Unified · Split
| @@ -0,0 +1,36 @@ | |||
| 1 | /* | ||
| 2 | * Copyright (c) 2019 Paul B Mahol | ||
| 3 | * | ||
| 4 | * This file is part of FFmpeg. | ||
| 5 | * | ||
| 6 | * FFmpeg is free software; you can redistribute it and/or | ||
| 7 | * modify it under the terms of the GNU Lesser General Public | ||
| 8 | * License as published by the Free Software Foundation; either | ||
| 9 | * version 2.1 of the License, or (at your option) any later version. | ||
| 10 | * | ||
| 11 | * FFmpeg is distributed in the hope that it will be useful, | ||
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
| 14 | * Lesser General Public License for more details. | ||
| 15 | * | ||
| 16 | * You should have received a copy of the GNU Lesser General Public | ||
| 17 | * License along with FFmpeg; if not, write to the Free Software | ||
| 18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | ||
| 19 | */ | ||
| 20 | |||
| 21 | #ifndef AVFILTER_ATADENOISE_H | ||
| 22 | #define AVFILTER_ATADENOISE_H | ||
| 23 | |||
| 24 | #include <stddef.h> | ||
| 25 | #include <stdint.h> | ||
| 26 | |||
| 27 | typedef struct ATADenoiseDSPContext { | ||
| 28 | void (*filter_row)(const uint8_t *src, uint8_t *dst, | ||
| 29 | const uint8_t **srcf, | ||
| 30 | int w, int mid, int size, | ||
| 31 | int thra, int thrb); | ||
| 32 | } ATADenoiseDSPContext; | ||
| 33 | |||
| 34 | void ff_atadenoise_init_x86(ATADenoiseDSPContext *dsp, int depth); | ||
| 35 | |||
| 36 | #endif /* AVFILTER_ATADENOISE_H */ | ||
| @@ -33,6 +33,7 @@ · 32 unmodified lines | |||
| 33 | #define FF_BUFQUEUE_SIZE 129 | 33 | #define FF_BUFQUEUE_SIZE 129 |
| 34 | #include "bufferqueue.h" | 34 | #include "bufferqueue.h" |
| 35 | 35 | ||
| 36 | #include "atadenoise.h" | ||
| 36 | #include "formats.h" | 37 | #include "formats.h" |
| 37 | #include "internal.h" | 38 | #include "internal.h" |
| 38 | #include "video.h" | 39 | #include "video.h" |
| @@ -57,10 +58,8 @@ typedef struct ATADenoiseContext { · 18 unmodified lines | |||
| 57 | int available; | 58 | int available; |
| 58 | 59 | ||
| 59 | int (*filter_slice)(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs); | 60 | int (*filter_slice)(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs); |
| 60 | void (*filter_row)(const uint8_t *src, uint8_t *dst, | 61 | |
| 61 | const uint8_t *srcf[SIZE], | 62 | ATADenoiseDSPContext dsp; |
| 62 | int w, int mid, int size, | ||
| 63 | int thra, int thrb); | ||
| 64 | } ATADenoiseContext; | 63 | } ATADenoiseContext; |
| 65 | 64 | ||
| 66 | #define OFFSET(x) offsetof(ATADenoiseContext, x) | 65 | #define OFFSET(x) offsetof(ATADenoiseContext, x) |
| @@ -209,7 +208,7 @@ static int filter_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) · 142 unmodified lines | |||
| 209 | srcf[i] = data[i] + slice_start * linesize[i]; | 208 | srcf[i] = data[i] + slice_start * linesize[i]; |
| 210 | 209 | ||
| 211 | for (y = slice_start; y < slice_end; y++) { | 210 | for (y = slice_start; y < slice_end; y++) { |
| 212 | s->filter_row(src, dst, srcf, w, mid, size, thra, thrb); | 211 | s->dsp.filter_row(src, dst, srcf, w, mid, size, thra, thrb); |
| 213 | 212 | ||
| 214 | dst += out->linesize[p]; | 213 | dst += out->linesize[p]; |
| 215 | src += in->linesize[p]; | 214 | src += in->linesize[p]; |
| @@ -239,9 +238,9 @@ static int config_input(AVFilterLink *inlink) · 23 unmodified lines | |||
| 239 | depth = desc->comp[0].depth; | 238 | depth = desc->comp[0].depth; |
| 240 | s->filter_slice = filter_slice; | 239 | s->filter_slice = filter_slice; |
| 241 | if (depth == 8) | 240 | if (depth == 8) |
| 242 | s->filter_row = filter_row8; | 241 | s->dsp.filter_row = filter_row8; |
| 243 | else | 242 | else |
| 244 | s->filter_row = filter_row16; | 243 | s->dsp.filter_row = filter_row16; |
| 245 | 244 | ||
| 246 | s->thra[0] = s->fthra[0] * (1 << depth) - 1; | 245 | s->thra[0] = s->fthra[0] * (1 << depth) - 1; |
| 247 | s->thra[1] = s->fthra[1] * (1 << depth) - 1; | 246 | s->thra[1] = s->fthra[1] * (1 << depth) - 1; |
| @@ -250,6 +249,9 @@ static int config_input(AVFilterLink *inlink) · 2 unmodified lines | |||
| 250 | s->thrb[1] = s->fthrb[1] * (1 << depth) - 1; | 249 | s->thrb[1] = s->fthrb[1] * (1 << depth) - 1; |
| 251 | s->thrb[2] = s->fthrb[2] * (1 << depth) - 1; | 250 | s->thrb[2] = s->fthrb[2] * (1 << depth) - 1; |
| 252 | 251 | ||
| 252 | if (ARCH_X86) | ||
| 253 | ff_atadenoise_init_x86(&s->dsp, depth); | ||
| 254 | |||
| 253 | return 0; | 255 | return 0; |
| 254 | } | 256 | } |
| 255 | 257 | ||
| @@ -2,6 +2,7 @@ OBJS-$(CONFIG_SCENE_SAD) += x86/scene_sad_init.o · 1 unmodified line | |||
| 2 | 2 | ||
| 3 | OBJS-$(CONFIG_AFIR_FILTER) += x86/af_afir_init.o | 3 | OBJS-$(CONFIG_AFIR_FILTER) += x86/af_afir_init.o |
| 4 | OBJS-$(CONFIG_ANLMDN_FILTER) += x86/af_anlmdn_init.o | 4 | OBJS-$(CONFIG_ANLMDN_FILTER) += x86/af_anlmdn_init.o |
| 5 | OBJS-$(CONFIG_ATADENOISE_FILTER) += x86/vf_atadenoise_init.o | ||
| 5 | OBJS-$(CONFIG_BLEND_FILTER) += x86/vf_blend_init.o | 6 | OBJS-$(CONFIG_BLEND_FILTER) += x86/vf_blend_init.o |
| 6 | OBJS-$(CONFIG_BWDIF_FILTER) += x86/vf_bwdif_init.o | 7 | OBJS-$(CONFIG_BWDIF_FILTER) += x86/vf_bwdif_init.o |
| 7 | OBJS-$(CONFIG_COLORSPACE_FILTER) += x86/colorspacedsp_init.o | 8 | OBJS-$(CONFIG_COLORSPACE_FILTER) += x86/colorspacedsp_init.o |
| @@ -39,6 +40,7 @@ X86ASM-OBJS-$(CONFIG_SCENE_SAD) += x86/scene_sad.o · 31 unmodified lines | |||
| 39 | 40 | ||
| 40 | X86ASM-OBJS-$(CONFIG_AFIR_FILTER) += x86/af_afir.o | 41 | X86ASM-OBJS-$(CONFIG_AFIR_FILTER) += x86/af_afir.o |
| 41 | X86ASM-OBJS-$(CONFIG_ANLMDN_FILTER) += x86/af_anlmdn.o | 42 | X86ASM-OBJS-$(CONFIG_ANLMDN_FILTER) += x86/af_anlmdn.o |
| 43 | X86ASM-OBJS-$(CONFIG_ATADENOISE_FILTER) += x86/vf_atadenoise.o | ||
| 42 | X86ASM-OBJS-$(CONFIG_BLEND_FILTER) += x86/vf_blend.o | 44 | X86ASM-OBJS-$(CONFIG_BLEND_FILTER) += x86/vf_blend.o |
| 43 | X86ASM-OBJS-$(CONFIG_BWDIF_FILTER) += x86/vf_bwdif.o | 45 | X86ASM-OBJS-$(CONFIG_BWDIF_FILTER) += x86/vf_bwdif.o |
| 44 | X86ASM-OBJS-$(CONFIG_COLORSPACE_FILTER) += x86/colorspacedsp.o | 46 | X86ASM-OBJS-$(CONFIG_COLORSPACE_FILTER) += x86/colorspacedsp.o |
| @@ -0,0 +1,154 @@ | |||
| 1 | ;***************************************************************************** | ||
| 2 | ;* x86-optimized functions for blend filter | ||
| 3 | ;* | ||
| 4 | ;* Copyright (C) 2019 Paul B Mahol | ||
| 5 | ;* | ||
| 6 | ;* This file is part of FFmpeg. | ||
| 7 | ;* | ||
| 8 | ;* FFmpeg is free software; you can redistribute it and/or | ||
| 9 | ;* modify it under the terms of the GNU Lesser General Public | ||
| 10 | ;* License as published by the Free Software Foundation; either | ||
| 11 | ;* version 2.1 of the License, or (at your option) any later version. | ||
| 12 | ;* | ||
| 13 | ;* FFmpeg is distributed in the hope that it will be useful, | ||
| 14 | ;* but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 15 | ;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
| 16 | ;* Lesser General Public License for more details. | ||
| 17 | ;* | ||
| 18 | ;* You should have received a copy of the GNU Lesser General Public | ||
| 19 | ;* License along with FFmpeg; if not, write to the Free Software | ||
| 20 | ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | ||
| 21 | ;****************************************************************************** | ||
| 22 | |||
| 23 | %if ARCH_X86_64 | ||
| 24 | |||
| 25 | %include "libavutil/x86/x86util.asm" | ||
| 26 | |||
| 27 | SECTION_RODATA | ||
| 28 | pw_one: times 8 dw 1 | ||
| 29 | pw_ones: times 8 dw 65535 | ||
| 30 | |||
| 31 | SECTION .text | ||
| 32 | |||
| 33 | ;------------------------------------------------------------------------------ | ||
| 34 | ; void ff_filter_row(const uint8_t *src, uint8_t *dst, | ||
| 35 | ; const uint8_t **srcf, | ||
| 36 | ; int w, int mid, int size, | ||
| 37 | ; int thra, int thrb) | ||
| 38 | ;------------------------------------------------------------------------------ | ||
| 39 | |||
| 40 | INIT_XMM sse4 | ||
| 41 | cglobal atadenoise_filter_row8, 8,10,13, src, dst, srcf, w, mid, size, i, j, srcfx, x | ||
| 42 | movsxdifnidn wq, wd | ||
| 43 | movsxdifnidn midq, midd | ||
| 44 | movsxdifnidn sizeq, sized | ||
| 45 | add srcq, wq | ||
| 46 | add dstq, wq | ||
| 47 | mov xq, wq | ||
| 48 | dec sizeq | ||
| 49 | neg xq | ||
| 50 | movd m4, r6m | ||
| 51 | SPLATW m4, m4 | ||
| 52 | movd m5, r7m | ||
| 53 | SPLATW m5, m5 | ||
| 54 | pxor m2, m2 | ||
| 55 | mova m10, [pw_ones] | ||
| 56 | |||
| 57 | .loop: | ||
| 58 | mov iq, midq | ||
| 59 | mov jq, midq | ||
| 60 | pxor m3, m3 | ||
| 61 | pxor m11, m11 | ||
| 62 | movu m0, [srcq + xq] | ||
| 63 | punpcklbw m0, m2 | ||
| 64 | mova m7, m0 | ||
| 65 | mova m8, [pw_one] | ||
| 66 | mova m12, [pw_ones] | ||
| 67 | |||
| 68 | .loop0: | ||
| 69 | inc iq | ||
| 70 | dec jq | ||
| 71 | |||
| 72 | mov srcfxq, [srcfq + jq * 8] | ||
| 73 | add srcfxq, wq | ||
| 74 | |||
| 75 | movu m1, [srcfxq + xq] | ||
| 76 | punpcklbw m1, m2 | ||
| 77 | mova m9, m1 | ||
| 78 | psubw m1, m0 | ||
| 79 | pabsw m1, m1 | ||
| 80 | paddw m11, m1 | ||
| 81 | pcmpgtw m1, m4 | ||
| 82 | mova m6, m11 | ||
| 83 | pcmpgtw m6, m5 | ||
| 84 | por m6, m1 | ||
| 85 | pxor m6, m10 | ||
| 86 | pand m12, m6 | ||
| 87 | pand m9, m12 | ||
| 88 | paddw m7, m9 | ||
| 89 | mova m6, m12 | ||
| 90 | psrlw m6, 15 | ||
| 91 | paddw m8, m6 | ||
| 92 | |||
| 93 | mov srcfxq, [srcfq + iq * 8] | ||
| 94 | add srcfxq, wq | ||
| 95 | |||
| 96 | movu m1, [srcfxq + xq] | ||
| 97 | punpcklbw m1, m2 | ||
| 98 | mova m9, m1 | ||
| 99 | psubw m1, m0 | ||
| 100 | pabsw m1, m1 | ||
| 101 | paddw m3, m1 | ||
| 102 | pcmpgtw m1, m4 | ||
| 103 | mova m6, m3 | ||
| 104 | pcmpgtw m6, m5 | ||
| 105 | por m6, m1 | ||
| 106 | pxor m6, m10 | ||
| 107 | pand m12, m6 | ||
| 108 | pand m9, m12 | ||
| 109 | paddw m7, m9 | ||
| 110 | mova m6, m12 | ||
| 111 | psrlw m6, 15 | ||
| 112 | paddw m8, m6 | ||
| 113 | |||
| 114 | ptest m12, m12 | ||
| 115 | jz .finish | ||
| 116 | |||
| 117 | cmp iq, sizeq | ||
| 118 | jl .loop0 | ||
| 119 | |||
| 120 | .finish: | ||
| 121 | mova m9, m8 | ||
| 122 | psrlw m9, 1 | ||
| 123 | paddw m7, m9 | ||
| 124 | |||
| 125 | mova m1, m7 | ||
| 126 | mova m6, m8 | ||
| 127 | |||
| 128 | punpcklwd m7, m2 | ||
| 129 | punpcklwd m8, m2 | ||
| 130 | cvtdq2ps m7, m7 | ||
| 131 | cvtdq2ps m8, m8 | ||
| 132 | divps m7, m8 | ||
| 133 | cvttps2dq m7, m7 | ||
| 134 | packssdw m7, m7 | ||
| 135 | packuswb m7, m7 | ||
| 136 | |||
| 137 | movd [dstq + xq], m7 | ||
| 138 | |||
| 139 | punpckhwd m1, m2 | ||
| 140 | punpckhwd m6, m2 | ||
| 141 | cvtdq2ps m1, m1 | ||
| 142 | cvtdq2ps m6, m6 | ||
| 143 | divps m1, m6 | ||
| 144 | cvttps2dq m1, m1 | ||
| 145 | packssdw m1, m1 | ||
| 146 | packuswb m1, m1 | ||
| 147 | |||
| 148 | movd [dstq + xq + 4], m1 | ||
| 149 | |||
| 150 | add xq, mmsize/2 | ||
| 151 | jl .loop | ||
| 152 | RET | ||
| 153 | |||
| 154 | %endif | ||
| @@ -0,0 +1,40 @@ | |||
| 1 | /* | ||
| 2 | * Copyright (C) 2019 Paul B Mahol | ||
| 3 | * | ||
| 4 | * This file is part of FFmpeg. | ||
| 5 | * | ||
| 6 | * FFmpeg is free software; you can redistribute it and/or | ||
| 7 | * modify it under the terms of the GNU Lesser General Public | ||
| 8 | * License as published by the Free Software Foundation; either | ||
| 9 | * version 2.1 of the License, or (at your option) any later version. | ||
| 10 | * | ||
| 11 | * FFmpeg is distributed in the hope that it will be useful, | ||
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
| 14 | * Lesser General Public License for more details. | ||
| 15 | * | ||
| 16 | * You should have received a copy of the GNU Lesser General Public | ||
| 17 | * License along with FFmpeg; if not, write to the Free Software | ||
| 18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | ||
| 19 | */ | ||
| 20 | |||
| 21 | #include "libavutil/attributes.h" | ||
| 22 | #include "libavutil/cpu.h" | ||
| 23 | #include "libavutil/mem.h" | ||
| 24 | #include "libavutil/x86/asm.h" | ||
| 25 | #include "libavutil/x86/cpu.h" | ||
| 26 | #include "libavfilter/atadenoise.h" | ||
| 27 | |||
| 28 | void ff_atadenoise_filter_row8_sse4(const uint8_t *src, uint8_t *dst, | ||
| 29 | const uint8_t **srcf, | ||
| 30 | int w, int mid, int size, | ||
| 31 | int thra, int thrb); | ||
| 32 | |||
| 33 | av_cold void ff_atadenoise_init_x86(ATADenoiseDSPContext *dsp, int depth) | ||
| 34 | { | ||
| 35 | int cpu_flags = av_get_cpu_flags(); | ||
| 36 | |||
| 37 | if (ARCH_X86_64 && EXTERNAL_SSE4(cpu_flags) && depth <= 8) { | ||
| 38 | dsp->filter_row = ff_atadenoise_filter_row8_sse4; | ||
| 39 | } | ||
| 40 | } | ||