| 1 | /* |
| 2 | * Opus decoder |
| 3 | * Copyright (c) 2012 Andrew D'Addesio |
| 4 | * Copyright (c) 2013-2014 Mozilla Corporation |
| 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 | /** |
| 24 | * @file |
| 25 | * Opus decoder |
| 26 | * @author Andrew D'Addesio, Anton Khirnov |
| 27 | * |
| 28 | * Codec homepage: http://opus-codec.org/ |
| 29 | * Specification: http://tools.ietf.org/html/rfc6716 |
| 30 | * Ogg Opus specification: https://tools.ietf.org/html/draft-ietf-codec-oggopus-03 |
| 31 | * |
| 32 | * Ogg-contained .opus files can be produced with opus-tools: |
| 33 | * http://git.xiph.org/?p=opus-tools.git |
| 34 | */ |
| 35 | |
| 36 | #include <stdint.h> |
| 37 | |
| 38 | #include "libavutil/attributes.h" |
| 39 | #include "libavutil/audio_fifo.h" |
| 40 | #include "libavutil/channel_layout.h" |
| 41 | #include "libavutil/opt.h" |
| 42 | |
| 43 | #include "libswresample/swresample.h" |
| 44 | |
| 45 | #include "avcodec.h" |
| 46 | #include "celp_filters.h" |
| 47 | #include "fft.h" |
| 48 | #include "get_bits.h" |
| 49 | #include "internal.h" |
| 50 | #include "mathops.h" |
| 51 | #include "opus.h" |
| 52 | |
| 53 | static const uint16_t silk_frame_duration_ms[16] = { |
| 54 | 10, 20, 40, 60, |
| 55 | 10, 20, 40, 60, |
| 56 | 10, 20, 40, 60, |
| 57 | 10, 20, |
| 58 | 10, 20, |
| 59 | }; |
| 60 | |
| 61 | /* number of samples of silence to feed to the resampler |
| 62 | * at the beginning */ |
| 63 | static const int silk_resample_delay[] = { |
| 64 | 4, 8, 11, 11, 11 |
| 65 | }; |
| 66 | |
| 67 | static const uint8_t celt_band_end[] = { 13, 17, 17, 19, 21 }; |
| 68 | |
| 69 | static int get_silk_samplerate(int config) |
| 70 | { |
| 71 | if (config < 4) |
| 72 | return 8000; |
| 73 | else if (config < 8) |
| 74 | return 12000; |
| 75 | return 16000; |
| 76 | } |
| 77 | |
| 78 | /** |
| 79 | * Range decoder |
| 80 | */ |
| 81 | static int opus_rc_init(OpusRangeCoder *rc, const uint8_t *data, int size) |
| 82 | { |
| 83 | int ret = init_get_bits8(&rc->gb, data, size); |
| 84 | if (ret < 0) |
| 85 | return ret; |
| 86 | |
| 87 | rc->range = 128; |
| 88 | rc->value = 127 - get_bits(&rc->gb, 7); |
| 89 | rc->total_read_bits = 9; |
| 90 | opus_rc_normalize(rc); |
| 91 | |
| 92 | return 0; |
| 93 | } |
| 94 | |
| 95 | static void opus_raw_init(OpusRangeCoder *rc, const uint8_t *rightend, |
| 96 | unsigned int bytes) |
| 97 | { |
| 98 | rc->rb.position = rightend; |
| 99 | rc->rb.bytes = bytes; |
| 100 | rc->rb.cachelen = 0; |
| 101 | rc->rb.cacheval = 0; |
| 102 | } |
| 103 | |
| 104 | static void opus_fade(float *out, |
| 105 | const float *in1, const float *in2, |
| 106 | const float *window, int len) |
| 107 | { |
| 108 | int i; |
| 109 | for (i = 0; i < len; i++) |
| 110 | out[i] = in2[i] * window[i] + in1[i] * (1.0 - window[i]); |
| 111 | } |
| 112 | |
| 113 | static int opus_flush_resample(OpusStreamContext *s, int nb_samples) |
| 114 | { |
| 115 | int celt_size = av_audio_fifo_size(s->celt_delay); |
| 116 | int ret, i; |
| 117 | ret = swr_convert(s->swr, |
| 118 | (uint8_t**)s->out, nb_samples, |
| 119 | NULL, 0); |
| 120 | if (ret < 0) |
| 121 | return ret; |
| 122 | else if (ret != nb_samples) { |
| 123 | av_log(s->avctx, AV_LOG_ERROR, "Wrong number of flushed samples: %d\n", |
| 124 | ret); |
| 125 | return AVERROR_BUG; |
| 126 | } |
| 127 | |
| 128 | if (celt_size) { |
| 129 | if (celt_size != nb_samples) { |
| 130 | av_log(s->avctx, AV_LOG_ERROR, "Wrong number of CELT delay samples.\n"); |
| 131 | return AVERROR_BUG; |
| 132 | } |
| 133 | av_audio_fifo_read(s->celt_delay, (void**)s->celt_output, nb_samples); |
| 134 | for (i = 0; i < s->output_channels; i++) { |
| 135 | s->fdsp->vector_fmac_scalar(s->out[i], |
| 136 | s->celt_output[i], 1.0, |
| 137 | nb_samples); |
| 138 | } |
| 139 | } |
| 140 | |
| 141 | if (s->redundancy_idx) { |
| 142 | for (i = 0; i < s->output_channels; i++) |
| 143 | opus_fade(s->out[i], s->out[i], |
| 144 | s->redundancy_output[i] + 120 + s->redundancy_idx, |
| 145 | ff_celt_window2 + s->redundancy_idx, 120 - s->redundancy_idx); |
| 146 | s->redundancy_idx = 0; |
| 147 | } |
| 148 | |
| 149 | s->out[0] += nb_samples; |
| 150 | s->out[1] += nb_samples; |
| 151 | s->out_size -= nb_samples * sizeof(float); |
| 152 | |
| 153 | return 0; |
| 154 | } |
| 155 | |
| 156 | static int opus_init_resample(OpusStreamContext *s) |
| 157 | { |
| 158 | static const float delay[16] = { 0.0 }; |
| 159 | const uint8_t *delayptr[2] = { (uint8_t*)delay, (uint8_t*)delay }; |
| 160 | int ret; |
| 161 | |
| 162 | av_opt_set_int(s->swr, "in_sample_rate", s->silk_samplerate, 0); |
| 163 | ret = swr_init(s->swr); |
| 164 | if (ret < 0) { |
| 165 | av_log(s->avctx, AV_LOG_ERROR, "Error opening the resampler.\n"); |
| 166 | return ret; |
| 167 | } |
| 168 | |
| 169 | ret = swr_convert(s->swr, |
| 170 | NULL, 0, |
| 171 | delayptr, silk_resample_delay[s->packet.bandwidth]); |
| 172 | if (ret < 0) { |
| 173 | av_log(s->avctx, AV_LOG_ERROR, |
| 174 | "Error feeding initial silence to the resampler.\n"); |
| 175 | return ret; |
| 176 | } |
| 177 | |
| 178 | return 0; |
| 179 | } |
| 180 | |
| 181 | static int opus_decode_redundancy(OpusStreamContext *s, const uint8_t *data, int size) |
| 182 | { |
| 183 | int ret; |
| 184 | enum OpusBandwidth bw = s->packet.bandwidth; |
| 185 | |
| 186 | if (s->packet.mode == OPUS_MODE_SILK && |
| 187 | bw == OPUS_BANDWIDTH_MEDIUMBAND) |
| 188 | bw = OPUS_BANDWIDTH_WIDEBAND; |
| 189 | |
| 190 | ret = opus_rc_init(&s->redundancy_rc, data, size); |
| 191 | if (ret < 0) |
| 192 | goto fail; |
| 193 | opus_raw_init(&s->redundancy_rc, data + size, size); |
| 194 | |
| 195 | ret = ff_celt_decode_frame(s->celt, &s->redundancy_rc, |
| 196 | s->redundancy_output, |
| 197 | s->packet.stereo + 1, 240, |
| 198 | 0, celt_band_end[s->packet.bandwidth]); |
| 199 | if (ret < 0) |
| 200 | goto fail; |
| 201 | |
| 202 | return 0; |
| 203 | fail: |
| 204 | av_log(s->avctx, AV_LOG_ERROR, "Error decoding the redundancy frame.\n"); |
| 205 | return ret; |
| 206 | } |
| 207 | |
| 208 | static int opus_decode_frame(OpusStreamContext *s, const uint8_t *data, int size) |
| 209 | { |
| 210 | int samples = s->packet.frame_duration; |
| 211 | int redundancy = 0; |
| 212 | int redundancy_size, redundancy_pos; |
| 213 | int ret, i, consumed; |
| 214 | int delayed_samples = s->delayed_samples; |
| 215 | |
| 216 | ret = opus_rc_init(&s->rc, data, size); |
| 217 | if (ret < 0) |
| 218 | return ret; |
| 219 | |
| 220 | /* decode the silk frame */ |
| 221 | if (s->packet.mode == OPUS_MODE_SILK || s->packet.mode == OPUS_MODE_HYBRID) { |
| 222 | if (!swr_is_initialized(s->swr)) { |
| 223 | ret = opus_init_resample(s); |
| 224 | if (ret < 0) |
| 225 | return ret; |
| 226 | } |
| 227 | |
| 228 | samples = ff_silk_decode_superframe(s->silk, &s->rc, s->silk_output, |
| 229 | FFMIN(s->packet.bandwidth, OPUS_BANDWIDTH_WIDEBAND), |
| 230 | s->packet.stereo + 1, |
| 231 | silk_frame_duration_ms[s->packet.config]); |
| 232 | if (samples < 0) { |
| 233 | av_log(s->avctx, AV_LOG_ERROR, "Error decoding a SILK frame.\n"); |
| 234 | return samples; |
| 235 | } |
| 236 | samples = swr_convert(s->swr, |
| 237 | (uint8_t**)s->out, s->packet.frame_duration, |
| 238 | (const uint8_t**)s->silk_output, samples); |
| 239 | if (samples < 0) { |
| 240 | av_log(s->avctx, AV_LOG_ERROR, "Error resampling SILK data.\n"); |
| 241 | return samples; |
| 242 | } |
| 243 | av_assert2((samples & 7) == 0); |
| 244 | s->delayed_samples += s->packet.frame_duration - samples; |
| 245 | } else |
| 246 | ff_silk_flush(s->silk); |
| 247 | |
| 248 | // decode redundancy information |
| 249 | consumed = opus_rc_tell(&s->rc); |
| 250 | if (s->packet.mode == OPUS_MODE_HYBRID && consumed + 37 <= size * 8) |
| 251 | redundancy = opus_rc_p2model(&s->rc, 12); |
| 252 | else if (s->packet.mode == OPUS_MODE_SILK && consumed + 17 <= size * 8) |
| 253 | redundancy = 1; |
| 254 | |
| 255 | if (redundancy) { |
| 256 | redundancy_pos = opus_rc_p2model(&s->rc, 1); |
| 257 | |
| 258 | if (s->packet.mode == OPUS_MODE_HYBRID) |
| 259 | redundancy_size = opus_rc_unimodel(&s->rc, 256) + 2; |
| 260 | else |
| 261 | redundancy_size = size - (consumed + 7) / 8; |
| 262 | size -= redundancy_size; |
| 263 | if (size < 0) { |
| 264 | av_log(s->avctx, AV_LOG_ERROR, "Invalid redundancy frame size.\n"); |
| 265 | return AVERROR_INVALIDDATA; |
| 266 | } |
| 267 | |
| 268 | if (redundancy_pos) { |
| 269 | ret = opus_decode_redundancy(s, data + size, redundancy_size); |
| 270 | if (ret < 0) |
| 271 | return ret; |
| 272 | ff_celt_flush(s->celt); |
| 273 | } |
| 274 | } |
| 275 | |
| 276 | /* decode the CELT frame */ |
| 277 | if (s->packet.mode == OPUS_MODE_CELT || s->packet.mode == OPUS_MODE_HYBRID) { |
| 278 | float *out_tmp[2] = { s->out[0], s->out[1] }; |
| 279 | float **dst = (s->packet.mode == OPUS_MODE_CELT) ? |
| 280 | out_tmp : s->celt_output; |
| 281 | int celt_output_samples = samples; |
| 282 | int delay_samples = av_audio_fifo_size(s->celt_delay); |
| 283 | |
| 284 | if (delay_samples) { |
| 285 | if (s->packet.mode == OPUS_MODE_HYBRID) { |
| 286 | av_audio_fifo_read(s->celt_delay, (void**)s->celt_output, delay_samples); |
| 287 | |
| 288 | for (i = 0; i < s->output_channels; i++) { |
| 289 | s->fdsp->vector_fmac_scalar(out_tmp[i], s->celt_output[i], 1.0, |
| 290 | delay_samples); |
| 291 | out_tmp[i] += delay_samples; |
| 292 | } |
| 293 | celt_output_samples -= delay_samples; |
| 294 | } else { |
| 295 | av_log(s->avctx, AV_LOG_WARNING, |
| 296 | "Spurious CELT delay samples present.\n"); |
| 297 | av_audio_fifo_drain(s->celt_delay, delay_samples); |
| 298 | if (s->avctx->err_recognition & AV_EF_EXPLODE) |
| 299 | return AVERROR_BUG; |
| 300 | } |
| 301 | } |
| 302 | |
| 303 | opus_raw_init(&s->rc, data + size, size); |
| 304 | |
| 305 | ret = ff_celt_decode_frame(s->celt, &s->rc, dst, |
| 306 | s->packet.stereo + 1, |
| 307 | s->packet.frame_duration, |
| 308 | (s->packet.mode == OPUS_MODE_HYBRID) ? 17 : 0, |
| 309 | celt_band_end[s->packet.bandwidth]); |
| 310 | if (ret < 0) |
| 311 | return ret; |
| 312 | |
| 313 | if (s->packet.mode == OPUS_MODE_HYBRID) { |
| 314 | int celt_delay = s->packet.frame_duration - celt_output_samples; |
| 315 | void *delaybuf[2] = { s->celt_output[0] + celt_output_samples, |
| 316 | s->celt_output[1] + celt_output_samples }; |
| 317 | |
| 318 | for (i = 0; i < s->output_channels; i++) { |
| 319 | s->fdsp->vector_fmac_scalar(out_tmp[i], |
| 320 | s->celt_output[i], 1.0, |
| 321 | celt_output_samples); |
| 322 | } |
| 323 | |
| 324 | ret = av_audio_fifo_write(s->celt_delay, delaybuf, celt_delay); |
| 325 | if (ret < 0) |
| 326 | return ret; |
| 327 | } |
| 328 | } else |
| 329 | ff_celt_flush(s->celt); |
| 330 | |
| 331 | if (s->redundancy_idx) { |
| 332 | for (i = 0; i < s->output_channels; i++) |
| 333 | opus_fade(s->out[i], s->out[i], |
| 334 | s->redundancy_output[i] + 120 + s->redundancy_idx, |
| 335 | ff_celt_window2 + s->redundancy_idx, 120 - s->redundancy_idx); |
| 336 | s->redundancy_idx = 0; |
| 337 | } |
| 338 | if (redundancy) { |
| 339 | if (!redundancy_pos) { |
| 340 | ff_celt_flush(s->celt); |
| 341 | ret = opus_decode_redundancy(s, data + size, redundancy_size); |
| 342 | if (ret < 0) |
| 343 | return ret; |
| 344 | |
| 345 | for (i = 0; i < s->output_channels; i++) { |
| 346 | opus_fade(s->out[i] + samples - 120 + delayed_samples, |
| 347 | s->out[i] + samples - 120 + delayed_samples, |
| 348 | s->redundancy_output[i] + 120, |
| 349 | ff_celt_window2, 120 - delayed_samples); |
| 350 | if (delayed_samples) |
| 351 | s->redundancy_idx = 120 - delayed_samples; |
| 352 | } |
| 353 | } else { |
| 354 | for (i = 0; i < s->output_channels; i++) { |
| 355 | memcpy(s->out[i] + delayed_samples, s->redundancy_output[i], 120 * sizeof(float)); |
| 356 | opus_fade(s->out[i] + 120 + delayed_samples, |
| 357 | s->redundancy_output[i] + 120, |
| 358 | s->out[i] + 120 + delayed_samples, |
| 359 | ff_celt_window2, 120); |
| 360 | } |
| 361 | } |
| 362 | } |
| 363 | |
| 364 | return samples; |
| 365 | } |
| 366 | |
| 367 | static int opus_decode_subpacket(OpusStreamContext *s, |
| 368 | const uint8_t *buf, int buf_size, |
| 369 | int nb_samples) |
| 370 | { |
| 371 | int output_samples = 0; |
| 372 | int flush_needed = 0; |
| 373 | int i, j, ret; |
| 374 | |
| 375 | /* check if we need to flush the resampler */ |
| 376 | if (swr_is_initialized(s->swr)) { |
| 377 | if (buf) { |
| 378 | int64_t cur_samplerate; |
| 379 | av_opt_get_int(s->swr, "in_sample_rate", 0, &cur_samplerate); |
| 380 | flush_needed = (s->packet.mode == OPUS_MODE_CELT) || (cur_samplerate != s->silk_samplerate); |
| 381 | } else { |
| 382 | flush_needed = !!s->delayed_samples; |
| 383 | } |
| 384 | } |
| 385 | |
| 386 | if (!buf && !flush_needed) |
| 387 | return 0; |
| 388 | |
| 389 | /* use dummy output buffers if the channel is not mapped to anything */ |
| 390 | if (!s->out[0] || |
| 391 | (s->output_channels == 2 && !s->out[1])) { |
| 392 | av_fast_malloc(&s->out_dummy, &s->out_dummy_allocated_size, s->out_size); |
| 393 | if (!s->out_dummy) |
| 394 | return AVERROR(ENOMEM); |
| 395 | if (!s->out[0]) |
| 396 | s->out[0] = s->out_dummy; |
| 397 | if (!s->out[1]) |
| 398 | s->out[1] = s->out_dummy; |
| 399 | } |
| 400 | |
| 401 | /* flush the resampler if necessary */ |
| 402 | if (flush_needed) { |
| 403 | ret = opus_flush_resample(s, s->delayed_samples); |
| 404 | if (ret < 0) { |
| 405 | av_log(s->avctx, AV_LOG_ERROR, "Error flushing the resampler.\n"); |
| 406 | return ret; |
| 407 | } |
| 408 | swr_close(s->swr); |
| 409 | output_samples += s->delayed_samples; |
| 410 | s->delayed_samples = 0; |
| 411 | |
| 412 | if (!buf) |
| 413 | goto finish; |
| 414 | } |
| 415 | |
| 416 | /* decode all the frames in the packet */ |
| 417 | for (i = 0; i < s->packet.frame_count; i++) { |
| 418 | int size = s->packet.frame_size[i]; |
| 419 | int samples = opus_decode_frame(s, buf + s->packet.frame_offset[i], size); |
| 420 | |
| 421 | if (samples < 0) { |
| 422 | av_log(s->avctx, AV_LOG_ERROR, "Error decoding an Opus frame.\n"); |
| 423 | if (s->avctx->err_recognition & AV_EF_EXPLODE) |
| 424 | return samples; |
| 425 | |
| 426 | for (j = 0; j < s->output_channels; j++) |
| 427 | memset(s->out[j], 0, s->packet.frame_duration * sizeof(float)); |
| 428 | samples = s->packet.frame_duration; |
| 429 | } |
| 430 | output_samples += samples; |
| 431 | |
| 432 | for (j = 0; j < s->output_channels; j++) |
| 433 | s->out[j] += samples; |
| 434 | s->out_size -= samples * sizeof(float); |
| 435 | } |
| 436 | |
| 437 | finish: |
| 438 | s->out[0] = s->out[1] = NULL; |
| 439 | s->out_size = 0; |
| 440 | |
| 441 | return output_samples; |
| 442 | } |
| 443 | |
| 444 | static int opus_decode_packet(AVCodecContext *avctx, void *data, |
| 445 | int *got_frame_ptr, AVPacket *avpkt) |
| 446 | { |
| 447 | OpusContext *c = avctx->priv_data; |
| 448 | AVFrame *frame = data; |
| 449 | const uint8_t *buf = avpkt->data; |
| 450 | int buf_size = avpkt->size; |
| 451 | int coded_samples = 0; |
| 452 | int decoded_samples = 0; |
| 453 | int i, ret; |
| 454 | |
| 455 | /* decode the header of the first sub-packet to find out the sample count */ |
| 456 | if (buf) { |
| 457 | OpusPacket *pkt = &c->streams[0].packet; |
| 458 | ret = ff_opus_parse_packet(pkt, buf, buf_size, c->nb_streams > 1); |
| 459 | if (ret < 0) { |
| 460 | av_log(avctx, AV_LOG_ERROR, "Error parsing the packet header.\n"); |
| 461 | return ret; |
| 462 | } |
| 463 | coded_samples += pkt->frame_count * pkt->frame_duration; |
| 464 | c->streams[0].silk_samplerate = get_silk_samplerate(pkt->config); |
| 465 | } |
| 466 | |
| 467 | frame->nb_samples = coded_samples + c->streams[0].delayed_samples; |
| 468 | |
| 469 | /* no input or buffered data => nothing to do */ |
| 470 | if (!frame->nb_samples) { |
| 471 | *got_frame_ptr = 0; |
| 472 | return 0; |
| 473 | } |
| 474 | |
| 475 | /* setup the data buffers */ |
| 476 | ret = ff_get_buffer(avctx, frame, 0); |
| 477 | if (ret < 0) { |
| 478 | av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n"); |
| 479 | return ret; |
| 480 | } |
| 481 | frame->nb_samples = 0; |
| 482 | |
| 483 | for (i = 0; i < avctx->channels; i++) { |
| 484 | ChannelMap *map = &c->channel_maps[i]; |
| 485 | if (!map->copy) |
| 486 | c->streams[map->stream_idx].out[map->channel_idx] = (float*)frame->extended_data[i]; |
| 487 | } |
| 488 | |
| 489 | for (i = 0; i < c->nb_streams; i++) |
| 490 | c->streams[i].out_size = frame->linesize[0]; |
| 491 | |
| 492 | /* decode each sub-packet */ |
| 493 | for (i = 0; i < c->nb_streams; i++) { |
| 494 | OpusStreamContext *s = &c->streams[i]; |
| 495 | |
| 496 | if (i && buf) { |
| 497 | ret = ff_opus_parse_packet(&s->packet, buf, buf_size, i != c->nb_streams - 1); |
| 498 | if (ret < 0) { |
| 499 | av_log(avctx, AV_LOG_ERROR, "Error parsing the packet header.\n"); |
| 500 | return ret; |
| 501 | } |
| 502 | s->silk_samplerate = get_silk_samplerate(s->packet.config); |
| 503 | } |
| 504 | |
| 505 | ret = opus_decode_subpacket(&c->streams[i], buf, |
| 506 | s->packet.data_size, coded_samples); |
| 507 | if (ret < 0) |
| 508 | return ret; |
| 509 | if (decoded_samples && ret != decoded_samples) { |
| 510 | av_log(avctx, AV_LOG_ERROR, "Different numbers of decoded samples " |
| 511 | "in a multi-channel stream\n"); |
| 512 | return AVERROR_INVALIDDATA; |
| 513 | } |
| 514 | decoded_samples = ret; |
| 515 | buf += s->packet.packet_size; |
| 516 | buf_size -= s->packet.packet_size; |
| 517 | } |
| 518 | |
| 519 | for (i = 0; i < avctx->channels; i++) { |
| 520 | ChannelMap *map = &c->channel_maps[i]; |
| 521 | |
| 522 | /* handle copied channels */ |
| 523 | if (map->copy) { |
| 524 | memcpy(frame->extended_data[i], |
| 525 | frame->extended_data[map->copy_idx], |
| 526 | frame->linesize[0]); |
| 527 | } else if (map->silence) { |
| 528 | memset(frame->extended_data[i], 0, frame->linesize[0]); |
| 529 | } |
| 530 | |
| 531 | if (c->gain_i) { |
| 532 | c->fdsp.vector_fmul_scalar((float*)frame->extended_data[i], |
| 533 | (float*)frame->extended_data[i], |
| 534 | c->gain, FFALIGN(decoded_samples, 8)); |
| 535 | } |
| 536 | } |
| 537 | |
| 538 | frame->nb_samples = decoded_samples; |
| 539 | *got_frame_ptr = !!decoded_samples; |
| 540 | |
| 541 | return avpkt->size; |
| 542 | } |
| 543 | |
| 544 | static av_cold void opus_decode_flush(AVCodecContext *ctx) |
| 545 | { |
| 546 | OpusContext *c = ctx->priv_data; |
| 547 | int i; |
| 548 | |
| 549 | for (i = 0; i < c->nb_streams; i++) { |
| 550 | OpusStreamContext *s = &c->streams[i]; |
| 551 | |
| 552 | memset(&s->packet, 0, sizeof(s->packet)); |
| 553 | s->delayed_samples = 0; |
| 554 | |
| 555 | if (s->celt_delay) |
| 556 | av_audio_fifo_drain(s->celt_delay, av_audio_fifo_size(s->celt_delay)); |
| 557 | swr_close(s->swr); |
| 558 | |
| 559 | ff_silk_flush(s->silk); |
| 560 | ff_celt_flush(s->celt); |
| 561 | } |
| 562 | } |
| 563 | |
| 564 | static av_cold int opus_decode_close(AVCodecContext *avctx) |
| 565 | { |
| 566 | OpusContext *c = avctx->priv_data; |
| 567 | int i; |
| 568 | |
| 569 | for (i = 0; i < c->nb_streams; i++) { |
| 570 | OpusStreamContext *s = &c->streams[i]; |
| 571 | |
| 572 | ff_silk_free(&s->silk); |
| 573 | ff_celt_free(&s->celt); |
| 574 | |
| 575 | av_freep(&s->out_dummy); |
| 576 | s->out_dummy_allocated_size = 0; |
| 577 | |
| 578 | av_audio_fifo_free(s->celt_delay); |
| 579 | swr_free(&s->swr); |
| 580 | } |
| 581 | |
| 582 | av_freep(&c->streams); |
| 583 | c->nb_streams = 0; |
| 584 | |
| 585 | av_freep(&c->channel_maps); |
| 586 | |
| 587 | return 0; |
| 588 | } |
| 589 | |
| 590 | static av_cold int opus_decode_init(AVCodecContext *avctx) |
| 591 | { |
| 592 | OpusContext *c = avctx->priv_data; |
| 593 | int ret, i, j; |
| 594 | |
| 595 | avctx->sample_fmt = AV_SAMPLE_FMT_FLTP; |
| 596 | avctx->sample_rate = 48000; |
| 597 | |
| 598 | avpriv_float_dsp_init(&c->fdsp, 0); |
| 599 | |
| 600 | /* find out the channel configuration */ |
| 601 | ret = ff_opus_parse_extradata(avctx, c); |
| 602 | if (ret < 0) |
| 603 | return ret; |
| 604 | |
| 605 | /* allocate and init each independent decoder */ |
| 606 | c->streams = av_mallocz_array(c->nb_streams, sizeof(*c->streams)); |
| 607 | if (!c->streams) { |
| 608 | c->nb_streams = 0; |
| 609 | ret = AVERROR(ENOMEM); |
| 610 | goto fail; |
| 611 | } |
| 612 | |
| 613 | for (i = 0; i < c->nb_streams; i++) { |
| 614 | OpusStreamContext *s = &c->streams[i]; |
| 615 | uint64_t layout; |
| 616 | |
| 617 | s->output_channels = (i < c->nb_stereo_streams) ? 2 : 1; |
| 618 | |
| 619 | s->avctx = avctx; |
| 620 | |
| 621 | for (j = 0; j < s->output_channels; j++) { |
| 622 | s->silk_output[j] = s->silk_buf[j]; |
| 623 | s->celt_output[j] = s->celt_buf[j]; |
| 624 | s->redundancy_output[j] = s->redundancy_buf[j]; |
| 625 | } |
| 626 | |
| 627 | s->fdsp = &c->fdsp; |
| 628 | |
| 629 | s->swr =swr_alloc(); |
| 630 | if (!s->swr) |
| 631 | goto fail; |
| 632 | |
| 633 | layout = (s->output_channels == 1) ? AV_CH_LAYOUT_MONO : AV_CH_LAYOUT_STEREO; |
| 634 | av_opt_set_int(s->swr, "in_sample_fmt", avctx->sample_fmt, 0); |
| 635 | av_opt_set_int(s->swr, "out_sample_fmt", avctx->sample_fmt, 0); |
| 636 | av_opt_set_int(s->swr, "in_channel_layout", layout, 0); |
| 637 | av_opt_set_int(s->swr, "out_channel_layout", layout, 0); |
| 638 | av_opt_set_int(s->swr, "out_sample_rate", avctx->sample_rate, 0); |
| 639 | av_opt_set_int(s->swr, "filter_size", 16, 0); |
| 640 | |
| 641 | ret = ff_silk_init(avctx, &s->silk, s->output_channels); |
| 642 | if (ret < 0) |
| 643 | goto fail; |
| 644 | |
| 645 | ret = ff_celt_init(avctx, &s->celt, s->output_channels); |
| 646 | if (ret < 0) |
| 647 | goto fail; |
| 648 | |
| 649 | s->celt_delay = av_audio_fifo_alloc(avctx->sample_fmt, |
| 650 | s->output_channels, 1024); |
| 651 | if (!s->celt_delay) { |
| 652 | ret = AVERROR(ENOMEM); |
| 653 | goto fail; |
| 654 | } |
| 655 | } |
| 656 | |
| 657 | return 0; |
| 658 | fail: |
| 659 | opus_decode_close(avctx); |
| 660 | return ret; |
| 661 | } |
| 662 | |
| 663 | AVCodec ff_opus_decoder = { |
| 664 | .name = "opus", |
| 665 | .long_name = NULL_IF_CONFIG_SMALL("Opus"), |
| 666 | .type = AVMEDIA_TYPE_AUDIO, |
| 667 | .id = AV_CODEC_ID_OPUS, |
| 668 | .priv_data_size = sizeof(OpusContext), |
| 669 | .init = opus_decode_init, |
| 670 | .close = opus_decode_close, |
| 671 | .decode = opus_decode_packet, |
| 672 | .flush = opus_decode_flush, |
| 673 | .capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY, |
| 674 | }; |