...
This commit is contained in:
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,301 @@
|
||||
;******************************************************************************
|
||||
;* SIMD optimized SAO functions for HEVC/VVC 8bit decoding
|
||||
;*
|
||||
;* Copyright (c) 2013 Pierre-Edouard LEPERE
|
||||
;* Copyright (c) 2014 James Almer
|
||||
;*
|
||||
;* This file is part of FFmpeg.
|
||||
;*
|
||||
;* FFmpeg is free software; you can redistribute it and/or
|
||||
;* modify it under the terms of the GNU Lesser General Public
|
||||
;* License as published by the Free Software Foundation; either
|
||||
;* version 2.1 of the License, or (at your option) any later version.
|
||||
;*
|
||||
;* FFmpeg is distributed in the hope that it will be useful,
|
||||
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
;* Lesser General Public License for more details.
|
||||
;*
|
||||
;* You should have received a copy of the GNU Lesser General Public
|
||||
;* License along with FFmpeg; if not, write to the Free Software
|
||||
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
;******************************************************************************
|
||||
|
||||
%include "libavutil/x86/x86util.asm"
|
||||
|
||||
SECTION_RODATA 32
|
||||
|
||||
pb_edge_shuffle: times 2 db 1, 2, 0, 3, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1
|
||||
pb_eo: db -1, 0, 1, 0, 0, -1, 0, 1, -1, -1, 1, 1, 1, -1, -1, 1
|
||||
cextern pb_1
|
||||
cextern pb_2
|
||||
|
||||
SECTION .text
|
||||
|
||||
;******************************************************************************
|
||||
;SAO Band Filter
|
||||
;******************************************************************************
|
||||
|
||||
%macro H2656_SAO_BAND_FILTER_INIT 0
|
||||
and leftq, 31
|
||||
movd xm0, leftd
|
||||
add leftq, 1
|
||||
and leftq, 31
|
||||
movd xm1, leftd
|
||||
add leftq, 1
|
||||
and leftq, 31
|
||||
movd xm2, leftd
|
||||
add leftq, 1
|
||||
and leftq, 31
|
||||
movd xm3, leftd
|
||||
|
||||
SPLATW m0, xm0
|
||||
SPLATW m1, xm1
|
||||
SPLATW m2, xm2
|
||||
SPLATW m3, xm3
|
||||
%if mmsize > 16
|
||||
SPLATW m4, [offsetq + 2]
|
||||
SPLATW m5, [offsetq + 4]
|
||||
SPLATW m6, [offsetq + 6]
|
||||
SPLATW m7, [offsetq + 8]
|
||||
%else
|
||||
movq m7, [offsetq + 2]
|
||||
SPLATW m4, m7, 0
|
||||
SPLATW m5, m7, 1
|
||||
SPLATW m6, m7, 2
|
||||
SPLATW m7, m7, 3
|
||||
%endif
|
||||
|
||||
%if ARCH_X86_64
|
||||
pxor m14, m14
|
||||
|
||||
%else ; ARCH_X86_32
|
||||
mova [rsp+mmsize*0], m0
|
||||
mova [rsp+mmsize*1], m1
|
||||
mova [rsp+mmsize*2], m2
|
||||
mova [rsp+mmsize*3], m3
|
||||
mova [rsp+mmsize*4], m4
|
||||
mova [rsp+mmsize*5], m5
|
||||
mova [rsp+mmsize*6], m6
|
||||
pxor m0, m0
|
||||
%assign MMSIZE mmsize
|
||||
%define m14 m0
|
||||
%define m13 m1
|
||||
%define m9 m2
|
||||
%define m8 m3
|
||||
%endif ; ARCH
|
||||
DEFINE_ARGS dst, src, dststride, srcstride, offset, height
|
||||
mov heightd, r7m
|
||||
%endmacro
|
||||
|
||||
%macro H2656_SAO_BAND_FILTER_COMPUTE 2
|
||||
psraw %1, %2, 3
|
||||
%if ARCH_X86_64
|
||||
pcmpeqw m10, %1, m0
|
||||
pcmpeqw m11, %1, m1
|
||||
pcmpeqw m12, %1, m2
|
||||
pcmpeqw %1, m3
|
||||
pand m10, m4
|
||||
pand m11, m5
|
||||
pand m12, m6
|
||||
pand %1, m7
|
||||
por m10, m11
|
||||
por m12, %1
|
||||
por m10, m12
|
||||
paddw %2, m10
|
||||
%else ; ARCH_X86_32
|
||||
pcmpeqw m4, %1, [rsp+MMSIZE*0]
|
||||
pcmpeqw m5, %1, [rsp+MMSIZE*1]
|
||||
pcmpeqw m6, %1, [rsp+MMSIZE*2]
|
||||
pcmpeqw %1, [rsp+MMSIZE*3]
|
||||
pand m4, [rsp+MMSIZE*4]
|
||||
pand m5, [rsp+MMSIZE*5]
|
||||
pand m6, [rsp+MMSIZE*6]
|
||||
pand %1, m7
|
||||
por m4, m5
|
||||
por m6, %1
|
||||
por m4, m6
|
||||
paddw %2, m4
|
||||
%endif ; ARCH
|
||||
%endmacro
|
||||
|
||||
;void ff_{hevc, vvc}_sao_band_filter_<width>_8_<opt>(uint8_t *_dst, const uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src,
|
||||
; int16_t *sao_offset_val, int sao_left_class, int width, int height);
|
||||
%macro H2656_SAO_BAND_FILTER 3
|
||||
cglobal %1_sao_band_filter_%2_8, 6, 6, 15, 7*mmsize*ARCH_X86_32, dst, src, dststride, srcstride, offset, left
|
||||
H2656_SAO_BAND_FILTER_INIT
|
||||
|
||||
align 16
|
||||
.loop:
|
||||
%if %2 == 8
|
||||
movq m8, [srcq]
|
||||
punpcklbw m8, m14
|
||||
H2656_SAO_BAND_FILTER_COMPUTE m9, m8
|
||||
packuswb m8, m14
|
||||
movq [dstq], m8
|
||||
%endif ; %2 == 8
|
||||
|
||||
%assign i 0
|
||||
%rep %3
|
||||
mova m13, [srcq + i]
|
||||
punpcklbw m8, m13, m14
|
||||
H2656_SAO_BAND_FILTER_COMPUTE m9, m8
|
||||
punpckhbw m13, m14
|
||||
H2656_SAO_BAND_FILTER_COMPUTE m9, m13
|
||||
packuswb m8, m13
|
||||
mova [dstq + i], m8
|
||||
%assign i i+mmsize
|
||||
%endrep
|
||||
|
||||
%if %2 == 48 || %2 == 80 || %2 == 112
|
||||
INIT_XMM cpuname
|
||||
|
||||
mova m13, [srcq + i]
|
||||
punpcklbw m8, m13, m14
|
||||
H2656_SAO_BAND_FILTER_COMPUTE m9, m8
|
||||
punpckhbw m13, m14
|
||||
H2656_SAO_BAND_FILTER_COMPUTE m9, m13
|
||||
packuswb m8, m13
|
||||
mova [dstq + i], m8
|
||||
%if cpuflag(avx2)
|
||||
INIT_YMM cpuname
|
||||
%endif
|
||||
%endif ; %2 == 48 || %2 == 80 || %2 == 112
|
||||
|
||||
add dstq, dststrideq ; dst += dststride
|
||||
add srcq, srcstrideq ; src += srcstride
|
||||
dec heightd ; cmp height
|
||||
jnz .loop ; height loop
|
||||
RET
|
||||
%endmacro
|
||||
|
||||
;******************************************************************************
|
||||
;SAO Edge Filter
|
||||
;******************************************************************************
|
||||
|
||||
%define PADDING_SIZE 64 ; AV_INPUT_BUFFER_PADDING_SIZE
|
||||
%define EDGE_SRCSTRIDE 2 * MAX_PB_SIZE + PADDING_SIZE
|
||||
|
||||
%macro H2656_SAO_EDGE_FILTER_INIT 0
|
||||
%if WIN64
|
||||
movsxd eoq, dword eom
|
||||
%elif ARCH_X86_64
|
||||
movsxd eoq, eod
|
||||
%else
|
||||
mov eoq, r4m
|
||||
%endif
|
||||
lea tmp2q, [pb_eo]
|
||||
movsx a_strideq, byte [tmp2q+eoq*4+1]
|
||||
movsx b_strideq, byte [tmp2q+eoq*4+3]
|
||||
imul a_strideq, EDGE_SRCSTRIDE
|
||||
imul b_strideq, EDGE_SRCSTRIDE
|
||||
movsx tmpq, byte [tmp2q+eoq*4]
|
||||
add a_strideq, tmpq
|
||||
movsx tmpq, byte [tmp2q+eoq*4+2]
|
||||
add b_strideq, tmpq
|
||||
%endmacro
|
||||
|
||||
%macro H2656_SAO_EDGE_FILTER_COMPUTE 1
|
||||
pminub m4, m1, m2
|
||||
pminub m5, m1, m3
|
||||
pcmpeqb m2, m4
|
||||
pcmpeqb m3, m5
|
||||
pcmpeqb m4, m1
|
||||
pcmpeqb m5, m1
|
||||
psubb m4, m2
|
||||
psubb m5, m3
|
||||
paddb m4, m6
|
||||
paddb m4, m5
|
||||
|
||||
pshufb m2, m0, m4
|
||||
%if %1 > 8
|
||||
punpckhbw m5, m7, m1
|
||||
punpckhbw m4, m2, m7
|
||||
punpcklbw m3, m7, m1
|
||||
punpcklbw m2, m7
|
||||
pmaddubsw m5, m4
|
||||
pmaddubsw m3, m2
|
||||
packuswb m3, m5
|
||||
%else
|
||||
punpcklbw m3, m7, m1
|
||||
punpcklbw m2, m7
|
||||
pmaddubsw m3, m2
|
||||
packuswb m3, m3
|
||||
%endif
|
||||
%endmacro
|
||||
|
||||
;void ff_{hevc, vvc}_sao_edge_filter_<width>_8_<opt>(uint8_t *_dst, uint8_t *_src, ptrdiff_t stride_dst, int16_t *sao_offset_val,
|
||||
; int eo, int width, int height);
|
||||
%macro H2656_SAO_EDGE_FILTER 3-4
|
||||
%if ARCH_X86_64
|
||||
cglobal %1_sao_edge_filter_%2_8, 4, 9, 8, dst, src, dststride, offset, eo, a_stride, b_stride, height, tmp
|
||||
%define tmp2q heightq
|
||||
H2656_SAO_EDGE_FILTER_INIT
|
||||
mov heightd, r6m
|
||||
|
||||
%else ; ARCH_X86_32
|
||||
cglobal %1_sao_edge_filter_%2_8, 1, 6, 8, dst, src, dststride, a_stride, b_stride, height
|
||||
%define eoq srcq
|
||||
%define tmpq heightq
|
||||
%define tmp2q dststrideq
|
||||
%define offsetq heightq
|
||||
H2656_SAO_EDGE_FILTER_INIT
|
||||
mov srcq, srcm
|
||||
mov offsetq, r3m
|
||||
mov dststrideq, dststridem
|
||||
%endif ; ARCH
|
||||
|
||||
%if mmsize > 16
|
||||
vbroadcasti128 m0, [offsetq]
|
||||
%else
|
||||
movu m0, [offsetq]
|
||||
%endif
|
||||
mova m1, [pb_edge_shuffle]
|
||||
packsswb m0, m0
|
||||
mova m7, [pb_1]
|
||||
pshufb m0, m1
|
||||
mova m6, [pb_2]
|
||||
%if ARCH_X86_32
|
||||
mov heightd, r6m
|
||||
%endif
|
||||
|
||||
align 16
|
||||
.loop:
|
||||
|
||||
%if %2 == 8
|
||||
movq m1, [srcq]
|
||||
movq m2, [srcq + a_strideq]
|
||||
movq m3, [srcq + b_strideq]
|
||||
H2656_SAO_EDGE_FILTER_COMPUTE %2
|
||||
movq [dstq], m3
|
||||
%endif
|
||||
|
||||
%assign i 0
|
||||
%rep %3
|
||||
mova m1, [srcq + i]
|
||||
movu m2, [srcq + a_strideq + i]
|
||||
movu m3, [srcq + b_strideq + i]
|
||||
H2656_SAO_EDGE_FILTER_COMPUTE %2
|
||||
mov%4 [dstq + i], m3
|
||||
%assign i i+mmsize
|
||||
%endrep
|
||||
|
||||
%if %2 == 48 || %2 == 80 || %2 == 112
|
||||
INIT_XMM cpuname
|
||||
|
||||
mova m1, [srcq + i]
|
||||
movu m2, [srcq + a_strideq + i]
|
||||
movu m3, [srcq + b_strideq + i]
|
||||
H2656_SAO_EDGE_FILTER_COMPUTE %2
|
||||
mova [dstq + i], m3
|
||||
%if cpuflag(avx2)
|
||||
INIT_YMM cpuname
|
||||
%endif
|
||||
%endif ; %2 == 48 || %2 == 80 || %2 == 112
|
||||
|
||||
add dstq, dststrideq
|
||||
add srcq, EDGE_SRCSTRIDE
|
||||
dec heightd
|
||||
jg .loop
|
||||
RET
|
||||
%endmacro
|
||||
@@ -0,0 +1,301 @@
|
||||
;******************************************************************************
|
||||
;* SIMD optimized SAO functions for HEVC/VVC 10/12bit decoding
|
||||
;*
|
||||
;* Copyright (c) 2013 Pierre-Edouard LEPERE
|
||||
;* Copyright (c) 2014 James Almer
|
||||
;*
|
||||
;* This file is part of FFmpeg.
|
||||
;*
|
||||
;* FFmpeg is free software; you can redistribute it and/or
|
||||
;* modify it under the terms of the GNU Lesser General Public
|
||||
;* License as published by the Free Software Foundation; either
|
||||
;* version 2.1 of the License, or (at your option) any later version.
|
||||
;*
|
||||
;* FFmpeg is distributed in the hope that it will be useful,
|
||||
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
;* Lesser General Public License for more details.
|
||||
;*
|
||||
;* You should have received a copy of the GNU Lesser General Public
|
||||
;* License along with FFmpeg; if not, write to the Free Software
|
||||
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
;******************************************************************************
|
||||
|
||||
%include "libavutil/x86/x86util.asm"
|
||||
|
||||
SECTION_RODATA 32
|
||||
|
||||
pw_m2: times 16 dw -2
|
||||
pw_mask10: times 16 dw 0x03FF
|
||||
pw_mask12: times 16 dw 0x0FFF
|
||||
pb_eo: db -1, 0, 1, 0, 0, -1, 0, 1, -1, -1, 1, 1, 1, -1, -1, 1
|
||||
cextern pw_m1
|
||||
cextern pw_1
|
||||
cextern pw_2
|
||||
|
||||
SECTION .text
|
||||
|
||||
;******************************************************************************
|
||||
;SAO Band Filter
|
||||
;******************************************************************************
|
||||
|
||||
%macro H2656_SAO_BAND_FILTER_INIT 1
|
||||
and leftq, 31
|
||||
movd xm0, leftd
|
||||
add leftq, 1
|
||||
and leftq, 31
|
||||
movd xm1, leftd
|
||||
add leftq, 1
|
||||
and leftq, 31
|
||||
movd xm2, leftd
|
||||
add leftq, 1
|
||||
and leftq, 31
|
||||
movd xm3, leftd
|
||||
|
||||
SPLATW m0, xm0
|
||||
SPLATW m1, xm1
|
||||
SPLATW m2, xm2
|
||||
SPLATW m3, xm3
|
||||
%if mmsize > 16
|
||||
SPLATW m4, [offsetq + 2]
|
||||
SPLATW m5, [offsetq + 4]
|
||||
SPLATW m6, [offsetq + 6]
|
||||
SPLATW m7, [offsetq + 8]
|
||||
%else
|
||||
movq m7, [offsetq + 2]
|
||||
SPLATW m4, m7, 0
|
||||
SPLATW m5, m7, 1
|
||||
SPLATW m6, m7, 2
|
||||
SPLATW m7, m7, 3
|
||||
%endif
|
||||
|
||||
%if ARCH_X86_64
|
||||
mova m13, [pw_mask %+ %1]
|
||||
pxor m14, m14
|
||||
|
||||
%else ; ARCH_X86_32
|
||||
mova [rsp+mmsize*0], m0
|
||||
mova [rsp+mmsize*1], m1
|
||||
mova [rsp+mmsize*2], m2
|
||||
mova [rsp+mmsize*3], m3
|
||||
mova [rsp+mmsize*4], m4
|
||||
mova [rsp+mmsize*5], m5
|
||||
mova [rsp+mmsize*6], m6
|
||||
mova m1, [pw_mask %+ %1]
|
||||
pxor m0, m0
|
||||
%define m14 m0
|
||||
%define m13 m1
|
||||
%define m9 m2
|
||||
%define m8 m3
|
||||
%endif ; ARCH
|
||||
DEFINE_ARGS dst, src, dststride, srcstride, offset, height
|
||||
mov heightd, r7m
|
||||
%endmacro
|
||||
|
||||
;void ff_{hevc, vvc}_sao_band_filter_<width>_<depth>_<opt>(uint8_t *_dst, const uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src,
|
||||
; int16_t *sao_offset_val, int sao_left_class, int width, int height);
|
||||
%macro H2656_SAO_BAND_FILTER 4
|
||||
cglobal %1_sao_band_filter_%3_%2, 6, 6, 15, 7*mmsize*ARCH_X86_32, dst, src, dststride, srcstride, offset, left
|
||||
H2656_SAO_BAND_FILTER_INIT %2
|
||||
|
||||
align 16
|
||||
.loop:
|
||||
|
||||
%assign i 0
|
||||
%assign j 0
|
||||
%rep %4
|
||||
%assign k 8+(j&1)
|
||||
%assign l 9-(j&1)
|
||||
mova m %+ k, [srcq + i]
|
||||
psraw m %+ l, m %+ k, %2-5
|
||||
%if ARCH_X86_64
|
||||
pcmpeqw m10, m %+ l, m0
|
||||
pcmpeqw m11, m %+ l, m1
|
||||
pcmpeqw m12, m %+ l, m2
|
||||
pcmpeqw m %+ l, m3
|
||||
pand m10, m4
|
||||
pand m11, m5
|
||||
pand m12, m6
|
||||
pand m %+ l, m7
|
||||
por m10, m11
|
||||
por m12, m %+ l
|
||||
por m10, m12
|
||||
paddw m %+ k, m10
|
||||
%else ; ARCH_X86_32
|
||||
pcmpeqw m4, m %+ l, [rsp+mmsize*0]
|
||||
pcmpeqw m5, m %+ l, [rsp+mmsize*1]
|
||||
pcmpeqw m6, m %+ l, [rsp+mmsize*2]
|
||||
pcmpeqw m %+ l, [rsp+mmsize*3]
|
||||
pand m4, [rsp+mmsize*4]
|
||||
pand m5, [rsp+mmsize*5]
|
||||
pand m6, [rsp+mmsize*6]
|
||||
pand m %+ l, m7
|
||||
por m4, m5
|
||||
por m6, m %+ l
|
||||
por m4, m6
|
||||
paddw m %+ k, m4
|
||||
%endif ; ARCH
|
||||
CLIPW m %+ k, m14, m13
|
||||
mova [dstq + i], m %+ k
|
||||
%assign i i+mmsize
|
||||
%assign j j+1
|
||||
%endrep
|
||||
|
||||
add dstq, dststrideq
|
||||
add srcq, srcstrideq
|
||||
dec heightd
|
||||
jg .loop
|
||||
RET
|
||||
%endmacro
|
||||
|
||||
;******************************************************************************
|
||||
;SAO Edge Filter
|
||||
;******************************************************************************
|
||||
|
||||
%define PADDING_SIZE 64 ; AV_INPUT_BUFFER_PADDING_SIZE
|
||||
%define EDGE_SRCSTRIDE 2 * MAX_PB_SIZE + PADDING_SIZE
|
||||
|
||||
%macro PMINUW 4
|
||||
%if cpuflag(sse4)
|
||||
pminuw %1, %2, %3
|
||||
%else
|
||||
psubusw %4, %2, %3
|
||||
psubw %1, %2, %4
|
||||
%endif
|
||||
%endmacro
|
||||
|
||||
%macro H2656_SAO_EDGE_FILTER_INIT 0
|
||||
%if WIN64
|
||||
movsxd eoq, dword eom
|
||||
%elif ARCH_X86_64
|
||||
movsxd eoq, eod
|
||||
%else
|
||||
mov eoq, r4m
|
||||
%endif
|
||||
lea tmp2q, [pb_eo]
|
||||
movsx a_strideq, byte [tmp2q+eoq*4+1]
|
||||
movsx b_strideq, byte [tmp2q+eoq*4+3]
|
||||
imul a_strideq, EDGE_SRCSTRIDE >> 1
|
||||
imul b_strideq, EDGE_SRCSTRIDE >> 1
|
||||
movsx tmpq, byte [tmp2q+eoq*4]
|
||||
add a_strideq, tmpq
|
||||
movsx tmpq, byte [tmp2q+eoq*4+2]
|
||||
add b_strideq, tmpq
|
||||
%endmacro
|
||||
|
||||
;void ff_{hevc, vvc}_sao_edge_filter_<width>_<depth>_<opt>(uint8_t *_dst, uint8_t *_src, ptrdiff_t stride_dst, int16_t *sao_offset_val,
|
||||
; int eo, int width, int height);
|
||||
%macro H2656_SAO_EDGE_FILTER 4
|
||||
%if ARCH_X86_64
|
||||
cglobal %1_sao_edge_filter_%3_%2, 4, 9, 16, dst, src, dststride, offset, eo, a_stride, b_stride, height, tmp
|
||||
%define tmp2q heightq
|
||||
H2656_SAO_EDGE_FILTER_INIT
|
||||
mov heightd, r6m
|
||||
add a_strideq, a_strideq
|
||||
add b_strideq, b_strideq
|
||||
|
||||
%else ; ARCH_X86_32
|
||||
cglobal %1_sao_edge_filter_%3_%2, 1, 6, 8, 5*mmsize, dst, src, dststride, a_stride, b_stride, height
|
||||
%define eoq srcq
|
||||
%define tmpq heightq
|
||||
%define tmp2q dststrideq
|
||||
%define offsetq heightq
|
||||
%define m8 m1
|
||||
%define m9 m2
|
||||
%define m10 m3
|
||||
%define m11 m4
|
||||
%define m12 m5
|
||||
H2656_SAO_EDGE_FILTER_INIT
|
||||
mov srcq, srcm
|
||||
mov offsetq, r3m
|
||||
mov dststrideq, dststridem
|
||||
add a_strideq, a_strideq
|
||||
add b_strideq, b_strideq
|
||||
|
||||
%endif ; ARCH
|
||||
|
||||
%if mmsize > 16
|
||||
SPLATW m8, [offsetq+2]
|
||||
SPLATW m9, [offsetq+4]
|
||||
SPLATW m10, [offsetq+0]
|
||||
SPLATW m11, [offsetq+6]
|
||||
SPLATW m12, [offsetq+8]
|
||||
%else
|
||||
movq m10, [offsetq+0]
|
||||
movd m12, [offsetq+6]
|
||||
SPLATW m8, xm10, 1
|
||||
SPLATW m9, xm10, 2
|
||||
SPLATW m10, xm10, 0
|
||||
SPLATW m11, xm12, 0
|
||||
SPLATW m12, xm12, 1
|
||||
%endif
|
||||
pxor m0, m0
|
||||
%if ARCH_X86_64
|
||||
mova m13, [pw_m1]
|
||||
mova m14, [pw_1]
|
||||
mova m15, [pw_2]
|
||||
%else
|
||||
mov heightd, r6m
|
||||
mova [rsp+mmsize*0], m8
|
||||
mova [rsp+mmsize*1], m9
|
||||
mova [rsp+mmsize*2], m10
|
||||
mova [rsp+mmsize*3], m11
|
||||
mova [rsp+mmsize*4], m12
|
||||
%endif
|
||||
|
||||
align 16
|
||||
.loop:
|
||||
|
||||
%assign i 0
|
||||
%rep %4
|
||||
mova m1, [srcq + i]
|
||||
movu m2, [srcq+a_strideq + i]
|
||||
movu m3, [srcq+b_strideq + i]
|
||||
PMINUW m4, m1, m2, m6
|
||||
PMINUW m5, m1, m3, m7
|
||||
pcmpeqw m2, m4
|
||||
pcmpeqw m3, m5
|
||||
pcmpeqw m4, m1
|
||||
pcmpeqw m5, m1
|
||||
psubw m4, m2
|
||||
psubw m5, m3
|
||||
|
||||
paddw m4, m5
|
||||
pcmpeqw m2, m4, [pw_m2]
|
||||
%if ARCH_X86_64
|
||||
pcmpeqw m3, m4, m13
|
||||
pcmpeqw m5, m4, m0
|
||||
pcmpeqw m6, m4, m14
|
||||
pcmpeqw m7, m4, m15
|
||||
pand m2, m8
|
||||
pand m3, m9
|
||||
pand m5, m10
|
||||
pand m6, m11
|
||||
pand m7, m12
|
||||
%else
|
||||
pcmpeqw m3, m4, [pw_m1]
|
||||
pcmpeqw m5, m4, m0
|
||||
pcmpeqw m6, m4, [pw_1]
|
||||
pcmpeqw m7, m4, [pw_2]
|
||||
pand m2, [rsp+mmsize*0]
|
||||
pand m3, [rsp+mmsize*1]
|
||||
pand m5, [rsp+mmsize*2]
|
||||
pand m6, [rsp+mmsize*3]
|
||||
pand m7, [rsp+mmsize*4]
|
||||
%endif
|
||||
paddw m2, m3
|
||||
paddw m5, m6
|
||||
paddw m2, m7
|
||||
paddw m2, m1
|
||||
paddw m2, m5
|
||||
CLIPW m2, m0, [pw_mask %+ %2]
|
||||
mova [dstq + i], m2
|
||||
%assign i i+mmsize
|
||||
%endrep
|
||||
|
||||
add dstq, dststrideq
|
||||
add srcq, EDGE_SRCSTRIDE
|
||||
dec heightd
|
||||
jg .loop
|
||||
RET
|
||||
%endmacro
|
||||
@@ -0,0 +1,104 @@
|
||||
/*
|
||||
* DSP for HEVC/VVC
|
||||
*
|
||||
* Copyright (C) 2022-2024 Nuo Mi
|
||||
* Copyright (c) 2023-2024 Wu Jianhua
|
||||
*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* modify it under the terms of the GNU Lesser General Public
|
||||
* License as published by the Free Software Foundation; either
|
||||
* version 2.1 of the License, or (at your option) any later version.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
* Lesser General Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the GNU Lesser General Public
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#include "h2656dsp.h"
|
||||
|
||||
#define mc_rep_func(name, bitd, step, W, opt) \
|
||||
void ff_h2656_put_##name##W##_##bitd##_##opt(int16_t *_dst, ptrdiff_t dststride, \
|
||||
const uint8_t *_src, ptrdiff_t _srcstride, int height, const int8_t *hf, const int8_t *vf, int width) \
|
||||
{ \
|
||||
int i; \
|
||||
int16_t *dst; \
|
||||
for (i = 0; i < W; i += step) { \
|
||||
const uint8_t *src = _src + (i * ((bitd + 7) / 8)); \
|
||||
dst = _dst + i; \
|
||||
ff_h2656_put_##name##step##_##bitd##_##opt(dst, dststride, src, _srcstride, height, hf, vf, width); \
|
||||
} \
|
||||
}
|
||||
|
||||
#define mc_rep_uni_func(name, bitd, step, W, opt) \
|
||||
void ff_h2656_put_uni_##name##W##_##bitd##_##opt(uint8_t *_dst, ptrdiff_t dststride, \
|
||||
const uint8_t *_src, ptrdiff_t _srcstride, int height, const int8_t *hf, const int8_t *vf, int width) \
|
||||
{ \
|
||||
int i; \
|
||||
uint8_t *dst; \
|
||||
for (i = 0; i < W; i += step) { \
|
||||
const uint8_t *src = _src + (i * ((bitd + 7) / 8)); \
|
||||
dst = _dst + (i * ((bitd + 7) / 8)); \
|
||||
ff_h2656_put_uni_##name##step##_##bitd##_##opt(dst, dststride, src, _srcstride, \
|
||||
height, hf, vf, width); \
|
||||
} \
|
||||
}
|
||||
|
||||
#define mc_rep_funcs(name, bitd, step, W, opt) \
|
||||
mc_rep_func(name, bitd, step, W, opt) \
|
||||
mc_rep_uni_func(name, bitd, step, W, opt)
|
||||
|
||||
#define MC_REP_FUNCS_SSE4(fname) \
|
||||
mc_rep_funcs(fname, 8, 16,128, sse4) \
|
||||
mc_rep_funcs(fname, 8, 16, 64, sse4) \
|
||||
mc_rep_funcs(fname, 8, 16, 32, sse4) \
|
||||
mc_rep_funcs(fname, 10, 8,128, sse4) \
|
||||
mc_rep_funcs(fname, 10, 8, 64, sse4) \
|
||||
mc_rep_funcs(fname, 10, 8, 32, sse4) \
|
||||
mc_rep_funcs(fname, 10, 8, 16, sse4) \
|
||||
mc_rep_funcs(fname, 12, 8,128, sse4) \
|
||||
mc_rep_funcs(fname, 12, 8, 64, sse4) \
|
||||
mc_rep_funcs(fname, 12, 8, 32, sse4) \
|
||||
mc_rep_funcs(fname, 12, 8, 16, sse4) \
|
||||
|
||||
#if ARCH_X86_64 && HAVE_SSE4_EXTERNAL
|
||||
|
||||
MC_REP_FUNCS_SSE4(pixels)
|
||||
MC_REP_FUNCS_SSE4(4tap_h)
|
||||
MC_REP_FUNCS_SSE4(4tap_v)
|
||||
MC_REP_FUNCS_SSE4(4tap_hv)
|
||||
MC_REP_FUNCS_SSE4(8tap_h)
|
||||
MC_REP_FUNCS_SSE4(8tap_v)
|
||||
MC_REP_FUNCS_SSE4(8tap_hv)
|
||||
mc_rep_funcs(8tap_hv, 8, 8, 16, sse4)
|
||||
|
||||
#if HAVE_AVX2_EXTERNAL
|
||||
|
||||
#define MC_REP_FUNCS_AVX2_NO8(fname) \
|
||||
mc_rep_funcs(fname,10, 16, 32, avx2) \
|
||||
mc_rep_funcs(fname,10, 16, 64, avx2) \
|
||||
mc_rep_funcs(fname,10, 16,128, avx2) \
|
||||
mc_rep_funcs(fname,12, 16, 32, avx2) \
|
||||
mc_rep_funcs(fname,12, 16, 64, avx2) \
|
||||
mc_rep_funcs(fname,12, 16,128, avx2) \
|
||||
|
||||
#define MC_REP_FUNCS_AVX2(fname) \
|
||||
mc_rep_funcs(fname, 8, 32, 64, avx2) \
|
||||
mc_rep_funcs(fname, 8, 32,128, avx2) \
|
||||
MC_REP_FUNCS_AVX2_NO8(fname)
|
||||
|
||||
MC_REP_FUNCS_AVX2(pixels)
|
||||
MC_REP_FUNCS_AVX2(8tap_h)
|
||||
MC_REP_FUNCS_AVX2(8tap_v)
|
||||
MC_REP_FUNCS_AVX2_NO8(8tap_hv)
|
||||
MC_REP_FUNCS_AVX2(4tap_h)
|
||||
MC_REP_FUNCS_AVX2(4tap_v)
|
||||
MC_REP_FUNCS_AVX2_NO8(4tap_hv)
|
||||
#endif
|
||||
#endif
|
||||
@@ -0,0 +1,103 @@
|
||||
/*
|
||||
* DSP for HEVC/VVC
|
||||
*
|
||||
* Copyright (C) 2022-2024 Nuo Mi
|
||||
* Copyright (c) 2023-2024 Wu Jianhua
|
||||
*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* modify it under the terms of the GNU Lesser General Public
|
||||
* License as published by the Free Software Foundation; either
|
||||
* version 2.1 of the License, or (at your option) any later version.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
||||
* Lesser General Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the GNU Lesser General Public
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#ifndef AVCODEC_X86_H26X_H2656DSP_H
|
||||
#define AVCODEC_X86_H26X_H2656DSP_H
|
||||
|
||||
#include "config.h"
|
||||
#include "libavutil/x86/asm.h"
|
||||
#include "libavutil/x86/cpu.h"
|
||||
#include <stdlib.h>
|
||||
|
||||
#define H2656_PEL_PROTOTYPE(name, D, opt) \
|
||||
void ff_h2656_put_ ## name ## _ ## D ## _##opt(int16_t *dst, ptrdiff_t dststride, const uint8_t *_src, ptrdiff_t _srcstride, int height, const int8_t *hf, const int8_t *vf, int width); \
|
||||
void ff_h2656_put_uni_ ## name ## _ ## D ## _##opt(uint8_t *_dst, ptrdiff_t _dststride, const uint8_t *_src, ptrdiff_t _srcstride, int height, const int8_t *hf, const int8_t *vf, int width) \
|
||||
|
||||
#define H2656_MC_8TAP_PROTOTYPES(fname, bitd, opt) \
|
||||
H2656_PEL_PROTOTYPE(fname##4, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##6, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##8, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##12, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##16, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##32, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##64, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##128, bitd, opt)
|
||||
|
||||
H2656_MC_8TAP_PROTOTYPES(pixels , 8, sse4);
|
||||
H2656_MC_8TAP_PROTOTYPES(pixels , 10, sse4);
|
||||
H2656_MC_8TAP_PROTOTYPES(pixels , 12, sse4);
|
||||
H2656_MC_8TAP_PROTOTYPES(8tap_h , 8, sse4);
|
||||
H2656_MC_8TAP_PROTOTYPES(8tap_h , 10, sse4);
|
||||
H2656_MC_8TAP_PROTOTYPES(8tap_h , 12, sse4);
|
||||
H2656_MC_8TAP_PROTOTYPES(8tap_v , 8, sse4);
|
||||
H2656_MC_8TAP_PROTOTYPES(8tap_v , 10, sse4);
|
||||
H2656_MC_8TAP_PROTOTYPES(8tap_v , 12, sse4);
|
||||
H2656_MC_8TAP_PROTOTYPES(8tap_hv , 8, sse4);
|
||||
H2656_MC_8TAP_PROTOTYPES(8tap_hv , 10, sse4);
|
||||
H2656_MC_8TAP_PROTOTYPES(8tap_hv , 12, sse4);
|
||||
|
||||
#define H2656_MC_4TAP_PROTOTYPES(fname, bitd, opt) \
|
||||
H2656_PEL_PROTOTYPE(fname##2, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##4, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##6, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##8, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##12, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##16, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##32, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##64, bitd, opt); \
|
||||
H2656_PEL_PROTOTYPE(fname##128, bitd, opt)
|
||||
|
||||
#define H2656_MC_4TAP_PROTOTYPES_SSE4(bitd) \
|
||||
H2656_PEL_PROTOTYPE(pixels2, bitd, sse4); \
|
||||
H2656_MC_4TAP_PROTOTYPES(4tap_h, bitd, sse4); \
|
||||
H2656_MC_4TAP_PROTOTYPES(4tap_v, bitd, sse4); \
|
||||
H2656_MC_4TAP_PROTOTYPES(4tap_hv, bitd, sse4); \
|
||||
|
||||
H2656_MC_4TAP_PROTOTYPES_SSE4(8)
|
||||
H2656_MC_4TAP_PROTOTYPES_SSE4(10)
|
||||
H2656_MC_4TAP_PROTOTYPES_SSE4(12)
|
||||
|
||||
#define H2656_MC_8TAP_PROTOTYPES_AVX2(fname) \
|
||||
H2656_PEL_PROTOTYPE(fname##32 , 8, avx2); \
|
||||
H2656_PEL_PROTOTYPE(fname##64 , 8, avx2); \
|
||||
H2656_PEL_PROTOTYPE(fname##128, 8, avx2); \
|
||||
H2656_PEL_PROTOTYPE(fname##16 ,10, avx2); \
|
||||
H2656_PEL_PROTOTYPE(fname##32 ,10, avx2); \
|
||||
H2656_PEL_PROTOTYPE(fname##64 ,10, avx2); \
|
||||
H2656_PEL_PROTOTYPE(fname##128,10, avx2); \
|
||||
H2656_PEL_PROTOTYPE(fname##16 ,12, avx2); \
|
||||
H2656_PEL_PROTOTYPE(fname##32 ,12, avx2); \
|
||||
H2656_PEL_PROTOTYPE(fname##64 ,12, avx2); \
|
||||
H2656_PEL_PROTOTYPE(fname##128,12, avx2) \
|
||||
|
||||
H2656_MC_8TAP_PROTOTYPES_AVX2(pixels);
|
||||
H2656_MC_8TAP_PROTOTYPES_AVX2(8tap_h);
|
||||
H2656_MC_8TAP_PROTOTYPES_AVX2(8tap_v);
|
||||
H2656_MC_8TAP_PROTOTYPES_AVX2(8tap_hv);
|
||||
H2656_PEL_PROTOTYPE(8tap_hv16, 8, avx2);
|
||||
|
||||
H2656_MC_8TAP_PROTOTYPES_AVX2(4tap_h);
|
||||
H2656_MC_8TAP_PROTOTYPES_AVX2(4tap_v);
|
||||
H2656_MC_8TAP_PROTOTYPES_AVX2(4tap_hv);
|
||||
|
||||
#endif
|
||||
Reference in New Issue
Block a user