av1dec.c (53330B)
1 /* 2 * AV1 video decoder 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 "config_components.h" 22 23 #include "libavutil/hdr_dynamic_metadata.h" 24 #include "libavutil/film_grain_params.h" 25 #include "libavutil/mastering_display_metadata.h" 26 #include "libavutil/mem.h" 27 #include "libavutil/pixdesc.h" 28 #include "libavutil/opt.h" 29 #include "avcodec.h" 30 #include "av1_parse.h" 31 #include "av1dec.h" 32 #include "atsc_a53.h" 33 #include "bytestream.h" 34 #include "codec_internal.h" 35 #include "decode.h" 36 #include "hwaccel_internal.h" 37 #include "internal.h" 38 #include "itut35.h" 39 #include "hwconfig.h" 40 #include "profiles.h" 41 #include "progressframe.h" 42 #include "libavutil/refstruct.h" 43 44 /**< same with Div_Lut defined in spec 7.11.3.7 */ 45 static const uint16_t div_lut[AV1_DIV_LUT_NUM] = { 46 16384, 16320, 16257, 16194, 16132, 16070, 16009, 15948, 15888, 15828, 15768, 47 15709, 15650, 15592, 15534, 15477, 15420, 15364, 15308, 15252, 15197, 15142, 48 15087, 15033, 14980, 14926, 14873, 14821, 14769, 14717, 14665, 14614, 14564, 49 14513, 14463, 14413, 14364, 14315, 14266, 14218, 14170, 14122, 14075, 14028, 50 13981, 13935, 13888, 13843, 13797, 13752, 13707, 13662, 13618, 13574, 13530, 51 13487, 13443, 13400, 13358, 13315, 13273, 13231, 13190, 13148, 13107, 13066, 52 13026, 12985, 12945, 12906, 12866, 12827, 12788, 12749, 12710, 12672, 12633, 53 12596, 12558, 12520, 12483, 12446, 12409, 12373, 12336, 12300, 12264, 12228, 54 12193, 12157, 12122, 12087, 12053, 12018, 11984, 11950, 11916, 11882, 11848, 55 11815, 11782, 11749, 11716, 11683, 11651, 11619, 11586, 11555, 11523, 11491, 56 11460, 11429, 11398, 11367, 11336, 11305, 11275, 11245, 11215, 11185, 11155, 57 11125, 11096, 11067, 11038, 11009, 10980, 10951, 10923, 10894, 10866, 10838, 58 10810, 10782, 10755, 10727, 10700, 10673, 10645, 10618, 10592, 10565, 10538, 59 10512, 10486, 10460, 10434, 10408, 10382, 10356, 10331, 10305, 10280, 10255, 60 10230, 10205, 10180, 10156, 10131, 10107, 10082, 10058, 10034, 10010, 9986, 61 9963, 9939, 9916, 9892, 9869, 9846, 9823, 9800, 9777, 9754, 9732, 62 9709, 9687, 9664, 9642, 9620, 9598, 9576, 9554, 9533, 9511, 9489, 63 9468, 9447, 9425, 9404, 9383, 9362, 9341, 9321, 9300, 9279, 9259, 64 9239, 9218, 9198, 9178, 9158, 9138, 9118, 9098, 9079, 9059, 9039, 65 9020, 9001, 8981, 8962, 8943, 8924, 8905, 8886, 8867, 8849, 8830, 66 8812, 8793, 8775, 8756, 8738, 8720, 8702, 8684, 8666, 8648, 8630, 67 8613, 8595, 8577, 8560, 8542, 8525, 8508, 8490, 8473, 8456, 8439, 68 8422, 8405, 8389, 8372, 8355, 8339, 8322, 8306, 8289, 8273, 8257, 69 8240, 8224, 8208, 8192 70 }; 71 72 static uint32_t inverse_recenter(int r, uint32_t v) 73 { 74 if (v > 2 * r) 75 return v; 76 else if (v & 1) 77 return r - ((v + 1) >> 1); 78 else 79 return r + (v >> 1); 80 } 81 82 static uint32_t decode_unsigned_subexp_with_ref(uint32_t sub_exp, 83 int mx, int r) 84 { 85 if ((r << 1) <= mx) { 86 return inverse_recenter(r, sub_exp); 87 } else { 88 return mx - 1 - inverse_recenter(mx - 1 - r, sub_exp); 89 } 90 } 91 92 static int32_t decode_signed_subexp_with_ref(uint32_t sub_exp, int low, 93 int high, int r) 94 { 95 int32_t x = decode_unsigned_subexp_with_ref(sub_exp, high - low, r - low); 96 return x + low; 97 } 98 99 static void read_global_param(AV1DecContext *s, int type, int ref, int idx) 100 { 101 uint8_t primary_frame, prev_frame; 102 uint32_t abs_bits, prec_bits, round, prec_diff, sub, mx; 103 int32_t r, prev_gm_param; 104 105 primary_frame = s->raw_frame_header->primary_ref_frame; 106 prev_frame = s->raw_frame_header->ref_frame_idx[primary_frame]; 107 abs_bits = AV1_GM_ABS_ALPHA_BITS; 108 prec_bits = AV1_GM_ALPHA_PREC_BITS; 109 110 /* setup_past_independence() sets PrevGmParams to default values. We can 111 * simply point to the current's frame gm_params as they will be initialized 112 * with defaults at this point. 113 */ 114 if (s->raw_frame_header->primary_ref_frame == AV1_PRIMARY_REF_NONE) 115 prev_gm_param = s->cur_frame.gm_params[ref][idx]; 116 else 117 prev_gm_param = s->ref[prev_frame].gm_params[ref][idx]; 118 119 if (idx < 2) { 120 if (type == AV1_WARP_MODEL_TRANSLATION) { 121 abs_bits = AV1_GM_ABS_TRANS_ONLY_BITS - 122 !s->raw_frame_header->allow_high_precision_mv; 123 prec_bits = AV1_GM_TRANS_ONLY_PREC_BITS - 124 !s->raw_frame_header->allow_high_precision_mv; 125 } else { 126 abs_bits = AV1_GM_ABS_TRANS_BITS; 127 prec_bits = AV1_GM_TRANS_PREC_BITS; 128 } 129 } 130 round = (idx % 3) == 2 ? (1 << AV1_WARPEDMODEL_PREC_BITS) : 0; 131 prec_diff = AV1_WARPEDMODEL_PREC_BITS - prec_bits; 132 sub = (idx % 3) == 2 ? (1 << prec_bits) : 0; 133 mx = 1 << abs_bits; 134 r = (prev_gm_param >> prec_diff) - sub; 135 136 s->cur_frame.gm_params[ref][idx] = 137 (decode_signed_subexp_with_ref(s->raw_frame_header->gm_params[ref][idx], 138 -mx, mx + 1, r) << prec_diff) + round; 139 } 140 141 static uint64_t round_two(uint64_t x, uint16_t n) 142 { 143 if (n == 0) 144 return x; 145 return ((x + ((uint64_t)1 << (n - 1))) >> n); 146 } 147 148 static int64_t round_two_signed(int64_t x, uint16_t n) 149 { 150 return ((x<0) ? -((int64_t)round_two(-x, n)) : (int64_t)round_two(x, n)); 151 } 152 153 /** 154 * Resolve divisor process. 155 * see spec 7.11.3.7 156 */ 157 static int16_t resolve_divisor(uint32_t d, uint16_t *shift) 158 { 159 int32_t e, f; 160 161 *shift = av_log2(d); 162 e = d - (1 << (*shift)); 163 if (*shift > AV1_DIV_LUT_BITS) 164 f = round_two(e, *shift - AV1_DIV_LUT_BITS); 165 else 166 f = e << (AV1_DIV_LUT_BITS - (*shift)); 167 168 *shift += AV1_DIV_LUT_PREC_BITS; 169 170 return div_lut[f]; 171 } 172 173 /** 174 * check if global motion params is valid. 175 * see spec 7.11.3.6 176 */ 177 static uint8_t get_shear_params_valid(AV1DecContext *s, int idx) 178 { 179 int16_t alpha, beta, gamma, delta, divf, divs; 180 int64_t v, w; 181 int32_t *param = &s->cur_frame.gm_params[idx][0]; 182 if (param[2] <= 0) 183 return 0; 184 185 alpha = av_clip_int16(param[2] - (1 << AV1_WARPEDMODEL_PREC_BITS)); 186 beta = av_clip_int16(param[3]); 187 divf = resolve_divisor(abs(param[2]), &divs); 188 v = (int64_t)param[4] * (1 << AV1_WARPEDMODEL_PREC_BITS); 189 w = (int64_t)param[3] * param[4]; 190 gamma = av_clip_int16((int)round_two_signed((v * divf), divs)); 191 delta = av_clip_int16(param[5] - (int)round_two_signed((w * divf), divs) - (1 << AV1_WARPEDMODEL_PREC_BITS)); 192 193 alpha = round_two_signed(alpha, AV1_WARP_PARAM_REDUCE_BITS) << AV1_WARP_PARAM_REDUCE_BITS; 194 beta = round_two_signed(beta, AV1_WARP_PARAM_REDUCE_BITS) << AV1_WARP_PARAM_REDUCE_BITS; 195 gamma = round_two_signed(gamma, AV1_WARP_PARAM_REDUCE_BITS) << AV1_WARP_PARAM_REDUCE_BITS; 196 delta = round_two_signed(delta, AV1_WARP_PARAM_REDUCE_BITS) << AV1_WARP_PARAM_REDUCE_BITS; 197 198 if ((4 * abs(alpha) + 7 * abs(beta)) >= (1 << AV1_WARPEDMODEL_PREC_BITS) || 199 (4 * abs(gamma) + 4 * abs(delta)) >= (1 << AV1_WARPEDMODEL_PREC_BITS)) 200 return 0; 201 202 return 1; 203 } 204 205 /** 206 * update gm type/params, since cbs already implemented part of this function, 207 * so we don't need to full implement spec. 208 */ 209 static void global_motion_params(AV1DecContext *s) 210 { 211 const AV1RawFrameHeader *header = s->raw_frame_header; 212 int type, ref; 213 214 for (ref = AV1_REF_FRAME_LAST; ref <= AV1_REF_FRAME_ALTREF; ref++) { 215 s->cur_frame.gm_type[ref] = AV1_WARP_MODEL_IDENTITY; 216 for (int i = 0; i < 6; i++) 217 s->cur_frame.gm_params[ref][i] = (i % 3 == 2) ? 218 1 << AV1_WARPEDMODEL_PREC_BITS : 0; 219 } 220 if (header->frame_type == AV1_FRAME_KEY || 221 header->frame_type == AV1_FRAME_INTRA_ONLY) 222 return; 223 224 for (ref = AV1_REF_FRAME_LAST; ref <= AV1_REF_FRAME_ALTREF; ref++) { 225 if (header->is_global[ref]) { 226 if (header->is_rot_zoom[ref]) { 227 type = AV1_WARP_MODEL_ROTZOOM; 228 } else { 229 type = header->is_translation[ref] ? AV1_WARP_MODEL_TRANSLATION 230 : AV1_WARP_MODEL_AFFINE; 231 } 232 } else { 233 type = AV1_WARP_MODEL_IDENTITY; 234 } 235 s->cur_frame.gm_type[ref] = type; 236 237 if (type >= AV1_WARP_MODEL_ROTZOOM) { 238 read_global_param(s, type, ref, 2); 239 read_global_param(s, type, ref, 3); 240 if (type == AV1_WARP_MODEL_AFFINE) { 241 read_global_param(s, type, ref, 4); 242 read_global_param(s, type, ref, 5); 243 } else { 244 s->cur_frame.gm_params[ref][4] = -s->cur_frame.gm_params[ref][3]; 245 s->cur_frame.gm_params[ref][5] = s->cur_frame.gm_params[ref][2]; 246 } 247 } 248 if (type >= AV1_WARP_MODEL_TRANSLATION) { 249 read_global_param(s, type, ref, 0); 250 read_global_param(s, type, ref, 1); 251 } 252 if (type <= AV1_WARP_MODEL_AFFINE) { 253 s->cur_frame.gm_invalid[ref] = !get_shear_params_valid(s, ref); 254 } 255 } 256 } 257 258 static int get_relative_dist(const AV1RawSequenceHeader *seq, 259 unsigned int a, unsigned int b) 260 { 261 unsigned int diff = a - b; 262 unsigned int m = 1 << seq->order_hint_bits_minus_1; 263 return (diff & (m - 1)) - (diff & m); 264 } 265 266 static void skip_mode_params(AV1DecContext *s) 267 { 268 const AV1RawFrameHeader *header = s->raw_frame_header; 269 const AV1RawSequenceHeader *seq = s->raw_seq; 270 271 int forward_idx, backward_idx; 272 int forward_hint, backward_hint; 273 int second_forward_idx, second_forward_hint; 274 int ref_hint, dist, i; 275 276 if (header->frame_type == AV1_FRAME_KEY || 277 header->frame_type == AV1_FRAME_INTRA_ONLY || 278 !header->reference_select || !seq->enable_order_hint) 279 return; 280 281 forward_idx = -1; 282 backward_idx = -1; 283 for (i = 0; i < AV1_REFS_PER_FRAME; i++) { 284 if (!s->ref[header->ref_frame_idx[i]].raw_frame_header) 285 return; 286 ref_hint = s->ref[header->ref_frame_idx[i]].raw_frame_header->order_hint; 287 dist = get_relative_dist(seq, ref_hint, header->order_hint); 288 if (dist < 0) { 289 if (forward_idx < 0 || 290 get_relative_dist(seq, ref_hint, forward_hint) > 0) { 291 forward_idx = i; 292 forward_hint = ref_hint; 293 } 294 } else if (dist > 0) { 295 if (backward_idx < 0 || 296 get_relative_dist(seq, ref_hint, backward_hint) < 0) { 297 backward_idx = i; 298 backward_hint = ref_hint; 299 } 300 } 301 } 302 303 if (forward_idx < 0) { 304 return; 305 } else if (backward_idx >= 0) { 306 s->cur_frame.skip_mode_frame_idx[0] = 307 AV1_REF_FRAME_LAST + FFMIN(forward_idx, backward_idx); 308 s->cur_frame.skip_mode_frame_idx[1] = 309 AV1_REF_FRAME_LAST + FFMAX(forward_idx, backward_idx); 310 return; 311 } 312 313 second_forward_idx = -1; 314 for (i = 0; i < AV1_REFS_PER_FRAME; i++) { 315 ref_hint = s->ref[header->ref_frame_idx[i]].raw_frame_header->order_hint; 316 if (get_relative_dist(seq, ref_hint, forward_hint) < 0) { 317 if (second_forward_idx < 0 || 318 get_relative_dist(seq, ref_hint, second_forward_hint) > 0) { 319 second_forward_idx = i; 320 second_forward_hint = ref_hint; 321 } 322 } 323 } 324 325 if (second_forward_idx < 0) 326 return; 327 328 s->cur_frame.skip_mode_frame_idx[0] = 329 AV1_REF_FRAME_LAST + FFMIN(forward_idx, second_forward_idx); 330 s->cur_frame.skip_mode_frame_idx[1] = 331 AV1_REF_FRAME_LAST + FFMAX(forward_idx, second_forward_idx); 332 } 333 334 static void coded_lossless_param(AV1DecContext *s) 335 { 336 const AV1RawFrameHeader *header = s->raw_frame_header; 337 int i; 338 339 if (header->delta_q_y_dc || header->delta_q_u_ac || 340 header->delta_q_u_dc || header->delta_q_v_ac || 341 header->delta_q_v_dc) { 342 s->cur_frame.coded_lossless = 0; 343 return; 344 } 345 346 s->cur_frame.coded_lossless = 1; 347 for (i = 0; i < AV1_MAX_SEGMENTS; i++) { 348 int qindex; 349 if (header->feature_enabled[i][AV1_SEG_LVL_ALT_Q]) { 350 qindex = (header->base_q_idx + 351 header->feature_value[i][AV1_SEG_LVL_ALT_Q]); 352 } else { 353 qindex = header->base_q_idx; 354 } 355 qindex = av_clip_uintp2(qindex, 8); 356 357 if (qindex) { 358 s->cur_frame.coded_lossless = 0; 359 return; 360 } 361 } 362 } 363 364 static void order_hint_info(AV1DecContext *s) 365 { 366 const AV1RawFrameHeader *header = s->raw_frame_header; 367 const AV1RawSequenceHeader *seq = s->raw_seq; 368 AV1Frame *frame = &s->cur_frame; 369 370 frame->order_hint = header->order_hint; 371 372 for (int i = 0; i < AV1_REFS_PER_FRAME; i++) { 373 int ref_name = i + AV1_REF_FRAME_LAST; 374 int ref_slot = header->ref_frame_idx[i]; 375 int ref_order_hint = s->ref[ref_slot].order_hint; 376 377 frame->order_hints[ref_name] = ref_order_hint; 378 if (!seq->enable_order_hint) { 379 frame->ref_frame_sign_bias[ref_name] = 0; 380 } else { 381 frame->ref_frame_sign_bias[ref_name] = 382 get_relative_dist(seq, ref_order_hint, 383 frame->order_hint) > 0; 384 } 385 } 386 } 387 388 static void load_grain_params(AV1DecContext *s) 389 { 390 const AV1RawFrameHeader *header = s->raw_frame_header; 391 const AV1RawFilmGrainParams *film_grain = &header->film_grain, *src; 392 AV1RawFilmGrainParams *dst = &s->cur_frame.film_grain; 393 394 if (!film_grain->apply_grain) 395 return; 396 397 if (film_grain->update_grain) { 398 memcpy(dst, film_grain, sizeof(*dst)); 399 return; 400 } 401 402 src = &s->ref[film_grain->film_grain_params_ref_idx].film_grain; 403 404 memcpy(dst, src, sizeof(*dst)); 405 dst->grain_seed = film_grain->grain_seed; 406 } 407 408 static int init_tile_data(AV1DecContext *s) 409 410 { 411 int cur_tile_num = 412 s->raw_frame_header->tile_cols * s->raw_frame_header->tile_rows; 413 if (s->tile_num < cur_tile_num) { 414 int ret = av_reallocp_array(&s->tile_group_info, cur_tile_num, 415 sizeof(TileGroupInfo)); 416 if (ret < 0) { 417 s->tile_num = 0; 418 return ret; 419 } 420 } 421 s->tile_num = cur_tile_num; 422 423 return 0; 424 } 425 426 static int get_tiles_info(AVCodecContext *avctx, const AV1RawTileGroup *tile_group) 427 { 428 AV1DecContext *s = avctx->priv_data; 429 GetByteContext gb; 430 uint16_t tile_num, tile_row, tile_col; 431 uint32_t size = 0, size_bytes = 0; 432 433 bytestream2_init(&gb, tile_group->tile_data.data, 434 tile_group->tile_data.data_size); 435 s->tg_start = tile_group->tg_start; 436 s->tg_end = tile_group->tg_end; 437 438 for (tile_num = tile_group->tg_start; tile_num <= tile_group->tg_end; tile_num++) { 439 tile_row = tile_num / s->raw_frame_header->tile_cols; 440 tile_col = tile_num % s->raw_frame_header->tile_cols; 441 442 if (tile_num == tile_group->tg_end) { 443 s->tile_group_info[tile_num].tile_size = bytestream2_get_bytes_left(&gb); 444 s->tile_group_info[tile_num].tile_offset = bytestream2_tell(&gb); 445 s->tile_group_info[tile_num].tile_row = tile_row; 446 s->tile_group_info[tile_num].tile_column = tile_col; 447 return 0; 448 } 449 size_bytes = s->raw_frame_header->tile_size_bytes_minus1 + 1; 450 if (bytestream2_get_bytes_left(&gb) < size_bytes) 451 return AVERROR_INVALIDDATA; 452 size = 0; 453 for (int i = 0; i < size_bytes; i++) 454 size |= bytestream2_get_byteu(&gb) << 8 * i; 455 if (bytestream2_get_bytes_left(&gb) <= size) 456 return AVERROR_INVALIDDATA; 457 size++; 458 459 s->tile_group_info[tile_num].tile_size = size; 460 s->tile_group_info[tile_num].tile_offset = bytestream2_tell(&gb); 461 s->tile_group_info[tile_num].tile_row = tile_row; 462 s->tile_group_info[tile_num].tile_column = tile_col; 463 464 bytestream2_skipu(&gb, size); 465 } 466 467 return 0; 468 469 } 470 471 static enum AVPixelFormat get_sw_pixel_format(void *logctx, 472 const AV1RawSequenceHeader *seq) 473 { 474 int bit_depth; 475 enum AVPixelFormat pix_fmt = AV_PIX_FMT_NONE; 476 477 if (seq->seq_profile == 2 && seq->color_config.high_bitdepth) 478 bit_depth = seq->color_config.twelve_bit ? 12 : 10; 479 else if (seq->seq_profile <= 2) 480 bit_depth = seq->color_config.high_bitdepth ? 10 : 8; 481 else { 482 av_log(logctx, AV_LOG_ERROR, 483 "Unknown AV1 profile %d.\n", seq->seq_profile); 484 return AV_PIX_FMT_NONE; 485 } 486 487 if (!seq->color_config.mono_chrome) { 488 // 4:4:4 x:0 y:0, 4:2:2 x:1 y:0, 4:2:0 x:1 y:1 489 if (seq->color_config.subsampling_x == 0 && 490 seq->color_config.subsampling_y == 0) { 491 if (bit_depth == 8) 492 pix_fmt = AV_PIX_FMT_YUV444P; 493 else if (bit_depth == 10) 494 pix_fmt = AV_PIX_FMT_YUV444P10; 495 else if (bit_depth == 12) 496 pix_fmt = AV_PIX_FMT_YUV444P12; 497 else 498 av_assert0(0); 499 } else if (seq->color_config.subsampling_x == 1 && 500 seq->color_config.subsampling_y == 0) { 501 if (bit_depth == 8) 502 pix_fmt = AV_PIX_FMT_YUV422P; 503 else if (bit_depth == 10) 504 pix_fmt = AV_PIX_FMT_YUV422P10; 505 else if (bit_depth == 12) 506 pix_fmt = AV_PIX_FMT_YUV422P12; 507 else 508 av_assert0(0); 509 } else if (seq->color_config.subsampling_x == 1 && 510 seq->color_config.subsampling_y == 1) { 511 if (bit_depth == 8) 512 pix_fmt = AV_PIX_FMT_YUV420P; 513 else if (bit_depth == 10) 514 pix_fmt = AV_PIX_FMT_YUV420P10; 515 else if (bit_depth == 12) 516 pix_fmt = AV_PIX_FMT_YUV420P12; 517 else 518 av_assert0(0); 519 } 520 } else { 521 if (bit_depth == 8) 522 pix_fmt = AV_PIX_FMT_GRAY8; 523 else if (bit_depth == 10) 524 pix_fmt = AV_PIX_FMT_GRAY10; 525 else if (bit_depth == 12) 526 pix_fmt = AV_PIX_FMT_GRAY12; 527 else 528 av_assert0(0); 529 } 530 531 return pix_fmt; 532 } 533 534 static int get_pixel_format(AVCodecContext *avctx) 535 { 536 AV1DecContext *s = avctx->priv_data; 537 const AV1RawSequenceHeader *seq = s->raw_seq; 538 int ret; 539 enum AVPixelFormat pix_fmt = get_sw_pixel_format(avctx, seq); 540 #define HWACCEL_MAX (CONFIG_AV1_DXVA2_HWACCEL + \ 541 CONFIG_AV1_D3D11VA_HWACCEL * 2 + \ 542 CONFIG_AV1_D3D12VA_HWACCEL + \ 543 CONFIG_AV1_NVDEC_HWACCEL + \ 544 CONFIG_AV1_VAAPI_HWACCEL + \ 545 CONFIG_AV1_VDPAU_HWACCEL + \ 546 CONFIG_AV1_VIDEOTOOLBOX_HWACCEL + \ 547 CONFIG_AV1_VULKAN_HWACCEL) 548 enum AVPixelFormat pix_fmts[HWACCEL_MAX + 2], *fmtp = pix_fmts; 549 550 if (pix_fmt == AV_PIX_FMT_NONE) 551 return -1; 552 553 switch (pix_fmt) { 554 case AV_PIX_FMT_YUV420P: 555 #if CONFIG_AV1_DXVA2_HWACCEL 556 *fmtp++ = AV_PIX_FMT_DXVA2_VLD; 557 #endif 558 #if CONFIG_AV1_D3D11VA_HWACCEL 559 *fmtp++ = AV_PIX_FMT_D3D11VA_VLD; 560 *fmtp++ = AV_PIX_FMT_D3D11; 561 #endif 562 #if CONFIG_AV1_D3D12VA_HWACCEL 563 *fmtp++ = AV_PIX_FMT_D3D12; 564 #endif 565 #if CONFIG_AV1_NVDEC_HWACCEL 566 *fmtp++ = AV_PIX_FMT_CUDA; 567 #endif 568 #if CONFIG_AV1_VAAPI_HWACCEL 569 *fmtp++ = AV_PIX_FMT_VAAPI; 570 #endif 571 #if CONFIG_AV1_VDPAU_HWACCEL 572 *fmtp++ = AV_PIX_FMT_VDPAU; 573 #endif 574 #if CONFIG_AV1_VIDEOTOOLBOX_HWACCEL 575 *fmtp++ = AV_PIX_FMT_VIDEOTOOLBOX; 576 #endif 577 #if CONFIG_AV1_VULKAN_HWACCEL 578 *fmtp++ = AV_PIX_FMT_VULKAN; 579 #endif 580 break; 581 case AV_PIX_FMT_YUV420P10: 582 #if CONFIG_AV1_DXVA2_HWACCEL 583 *fmtp++ = AV_PIX_FMT_DXVA2_VLD; 584 #endif 585 #if CONFIG_AV1_D3D11VA_HWACCEL 586 *fmtp++ = AV_PIX_FMT_D3D11VA_VLD; 587 *fmtp++ = AV_PIX_FMT_D3D11; 588 #endif 589 #if CONFIG_AV1_D3D12VA_HWACCEL 590 *fmtp++ = AV_PIX_FMT_D3D12; 591 #endif 592 #if CONFIG_AV1_NVDEC_HWACCEL 593 *fmtp++ = AV_PIX_FMT_CUDA; 594 #endif 595 #if CONFIG_AV1_VAAPI_HWACCEL 596 *fmtp++ = AV_PIX_FMT_VAAPI; 597 #endif 598 #if CONFIG_AV1_VDPAU_HWACCEL 599 *fmtp++ = AV_PIX_FMT_VDPAU; 600 #endif 601 #if CONFIG_AV1_VIDEOTOOLBOX_HWACCEL 602 *fmtp++ = AV_PIX_FMT_VIDEOTOOLBOX; 603 #endif 604 #if CONFIG_AV1_VULKAN_HWACCEL 605 *fmtp++ = AV_PIX_FMT_VULKAN; 606 #endif 607 break; 608 case AV_PIX_FMT_YUV420P12: 609 #if CONFIG_AV1_VULKAN_HWACCEL 610 *fmtp++ = AV_PIX_FMT_VULKAN; 611 #endif 612 break; 613 case AV_PIX_FMT_YUV422P: 614 #if CONFIG_AV1_VULKAN_HWACCEL 615 *fmtp++ = AV_PIX_FMT_VULKAN; 616 #endif 617 break; 618 case AV_PIX_FMT_YUV422P10: 619 #if CONFIG_AV1_VULKAN_HWACCEL 620 *fmtp++ = AV_PIX_FMT_VULKAN; 621 #endif 622 break; 623 case AV_PIX_FMT_YUV422P12: 624 #if CONFIG_AV1_VULKAN_HWACCEL 625 *fmtp++ = AV_PIX_FMT_VULKAN; 626 #endif 627 break; 628 case AV_PIX_FMT_YUV444P: 629 #if CONFIG_AV1_VULKAN_HWACCEL 630 *fmtp++ = AV_PIX_FMT_VULKAN; 631 #endif 632 break; 633 case AV_PIX_FMT_YUV444P10: 634 #if CONFIG_AV1_VULKAN_HWACCEL 635 *fmtp++ = AV_PIX_FMT_VULKAN; 636 #endif 637 break; 638 case AV_PIX_FMT_YUV444P12: 639 #if CONFIG_AV1_VULKAN_HWACCEL 640 *fmtp++ = AV_PIX_FMT_VULKAN; 641 #endif 642 break; 643 case AV_PIX_FMT_GRAY8: 644 #if CONFIG_AV1_NVDEC_HWACCEL 645 *fmtp++ = AV_PIX_FMT_CUDA; 646 #endif 647 break; 648 case AV_PIX_FMT_GRAY10: 649 #if CONFIG_AV1_NVDEC_HWACCEL 650 *fmtp++ = AV_PIX_FMT_CUDA; 651 #endif 652 break; 653 } 654 655 *fmtp++ = pix_fmt; 656 *fmtp = AV_PIX_FMT_NONE; 657 658 for (int i = 0; pix_fmts[i] != pix_fmt; i++) 659 if (pix_fmts[i] == avctx->pix_fmt) { 660 s->pix_fmt = pix_fmt; 661 return 1; 662 } 663 664 ret = ff_get_format(avctx, pix_fmts); 665 666 /** 667 * check if the HW accel is inited correctly. If not, return un-implemented. 668 * Since now the av1 decoder doesn't support native decode, if it will be 669 * implemented in the future, need remove this check. 670 */ 671 if (!avctx->hwaccel) { 672 av_log(avctx, AV_LOG_ERROR, "Your platform doesn't support" 673 " hardware accelerated AV1 decoding.\n"); 674 avctx->pix_fmt = AV_PIX_FMT_NONE; 675 return AVERROR(ENOSYS); 676 } 677 678 s->pix_fmt = pix_fmt; 679 avctx->pix_fmt = ret; 680 681 av_log(avctx, AV_LOG_DEBUG, "AV1 decode get format: %s.\n", 682 av_get_pix_fmt_name(avctx->pix_fmt)); 683 684 return 0; 685 } 686 687 static void av1_frame_unref(AV1Frame *f) 688 { 689 ff_progress_frame_unref(&f->pf); 690 av_refstruct_unref(&f->hwaccel_picture_private); 691 av_refstruct_unref(&f->header_ref); 692 f->raw_frame_header = NULL; 693 f->spatial_id = f->temporal_id = 0; 694 memset(f->skip_mode_frame_idx, 0, 695 2 * sizeof(uint8_t)); 696 memset(&f->film_grain, 0, sizeof(f->film_grain)); 697 f->coded_lossless = 0; 698 } 699 700 static void av1_frame_replace(AV1Frame *dst, const AV1Frame *src) 701 { 702 av_assert1(dst != src); 703 704 av_refstruct_replace(&dst->header_ref, src->header_ref); 705 706 dst->raw_frame_header = src->raw_frame_header; 707 708 ff_progress_frame_replace(&dst->pf, &src->pf); 709 710 av_refstruct_replace(&dst->hwaccel_picture_private, 711 src->hwaccel_picture_private); 712 713 dst->spatial_id = src->spatial_id; 714 dst->temporal_id = src->temporal_id; 715 memcpy(dst->gm_invalid, 716 src->gm_invalid, 717 AV1_NUM_REF_FRAMES * sizeof(uint8_t)); 718 memcpy(dst->gm_type, 719 src->gm_type, 720 AV1_NUM_REF_FRAMES * sizeof(uint8_t)); 721 memcpy(dst->gm_params, 722 src->gm_params, 723 AV1_NUM_REF_FRAMES * 6 * sizeof(int32_t)); 724 memcpy(dst->skip_mode_frame_idx, 725 src->skip_mode_frame_idx, 726 2 * sizeof(uint8_t)); 727 memcpy(&dst->film_grain, 728 &src->film_grain, 729 sizeof(dst->film_grain)); 730 dst->coded_lossless = src->coded_lossless; 731 732 dst->order_hint = src->order_hint; 733 memcpy(dst->ref_frame_sign_bias, src->ref_frame_sign_bias, 734 sizeof(dst->ref_frame_sign_bias)); 735 memcpy(dst->order_hints, src->order_hints, 736 sizeof(dst->order_hints)); 737 738 dst->force_integer_mv = src->force_integer_mv; 739 } 740 741 static av_cold int av1_decode_free(AVCodecContext *avctx) 742 { 743 AV1DecContext *s = avctx->priv_data; 744 AV1RawMetadataITUTT35 itut_t35; 745 746 for (int i = 0; i < FF_ARRAY_ELEMS(s->ref); i++) 747 av1_frame_unref(&s->ref[i]); 748 av1_frame_unref(&s->cur_frame); 749 av_buffer_unref(&s->seq_data_ref); 750 av_refstruct_unref(&s->seq_ref); 751 av_refstruct_unref(&s->header_ref); 752 av_refstruct_unref(&s->cll_ref); 753 av_refstruct_unref(&s->mdcv_ref); 754 av_freep(&s->tile_group_info); 755 756 while (s->itut_t35_fifo && av_fifo_read(s->itut_t35_fifo, &itut_t35, 1) >= 0) 757 av_buffer_unref(&itut_t35.payload_ref); 758 av_fifo_freep2(&s->itut_t35_fifo); 759 760 ff_cbs_fragment_free(&s->current_obu); 761 ff_cbs_close(&s->cbc); 762 ff_dovi_ctx_unref(&s->dovi); 763 764 return 0; 765 } 766 767 static int set_context_with_sequence(AVCodecContext *avctx, 768 const AV1RawSequenceHeader *seq) 769 { 770 int width = seq->max_frame_width_minus_1 + 1; 771 int height = seq->max_frame_height_minus_1 + 1; 772 773 avctx->profile = seq->seq_profile; 774 avctx->level = seq->seq_level_idx[0]; 775 776 avctx->color_range = 777 seq->color_config.color_range ? AVCOL_RANGE_JPEG : AVCOL_RANGE_MPEG; 778 avctx->color_primaries = seq->color_config.color_primaries; 779 avctx->colorspace = seq->color_config.matrix_coefficients; 780 avctx->color_trc = seq->color_config.transfer_characteristics; 781 782 switch (seq->color_config.chroma_sample_position) { 783 case AV1_CSP_VERTICAL: 784 avctx->chroma_sample_location = AVCHROMA_LOC_LEFT; 785 break; 786 case AV1_CSP_COLOCATED: 787 avctx->chroma_sample_location = AVCHROMA_LOC_TOPLEFT; 788 break; 789 } 790 791 #if FF_API_CODEC_PROPS 792 FF_DISABLE_DEPRECATION_WARNINGS 793 if (seq->film_grain_params_present) 794 avctx->properties |= FF_CODEC_PROPERTY_FILM_GRAIN; 795 else 796 avctx->properties &= ~FF_CODEC_PROPERTY_FILM_GRAIN; 797 FF_ENABLE_DEPRECATION_WARNINGS 798 #endif 799 800 if (avctx->width != width || avctx->height != height) { 801 int ret = ff_set_dimensions(avctx, width, height); 802 if (ret < 0) 803 return ret; 804 } 805 806 if (seq->timing_info_present_flag) 807 avctx->framerate = ff_av1_framerate(1LL + seq->timing_info.num_ticks_per_picture_minus_1, 808 seq->timing_info.num_units_in_display_tick, 809 seq->timing_info.time_scale); 810 811 if (avctx->pix_fmt == AV_PIX_FMT_NONE) 812 avctx->pix_fmt = get_sw_pixel_format(avctx, seq); 813 814 return 0; 815 } 816 817 static int update_context_with_frame_header(AVCodecContext *avctx, 818 const AV1RawFrameHeader *header) 819 { 820 AVRational aspect_ratio; 821 int width = header->frame_width_minus_1 + 1; 822 int height = header->frame_height_minus_1 + 1; 823 int r_width = header->render_width_minus_1 + 1; 824 int r_height = header->render_height_minus_1 + 1; 825 int ret; 826 827 if (avctx->width != width || avctx->height != height) { 828 ret = ff_set_dimensions(avctx, width, height); 829 if (ret < 0) 830 return ret; 831 } 832 833 av_reduce(&aspect_ratio.num, &aspect_ratio.den, 834 (int64_t)height * r_width, 835 (int64_t)width * r_height, 836 INT_MAX); 837 838 if (av_cmp_q(avctx->sample_aspect_ratio, aspect_ratio)) { 839 ret = ff_set_sar(avctx, aspect_ratio); 840 if (ret < 0) 841 return ret; 842 } 843 844 return 0; 845 } 846 847 static const CodedBitstreamUnitType decompose_unit_types[] = { 848 AV1_OBU_FRAME, 849 AV1_OBU_FRAME_HEADER, 850 AV1_OBU_METADATA, 851 AV1_OBU_REDUNDANT_FRAME_HEADER, 852 AV1_OBU_SEQUENCE_HEADER, 853 AV1_OBU_TEMPORAL_DELIMITER, 854 AV1_OBU_TILE_GROUP, 855 }; 856 857 static av_cold int av1_decode_init(AVCodecContext *avctx) 858 { 859 AV1DecContext *s = avctx->priv_data; 860 AV1RawSequenceHeader *seq; 861 const AVPacketSideData *sd; 862 int ret; 863 864 s->avctx = avctx; 865 s->pkt = avctx->internal->in_pkt; 866 s->pix_fmt = AV_PIX_FMT_NONE; 867 868 ret = ff_cbs_init(&s->cbc, AV_CODEC_ID_AV1, avctx); 869 if (ret < 0) 870 return ret; 871 872 s->cbc->decompose_unit_types = decompose_unit_types; 873 s->cbc->nb_decompose_unit_types = FF_ARRAY_ELEMS(decompose_unit_types); 874 875 s->itut_t35_fifo = av_fifo_alloc2(1, sizeof(AV1RawMetadataITUTT35), 876 AV_FIFO_FLAG_AUTO_GROW); 877 if (!s->itut_t35_fifo) 878 return AVERROR(ENOMEM); 879 880 av_opt_set_int(s->cbc->priv_data, "operating_point", s->operating_point, 0); 881 882 if (avctx->extradata && avctx->extradata_size) { 883 ret = ff_cbs_read_extradata_from_codec(s->cbc, 884 &s->current_obu, 885 avctx); 886 if (ret < 0) { 887 av_log(avctx, AV_LOG_WARNING, "Failed to read extradata.\n"); 888 goto end; 889 } 890 891 seq = ((CodedBitstreamAV1Context *)(s->cbc->priv_data))->sequence_header; 892 if (!seq) { 893 av_log(avctx, AV_LOG_WARNING, "No sequence header available.\n"); 894 goto end; 895 } 896 897 ret = set_context_with_sequence(avctx, seq); 898 if (ret < 0) { 899 av_log(avctx, AV_LOG_WARNING, "Failed to set decoder context.\n"); 900 goto end; 901 } 902 903 end: 904 ff_cbs_fragment_reset(&s->current_obu); 905 } 906 907 s->dovi.logctx = avctx; 908 s->dovi.cfg.dv_profile = 10; // default for AV1 909 sd = ff_get_coded_side_data(avctx, AV_PKT_DATA_DOVI_CONF); 910 if (sd && sd->size >= sizeof(s->dovi.cfg)) 911 s->dovi.cfg = *(AVDOVIDecoderConfigurationRecord *) sd->data; 912 913 return ret; 914 } 915 916 static int av1_frame_alloc(AVCodecContext *avctx, AV1Frame *f) 917 { 918 AV1DecContext *s = avctx->priv_data; 919 AV1RawFrameHeader *header= s->raw_frame_header; 920 AVFrame *frame; 921 int ret; 922 923 ret = update_context_with_frame_header(avctx, header); 924 if (ret < 0) { 925 av_log(avctx, AV_LOG_ERROR, "Failed to update context with frame header\n"); 926 return ret; 927 } 928 929 ret = ff_progress_frame_get_buffer(avctx, &f->pf, AV_GET_BUFFER_FLAG_REF); 930 if (ret < 0) 931 goto fail; 932 933 frame = f->f; 934 if (header->frame_type == AV1_FRAME_KEY) 935 frame->flags |= AV_FRAME_FLAG_KEY; 936 else 937 frame->flags &= ~AV_FRAME_FLAG_KEY; 938 939 switch (header->frame_type) { 940 case AV1_FRAME_KEY: 941 case AV1_FRAME_INTRA_ONLY: 942 frame->pict_type = AV_PICTURE_TYPE_I; 943 break; 944 case AV1_FRAME_INTER: 945 frame->pict_type = AV_PICTURE_TYPE_P; 946 break; 947 case AV1_FRAME_SWITCH: 948 frame->pict_type = AV_PICTURE_TYPE_SP; 949 break; 950 } 951 952 ret = ff_hwaccel_frame_priv_alloc(avctx, &f->hwaccel_picture_private); 953 if (ret < 0) 954 goto fail; 955 956 return 0; 957 958 fail: 959 av1_frame_unref(f); 960 return ret; 961 } 962 963 static int export_itut_t35(AVCodecContext *avctx, AVFrame *frame, 964 const AV1RawMetadataITUTT35 *itut_t35) 965 { 966 GetByteContext gb; 967 AV1DecContext *s = avctx->priv_data; 968 int ret, provider_code; 969 970 bytestream2_init(&gb, itut_t35->payload, itut_t35->payload_size); 971 972 provider_code = bytestream2_get_be16(&gb); 973 switch (provider_code) { 974 case ITU_T_T35_PROVIDER_CODE_ATSC: { 975 uint32_t user_identifier = bytestream2_get_be32(&gb); 976 switch (user_identifier) { 977 case MKBETAG('G', 'A', '9', '4'): { // closed captions 978 AVBufferRef *buf = NULL; 979 980 ret = ff_parse_a53_cc(&buf, gb.buffer, bytestream2_get_bytes_left(&gb)); 981 if (ret < 0) 982 return ret; 983 if (!ret) 984 break; 985 986 ret = ff_frame_new_side_data_from_buf(avctx, frame, AV_FRAME_DATA_A53_CC, &buf); 987 if (ret < 0) 988 return ret; 989 990 #if FF_API_CODEC_PROPS 991 FF_DISABLE_DEPRECATION_WARNINGS 992 avctx->properties |= FF_CODEC_PROPERTY_CLOSED_CAPTIONS; 993 FF_ENABLE_DEPRECATION_WARNINGS 994 #endif 995 break; 996 } 997 default: // ignore unsupported identifiers 998 break; 999 } 1000 break; 1001 } 1002 case ITU_T_T35_PROVIDER_CODE_SMTPE: { 1003 AVDynamicHDRPlus *hdrplus; 1004 int provider_oriented_code = bytestream2_get_be16(&gb); 1005 int application_identifier = bytestream2_get_byte(&gb); 1006 1007 if (itut_t35->itu_t_t35_country_code != ITU_T_T35_COUNTRY_CODE_US || 1008 provider_oriented_code != 1 || application_identifier != 4) 1009 break; 1010 1011 hdrplus = av_dynamic_hdr_plus_create_side_data(frame); 1012 if (!hdrplus) 1013 return AVERROR(ENOMEM); 1014 1015 ret = av_dynamic_hdr_plus_from_t35(hdrplus, gb.buffer, 1016 bytestream2_get_bytes_left(&gb)); 1017 if (ret < 0) 1018 return ret; 1019 break; 1020 } 1021 case ITU_T_T35_PROVIDER_CODE_DOLBY: { 1022 int provider_oriented_code = bytestream2_get_be32(&gb); 1023 if (itut_t35->itu_t_t35_country_code != ITU_T_T35_COUNTRY_CODE_US || 1024 provider_oriented_code != 0x800) 1025 break; 1026 1027 ret = ff_dovi_rpu_parse(&s->dovi, gb.buffer, gb.buffer_end - gb.buffer, 1028 avctx->err_recognition); 1029 if (ret < 0) { 1030 av_log(avctx, AV_LOG_WARNING, "Error parsing DOVI OBU.\n"); 1031 break; // ignore 1032 } 1033 1034 ret = ff_dovi_attach_side_data(&s->dovi, frame); 1035 if (ret < 0) 1036 return ret; 1037 break; 1038 } 1039 default: // ignore unsupported provider codes 1040 break; 1041 } 1042 1043 return 0; 1044 } 1045 1046 static int export_metadata(AVCodecContext *avctx, AVFrame *frame) 1047 { 1048 AV1DecContext *s = avctx->priv_data; 1049 AV1RawMetadataITUTT35 itut_t35; 1050 int ret = 0; 1051 1052 if (s->mdcv) { 1053 AVMasteringDisplayMetadata *mastering; 1054 1055 ret = ff_decode_mastering_display_new(avctx, frame, &mastering); 1056 if (ret < 0) 1057 return ret; 1058 1059 if (mastering) { 1060 for (int i = 0; i < 3; i++) { 1061 mastering->display_primaries[i][0] = av_make_q(s->mdcv->primary_chromaticity_x[i], 1 << 16); 1062 mastering->display_primaries[i][1] = av_make_q(s->mdcv->primary_chromaticity_y[i], 1 << 16); 1063 } 1064 mastering->white_point[0] = av_make_q(s->mdcv->white_point_chromaticity_x, 1 << 16); 1065 mastering->white_point[1] = av_make_q(s->mdcv->white_point_chromaticity_y, 1 << 16); 1066 1067 mastering->max_luminance = av_make_q(s->mdcv->luminance_max, 1 << 8); 1068 mastering->min_luminance = av_make_q(s->mdcv->luminance_min, 1 << 14); 1069 1070 mastering->has_primaries = 1; 1071 mastering->has_luminance = 1; 1072 } 1073 } 1074 1075 if (s->cll) { 1076 AVContentLightMetadata *light; 1077 1078 ret = ff_decode_content_light_new(avctx, frame, &light); 1079 if (ret < 0) 1080 return ret; 1081 1082 if (light) { 1083 light->MaxCLL = s->cll->max_cll; 1084 light->MaxFALL = s->cll->max_fall; 1085 } 1086 } 1087 1088 while (av_fifo_read(s->itut_t35_fifo, &itut_t35, 1) >= 0) { 1089 if (ret >= 0) 1090 ret = export_itut_t35(avctx, frame, &itut_t35); 1091 av_buffer_unref(&itut_t35.payload_ref); 1092 } 1093 1094 return ret; 1095 } 1096 1097 static int export_film_grain(AVCodecContext *avctx, AVFrame *frame) 1098 { 1099 AV1DecContext *s = avctx->priv_data; 1100 const AV1RawFilmGrainParams *film_grain = &s->cur_frame.film_grain; 1101 const AVPixFmtDescriptor *pixdesc = av_pix_fmt_desc_get(frame->format); 1102 AVFilmGrainParams *fgp; 1103 AVFilmGrainAOMParams *aom; 1104 1105 av_assert0(pixdesc); 1106 if (!film_grain->apply_grain) 1107 return 0; 1108 1109 fgp = av_film_grain_params_create_side_data(frame); 1110 if (!fgp) 1111 return AVERROR(ENOMEM); 1112 1113 fgp->type = AV_FILM_GRAIN_PARAMS_AV1; 1114 fgp->seed = film_grain->grain_seed; 1115 fgp->width = frame->width; 1116 fgp->height = frame->height; 1117 fgp->color_range = frame->color_range; 1118 fgp->color_primaries = frame->color_primaries; 1119 fgp->color_trc = frame->color_trc; 1120 fgp->color_space = frame->colorspace; 1121 fgp->subsampling_x = pixdesc->log2_chroma_w; 1122 fgp->subsampling_y = pixdesc->log2_chroma_h; 1123 1124 aom = &fgp->codec.aom; 1125 aom->chroma_scaling_from_luma = film_grain->chroma_scaling_from_luma; 1126 aom->scaling_shift = film_grain->grain_scaling_minus_8 + 8; 1127 aom->ar_coeff_lag = film_grain->ar_coeff_lag; 1128 aom->ar_coeff_shift = film_grain->ar_coeff_shift_minus_6 + 6; 1129 aom->grain_scale_shift = film_grain->grain_scale_shift; 1130 aom->overlap_flag = film_grain->overlap_flag; 1131 aom->limit_output_range = film_grain->clip_to_restricted_range; 1132 1133 aom->num_y_points = film_grain->num_y_points; 1134 for (int i = 0; i < film_grain->num_y_points; i++) { 1135 aom->y_points[i][0] = film_grain->point_y_value[i]; 1136 aom->y_points[i][1] = film_grain->point_y_scaling[i]; 1137 } 1138 aom->num_uv_points[0] = film_grain->num_cb_points; 1139 for (int i = 0; i < film_grain->num_cb_points; i++) { 1140 aom->uv_points[0][i][0] = film_grain->point_cb_value[i]; 1141 aom->uv_points[0][i][1] = film_grain->point_cb_scaling[i]; 1142 } 1143 aom->num_uv_points[1] = film_grain->num_cr_points; 1144 for (int i = 0; i < film_grain->num_cr_points; i++) { 1145 aom->uv_points[1][i][0] = film_grain->point_cr_value[i]; 1146 aom->uv_points[1][i][1] = film_grain->point_cr_scaling[i]; 1147 } 1148 1149 for (int i = 0; i < 24; i++) { 1150 aom->ar_coeffs_y[i] = film_grain->ar_coeffs_y_plus_128[i] - 128; 1151 } 1152 for (int i = 0; i < 25; i++) { 1153 aom->ar_coeffs_uv[0][i] = film_grain->ar_coeffs_cb_plus_128[i] - 128; 1154 aom->ar_coeffs_uv[1][i] = film_grain->ar_coeffs_cr_plus_128[i] - 128; 1155 } 1156 1157 aom->uv_mult[0] = film_grain->cb_mult; 1158 aom->uv_mult[1] = film_grain->cr_mult; 1159 aom->uv_mult_luma[0] = film_grain->cb_luma_mult; 1160 aom->uv_mult_luma[1] = film_grain->cr_luma_mult; 1161 aom->uv_offset[0] = film_grain->cb_offset; 1162 aom->uv_offset[1] = film_grain->cr_offset; 1163 1164 return 0; 1165 } 1166 1167 static int set_output_frame(AVCodecContext *avctx, AVFrame *frame) 1168 { 1169 AV1DecContext *s = avctx->priv_data; 1170 const AVFrame *srcframe = s->cur_frame.f; 1171 AVPacket *pkt = s->pkt; 1172 int ret; 1173 1174 // TODO: all layers 1175 if (s->operating_point_idc && 1176 av_log2(s->operating_point_idc >> 8) > s->cur_frame.spatial_id) 1177 return 0; 1178 1179 ret = av_frame_ref(frame, srcframe); 1180 if (ret < 0) 1181 return ret; 1182 1183 ret = export_metadata(avctx, frame); 1184 if (ret < 0) { 1185 av_frame_unref(frame); 1186 return ret; 1187 } 1188 1189 if (avctx->export_side_data & AV_CODEC_EXPORT_DATA_FILM_GRAIN) { 1190 ret = export_film_grain(avctx, frame); 1191 if (ret < 0) { 1192 av_frame_unref(frame); 1193 return ret; 1194 } 1195 } 1196 1197 frame->pts = pkt->pts; 1198 frame->pkt_dts = pkt->dts; 1199 #if FF_API_FRAME_PKT 1200 FF_DISABLE_DEPRECATION_WARNINGS 1201 frame->pkt_size = pkt->size; 1202 frame->pkt_pos = pkt->pos; 1203 FF_ENABLE_DEPRECATION_WARNINGS 1204 #endif 1205 1206 av_packet_unref(pkt); 1207 1208 return 0; 1209 } 1210 1211 static void update_reference_list(AVCodecContext *avctx) 1212 { 1213 AV1DecContext *s = avctx->priv_data; 1214 const AV1RawFrameHeader *header = s->raw_frame_header; 1215 1216 for (int i = 0; i < AV1_NUM_REF_FRAMES; i++) { 1217 if (header->refresh_frame_flags & (1 << i)) 1218 av1_frame_replace(&s->ref[i], &s->cur_frame); 1219 } 1220 } 1221 1222 static int get_current_frame(AVCodecContext *avctx) 1223 { 1224 AV1DecContext *s = avctx->priv_data; 1225 int ret; 1226 1227 av1_frame_unref(&s->cur_frame); 1228 1229 s->cur_frame.header_ref = av_refstruct_ref(s->header_ref); 1230 1231 s->cur_frame.raw_frame_header = s->raw_frame_header; 1232 1233 ret = init_tile_data(s); 1234 if (ret < 0) { 1235 av_log(avctx, AV_LOG_ERROR, "Failed to init tile data.\n"); 1236 return ret; 1237 } 1238 1239 if ((avctx->skip_frame >= AVDISCARD_NONINTRA && 1240 (s->raw_frame_header->frame_type != AV1_FRAME_KEY && 1241 s->raw_frame_header->frame_type != AV1_FRAME_INTRA_ONLY)) || 1242 (avctx->skip_frame >= AVDISCARD_NONKEY && 1243 s->raw_frame_header->frame_type != AV1_FRAME_KEY) || 1244 avctx->skip_frame >= AVDISCARD_ALL) 1245 return 0; 1246 1247 if (s->pix_fmt == AV_PIX_FMT_NONE) { 1248 ret = get_pixel_format(avctx); 1249 if (ret < 0) { 1250 av_log(avctx, AV_LOG_ERROR, "Failed to get pixel format.\n"); 1251 return ret; 1252 } 1253 1254 if (!ret && FF_HW_HAS_CB(avctx, decode_params)) { 1255 ret = FF_HW_CALL(avctx, decode_params, AV1_OBU_SEQUENCE_HEADER, 1256 s->seq_data_ref->data, s->seq_data_ref->size); 1257 if (ret < 0) { 1258 av_log(avctx, AV_LOG_ERROR, "HW accel decode params fail.\n"); 1259 return ret; 1260 } 1261 } 1262 } 1263 1264 ret = av1_frame_alloc(avctx, &s->cur_frame); 1265 if (ret < 0) { 1266 av_log(avctx, AV_LOG_ERROR, 1267 "Failed to allocate space for current frame.\n"); 1268 return ret; 1269 } 1270 1271 global_motion_params(s); 1272 skip_mode_params(s); 1273 coded_lossless_param(s); 1274 order_hint_info(s); 1275 load_grain_params(s); 1276 1277 s->cur_frame.force_integer_mv = 1278 s->raw_frame_header->force_integer_mv || 1279 s->raw_frame_header->frame_type == AV1_FRAME_KEY || 1280 s->raw_frame_header->frame_type == AV1_FRAME_INTRA_ONLY; 1281 1282 return ret; 1283 } 1284 1285 static int av1_receive_frame_internal(AVCodecContext *avctx, AVFrame *frame) 1286 { 1287 AV1DecContext *s = avctx->priv_data; 1288 AV1RawTileGroup *raw_tile_group = NULL; 1289 int i = 0, ret; 1290 1291 for (i = s->nb_unit; i < s->current_obu.nb_units; i++) { 1292 CodedBitstreamUnit *unit = &s->current_obu.units[i]; 1293 AV1RawOBU *obu = unit->content; 1294 const AV1RawOBUHeader *header; 1295 1296 av_log(avctx, AV_LOG_DEBUG, "OBU idx:%d, type:%d, content available:%d.\n", i, unit->type, !!obu); 1297 1298 if (unit->type == AV1_OBU_TILE_LIST) { 1299 av_log(avctx, AV_LOG_ERROR, "Large scale tile decoding is unsupported.\n"); 1300 ret = AVERROR_PATCHWELCOME; 1301 goto end; 1302 } 1303 1304 if (!obu) 1305 continue; 1306 1307 header = &obu->header; 1308 1309 switch (unit->type) { 1310 case AV1_OBU_SEQUENCE_HEADER: 1311 ret = av_buffer_replace(&s->seq_data_ref, unit->data_ref); 1312 if (ret < 0) 1313 goto end; 1314 1315 s->seq_data_ref->data = unit->data; 1316 s->seq_data_ref->size = unit->data_size; 1317 av_refstruct_replace(&s->seq_ref, unit->content_ref); 1318 1319 s->raw_seq = &obu->obu.sequence_header; 1320 1321 ret = set_context_with_sequence(avctx, s->raw_seq); 1322 if (ret < 0) { 1323 av_log(avctx, AV_LOG_ERROR, "Failed to set context.\n"); 1324 s->raw_seq = NULL; 1325 goto end; 1326 } 1327 1328 s->operating_point_idc = s->raw_seq->operating_point_idc[s->operating_point]; 1329 1330 s->pix_fmt = AV_PIX_FMT_NONE; 1331 1332 break; 1333 case AV1_OBU_REDUNDANT_FRAME_HEADER: 1334 if (s->raw_frame_header) 1335 break; 1336 // fall-through 1337 case AV1_OBU_FRAME: 1338 case AV1_OBU_FRAME_HEADER: 1339 if (!s->raw_seq) { 1340 av_log(avctx, AV_LOG_ERROR, "Missing Sequence Header.\n"); 1341 ret = AVERROR_INVALIDDATA; 1342 goto end; 1343 } 1344 1345 av_refstruct_replace(&s->header_ref, unit->content_ref); 1346 1347 if (unit->type == AV1_OBU_FRAME) 1348 s->raw_frame_header = &obu->obu.frame.header; 1349 else 1350 s->raw_frame_header = &obu->obu.frame_header; 1351 1352 if (s->raw_frame_header->show_existing_frame) { 1353 av1_frame_replace(&s->cur_frame, 1354 &s->ref[s->raw_frame_header->frame_to_show_map_idx]); 1355 1356 update_reference_list(avctx); 1357 1358 if (s->cur_frame.f) { 1359 ret = set_output_frame(avctx, frame); 1360 if (ret < 0) { 1361 av_log(avctx, AV_LOG_ERROR, "Set output frame error.\n"); 1362 goto end; 1363 } 1364 } 1365 1366 s->raw_frame_header = NULL; 1367 i++; 1368 ret = 0; 1369 1370 goto end; 1371 } 1372 1373 ret = get_current_frame(avctx); 1374 if (ret < 0) { 1375 av_log(avctx, AV_LOG_ERROR, "Get current frame error\n"); 1376 goto end; 1377 } 1378 1379 s->cur_frame.spatial_id = header->spatial_id; 1380 s->cur_frame.temporal_id = header->temporal_id; 1381 1382 if (avctx->hwaccel && s->cur_frame.f) { 1383 ret = FF_HW_CALL(avctx, start_frame, unit->data, unit->data_size); 1384 if (ret < 0) { 1385 av_log(avctx, AV_LOG_ERROR, "HW accel start frame fail.\n"); 1386 goto end; 1387 } 1388 } 1389 if (unit->type != AV1_OBU_FRAME) 1390 break; 1391 // fall-through 1392 case AV1_OBU_TILE_GROUP: 1393 if (!s->raw_frame_header) { 1394 av_log(avctx, AV_LOG_ERROR, "Missing Frame Header.\n"); 1395 ret = AVERROR_INVALIDDATA; 1396 goto end; 1397 } 1398 1399 if (unit->type == AV1_OBU_FRAME) 1400 raw_tile_group = &obu->obu.frame.tile_group; 1401 else 1402 raw_tile_group = &obu->obu.tile_group; 1403 1404 ret = get_tiles_info(avctx, raw_tile_group); 1405 if (ret < 0) 1406 goto end; 1407 1408 if (avctx->hwaccel && s->cur_frame.f) { 1409 ret = FF_HW_CALL(avctx, decode_slice, raw_tile_group->tile_data.data, 1410 raw_tile_group->tile_data.data_size); 1411 if (ret < 0) { 1412 av_log(avctx, AV_LOG_ERROR, 1413 "HW accel decode slice fail.\n"); 1414 goto end; 1415 } 1416 } 1417 break; 1418 case AV1_OBU_TILE_LIST: 1419 case AV1_OBU_TEMPORAL_DELIMITER: 1420 case AV1_OBU_PADDING: 1421 break; 1422 case AV1_OBU_METADATA: 1423 switch (obu->obu.metadata.metadata_type) { 1424 case AV1_METADATA_TYPE_HDR_CLL: 1425 av_refstruct_replace(&s->cll_ref, unit->content_ref); 1426 s->cll = &obu->obu.metadata.metadata.hdr_cll; 1427 break; 1428 case AV1_METADATA_TYPE_HDR_MDCV: 1429 av_refstruct_replace(&s->mdcv_ref, unit->content_ref); 1430 s->mdcv = &obu->obu.metadata.metadata.hdr_mdcv; 1431 break; 1432 case AV1_METADATA_TYPE_ITUT_T35: { 1433 AV1RawMetadataITUTT35 itut_t35; 1434 memcpy(&itut_t35, &obu->obu.metadata.metadata.itut_t35, sizeof(itut_t35)); 1435 itut_t35.payload_ref = av_buffer_ref(obu->obu.metadata.metadata.itut_t35.payload_ref); 1436 if (!itut_t35.payload_ref) { 1437 ret = AVERROR(ENOMEM); 1438 goto end; 1439 } 1440 ret = av_fifo_write(s->itut_t35_fifo, &itut_t35, 1); 1441 if (ret < 0) { 1442 av_buffer_unref(&itut_t35.payload_ref); 1443 goto end; 1444 } 1445 break; 1446 } 1447 default: 1448 break; 1449 } 1450 break; 1451 default: 1452 av_log(avctx, AV_LOG_DEBUG, 1453 "Unknown obu type: %d (%"SIZE_SPECIFIER" bits).\n", 1454 unit->type, unit->data_size); 1455 } 1456 1457 if (raw_tile_group && (s->tile_num == raw_tile_group->tg_end + 1)) { 1458 int show_frame = s->raw_frame_header->show_frame; 1459 // Set nb_unit to point at the next OBU, to indicate which 1460 // OBUs have been processed for this current frame. (If this 1461 // frame gets output, we set nb_unit to this value later too.) 1462 s->nb_unit = i + 1; 1463 if (avctx->hwaccel && s->cur_frame.f) { 1464 ret = FF_HW_SIMPLE_CALL(avctx, end_frame); 1465 if (ret < 0) { 1466 av_log(avctx, AV_LOG_ERROR, "HW accel end frame fail.\n"); 1467 goto end; 1468 } 1469 } 1470 1471 update_reference_list(avctx); 1472 1473 // Set start_unit to indicate the first OBU of the next frame. 1474 s->start_unit = s->nb_unit; 1475 raw_tile_group = NULL; 1476 s->raw_frame_header = NULL; 1477 1478 if (show_frame) { 1479 // cur_frame.f needn't exist due to skip_frame. 1480 if (s->cur_frame.f) { 1481 ret = set_output_frame(avctx, frame); 1482 if (ret < 0) { 1483 av_log(avctx, AV_LOG_ERROR, "Set output frame error\n"); 1484 goto end; 1485 } 1486 } 1487 i++; 1488 ret = 0; 1489 goto end; 1490 } 1491 } 1492 } 1493 1494 ret = AVERROR(EAGAIN); 1495 end: 1496 av_assert0(i <= s->current_obu.nb_units); 1497 s->nb_unit = i; 1498 1499 if ((ret < 0 && ret != AVERROR(EAGAIN)) || s->current_obu.nb_units == i) { 1500 if (ret < 0) 1501 s->raw_frame_header = NULL; 1502 av_packet_unref(s->pkt); 1503 ff_cbs_fragment_reset(&s->current_obu); 1504 s->nb_unit = s->start_unit = 0; 1505 } 1506 if (!ret && !frame->buf[0]) 1507 ret = AVERROR(EAGAIN); 1508 1509 return ret; 1510 } 1511 1512 static int av1_receive_frame(AVCodecContext *avctx, AVFrame *frame) 1513 { 1514 AV1DecContext *s = avctx->priv_data; 1515 int ret; 1516 1517 do { 1518 if (!s->current_obu.nb_units) { 1519 ret = ff_decode_get_packet(avctx, s->pkt); 1520 if (ret < 0) 1521 return ret; 1522 1523 ret = ff_cbs_read_packet(s->cbc, &s->current_obu, s->pkt); 1524 if (ret < 0) { 1525 ff_cbs_fragment_reset(&s->current_obu); 1526 av_packet_unref(s->pkt); 1527 av_log(avctx, AV_LOG_ERROR, "Failed to read packet.\n"); 1528 return ret; 1529 } 1530 1531 s->nb_unit = s->start_unit = 0; 1532 av_log(avctx, AV_LOG_DEBUG, "Total OBUs on this packet: %d.\n", 1533 s->current_obu.nb_units); 1534 } 1535 1536 ret = av1_receive_frame_internal(avctx, frame); 1537 } while (ret == AVERROR(EAGAIN)); 1538 1539 return ret; 1540 } 1541 1542 static void av1_decode_flush(AVCodecContext *avctx) 1543 { 1544 AV1DecContext *s = avctx->priv_data; 1545 AV1RawMetadataITUTT35 itut_t35; 1546 1547 for (int i = 0; i < FF_ARRAY_ELEMS(s->ref); i++) 1548 av1_frame_unref(&s->ref[i]); 1549 1550 av1_frame_unref(&s->cur_frame); 1551 s->operating_point_idc = 0; 1552 s->nb_unit = s->start_unit = 0; 1553 s->raw_frame_header = NULL; 1554 s->raw_seq = NULL; 1555 s->cll = NULL; 1556 s->mdcv = NULL; 1557 while (av_fifo_read(s->itut_t35_fifo, &itut_t35, 1) >= 0) 1558 av_buffer_unref(&itut_t35.payload_ref); 1559 1560 ff_cbs_fragment_reset(&s->current_obu); 1561 ff_cbs_flush(s->cbc); 1562 1563 if (FF_HW_HAS_CB(avctx, flush)) 1564 FF_HW_SIMPLE_CALL(avctx, flush); 1565 } 1566 1567 #define OFFSET(x) offsetof(AV1DecContext, x) 1568 #define VD AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_DECODING_PARAM 1569 static const AVOption av1_options[] = { 1570 { "operating_point", "Select an operating point of the scalable bitstream", 1571 OFFSET(operating_point), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, AV1_MAX_OPERATING_POINTS - 1, VD }, 1572 { NULL } 1573 }; 1574 1575 static const AVClass av1_class = { 1576 .class_name = "AV1 decoder", 1577 .item_name = av_default_item_name, 1578 .option = av1_options, 1579 .version = LIBAVUTIL_VERSION_INT, 1580 }; 1581 1582 const FFCodec ff_av1_decoder = { 1583 .p.name = "av1", 1584 CODEC_LONG_NAME("Alliance for Open Media AV1"), 1585 .p.type = AVMEDIA_TYPE_VIDEO, 1586 .p.id = AV_CODEC_ID_AV1, 1587 .priv_data_size = sizeof(AV1DecContext), 1588 .init = av1_decode_init, 1589 .close = av1_decode_free, 1590 FF_CODEC_RECEIVE_FRAME_CB(av1_receive_frame), 1591 .p.capabilities = AV_CODEC_CAP_DR1, 1592 .caps_internal = FF_CODEC_CAP_INIT_CLEANUP | 1593 FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM | 1594 FF_CODEC_CAP_USES_PROGRESSFRAMES, 1595 .flush = av1_decode_flush, 1596 .p.profiles = NULL_IF_CONFIG_SMALL(ff_av1_profiles), 1597 .p.priv_class = &av1_class, 1598 .hw_configs = (const AVCodecHWConfigInternal *const []) { 1599 #if CONFIG_AV1_DXVA2_HWACCEL 1600 HWACCEL_DXVA2(av1), 1601 #endif 1602 #if CONFIG_AV1_D3D11VA_HWACCEL 1603 HWACCEL_D3D11VA(av1), 1604 #endif 1605 #if CONFIG_AV1_D3D11VA2_HWACCEL 1606 HWACCEL_D3D11VA2(av1), 1607 #endif 1608 #if CONFIG_AV1_D3D12VA_HWACCEL 1609 HWACCEL_D3D12VA(av1), 1610 #endif 1611 #if CONFIG_AV1_NVDEC_HWACCEL 1612 HWACCEL_NVDEC(av1), 1613 #endif 1614 #if CONFIG_AV1_VAAPI_HWACCEL 1615 HWACCEL_VAAPI(av1), 1616 #endif 1617 #if CONFIG_AV1_VDPAU_HWACCEL 1618 HWACCEL_VDPAU(av1), 1619 #endif 1620 #if CONFIG_AV1_VIDEOTOOLBOX_HWACCEL 1621 HWACCEL_VIDEOTOOLBOX(av1), 1622 #endif 1623 #if CONFIG_AV1_VULKAN_HWACCEL 1624 HWACCEL_VULKAN(av1), 1625 #endif 1626 1627 NULL 1628 }, 1629 };