-cglobal vector_fmul_add, 5,5,2, dst, src0, src1, src2, len
- lea lenq, [lend*4 - 2*mmsize]
-ALIGN 16
-.loop:
- mova m0, [src0q + lenq]
- mova m1, [src0q + lenq + mmsize]
- mulps m0, m0, [src1q + lenq]
- mulps m1, m1, [src1q + lenq + mmsize]
- addps m0, m0, [src2q + lenq]
- addps m1, m1, [src2q + lenq + mmsize]
- mova [dstq + lenq], m0
- mova [dstq + lenq + mmsize], m1
-
- sub lenq, 2*mmsize
- jge .loop
- REP_RET
-%endmacro
-
-INIT_XMM sse
-VECTOR_FMUL_ADD
-INIT_YMM avx
-VECTOR_FMUL_ADD
-
; %1 = aligned/unaligned
%macro BSWAP_LOOPS 1
mov r3, r2
%if cpuflag(ssse3)
pshufb m0, m2
pshufb m1, m2
- mova [r0 + 0], m0
- mova [r0 + 16], m1
+ mov%1 [r0 + 0], m0
+ mov%1 [r0 + 16], m1
%else
pshuflw m0, m0, 10110001b
pshuflw m1, m1, 10110001b
psrlw m3, 8
por m2, m0
por m3, m1
- mova [r0 + 0], m2
- mova [r0 + 16], m3
+ mov%1 [r0 + 0], m2
+ mov%1 [r0 + 16], m3
%endif
add r0, 32
add r1, 32
mov%1 m0, [r1]
%if cpuflag(ssse3)
pshufb m0, m2
- mova [r0], m0
+ mov%1 [r0], m0
%else
pshuflw m0, m0, 10110001b
pshufhw m0, m0, 10110001b
psllw m0, 8
psrlw m2, 8
por m2, m0
- mova [r0], m2
+ mov%1 [r0], m2
%endif
add r1, 16
add r0, 16
INIT_XMM ssse3
BSWAP32_BUF
-
-%macro op_avgh 3
- movh %3, %2
- pavgb %1, %3
- movh %2, %1
-%endmacro
-
-%macro op_avg 2
- pavgb %1, %2
- mova %2, %1
-%endmacro
-
-%macro op_puth 2-3
- movh %2, %1
-%endmacro
-
-%macro op_put 2
- mova %2, %1
-%endmacro
-
-; void pixels4_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
-%macro PIXELS4_L2 1
-%define OP op_%1h
-cglobal %1_pixels4_l2, 6,6
- movsxdifnidn r3, r3d
- movsxdifnidn r4, r4d
- test r5d, 1
- je .loop
- movd m0, [r1]
- movd m1, [r2]
- add r1, r4
- add r2, 4
- pavgb m0, m1
- OP m0, [r0], m3
- add r0, r3
- dec r5d
-.loop:
- mova m0, [r1]
- mova m1, [r1+r4]
- lea r1, [r1+2*r4]
- pavgb m0, [r2]
- pavgb m1, [r2+4]
- OP m0, [r0], m3
- OP m1, [r0+r3], m3
- lea r0, [r0+2*r3]
- mova m0, [r1]
- mova m1, [r1+r4]
- lea r1, [r1+2*r4]
- pavgb m0, [r2+8]
- pavgb m1, [r2+12]
- OP m0, [r0], m3
- OP m1, [r0+r3], m3
- lea r0, [r0+2*r3]
- add r2, 16
- sub r5d, 4
- jne .loop
- REP_RET
-%endmacro
-
-INIT_MMX mmxext
-PIXELS4_L2 put
-PIXELS4_L2 avg
-
-; void pixels8_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
-%macro PIXELS8_L2 1
-%define OP op_%1
-cglobal %1_pixels8_l2, 6,6
- movsxdifnidn r3, r3d
- movsxdifnidn r4, r4d
- test r5d, 1
- je .loop
- mova m0, [r1]
- mova m1, [r2]
- add r1, r4
- add r2, 8
- pavgb m0, m1
- OP m0, [r0]
- add r0, r3
- dec r5d
-.loop:
- mova m0, [r1]
- mova m1, [r1+r4]
- lea r1, [r1+2*r4]
- pavgb m0, [r2]
- pavgb m1, [r2+8]
- OP m0, [r0]
- OP m1, [r0+r3]
- lea r0, [r0+2*r3]
- mova m0, [r1]
- mova m1, [r1+r4]
- lea r1, [r1+2*r4]
- pavgb m0, [r2+16]
- pavgb m1, [r2+24]
- OP m0, [r0]
- OP m1, [r0+r3]
- lea r0, [r0+2*r3]
- add r2, 32
- sub r5d, 4
- jne .loop
- REP_RET
-%endmacro
-
-INIT_MMX mmxext
-PIXELS8_L2 put
-PIXELS8_L2 avg
-
-; void pixels16_l2_mmxext(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
-%macro PIXELS16_L2 1
-%define OP op_%1
-cglobal %1_pixels16_l2, 6,6
- movsxdifnidn r3, r3d
- movsxdifnidn r4, r4d
- test r5d, 1
- je .loop
- mova m0, [r1]
- mova m1, [r1+8]
- pavgb m0, [r2]
- pavgb m1, [r2+8]
- add r1, r4
- add r2, 16
- OP m0, [r0]
- OP m1, [r0+8]
- add r0, r3
- dec r5d
-.loop:
- mova m0, [r1]
- mova m1, [r1+8]
- add r1, r4
- pavgb m0, [r2]
- pavgb m1, [r2+8]
- OP m0, [r0]
- OP m1, [r0+8]
- add r0, r3
- mova m0, [r1]
- mova m1, [r1+8]
- add r1, r4
- pavgb m0, [r2+16]
- pavgb m1, [r2+24]
- OP m0, [r0]
- OP m1, [r0+8]
- add r0, r3
- add r2, 32
- sub r5d, 2
- jne .loop
- REP_RET
-%endmacro
-
-INIT_MMX mmxext
-PIXELS16_L2 put
-PIXELS16_L2 avg
-
-INIT_MMX mmxext
-; void pixels(uint8_t *block, const uint8_t *pixels, int line_size, int h)
-%macro PIXELS48 2
-%if %2 == 4
-%define OP movh
-%else
-%define OP mova
-%endif
-cglobal %1_pixels%2, 4,5
- movsxdifnidn r2, r2d
- lea r4, [r2*3]
-.loop:
- OP m0, [r1]
- OP m1, [r1+r2]
- OP m2, [r1+r2*2]
- OP m3, [r1+r4]
- lea r1, [r1+r2*4]
-%ifidn %1, avg
- pavgb m0, [r0]
- pavgb m1, [r0+r2]
- pavgb m2, [r0+r2*2]
- pavgb m3, [r0+r4]
-%endif
- OP [r0], m0
- OP [r0+r2], m1
- OP [r0+r2*2], m2
- OP [r0+r4], m3
- sub r3d, 4
- lea r0, [r0+r2*4]
- jne .loop
- RET
-%endmacro
-
-PIXELS48 put, 4
-PIXELS48 avg, 4
-PIXELS48 put, 8
-PIXELS48 avg, 8
-
-INIT_XMM sse2
-; void put_pixels16_sse2(uint8_t *block, const uint8_t *pixels, int line_size, int h)
-cglobal put_pixels16, 4,5,4
- movsxdifnidn r2, r2d
- lea r4, [r2*3]
-.loop:
- movu m0, [r1]
- movu m1, [r1+r2]
- movu m2, [r1+r2*2]
- movu m3, [r1+r4]
- lea r1, [r1+r2*4]
- mova [r0], m0
- mova [r0+r2], m1
- mova [r0+r2*2], m2
- mova [r0+r4], m3
- sub r3d, 4
- lea r0, [r0+r2*4]
- jnz .loop
- REP_RET
-
-; void avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels, int line_size, int h)
-cglobal avg_pixels16, 4,5,4
- movsxdifnidn r2, r2d
- lea r4, [r2*3]
-.loop:
- movu m0, [r1]
- movu m1, [r1+r2]
- movu m2, [r1+r2*2]
- movu m3, [r1+r4]
- lea r1, [r1+r2*4]
- pavgb m0, [r0]
- pavgb m1, [r0+r2]
- pavgb m2, [r0+r2*2]
- pavgb m3, [r0+r4]
- mova [r0], m0
- mova [r0+r2], m1
- mova [r0+r2*2], m2
- mova [r0+r4], m3
- sub r3d, 4
- lea r0, [r0+r2*4]
- jnz .loop
- REP_RET
projects / ffmpeg / blobdiff