git.sesse.net Git - ffmpeg/blob - libavresample/utils.c

   1 /*
   2  * Copyright (c) 2012 Justin Ruggles <justin.ruggles@gmail.com>
   3  *
   4  * This file is part of Libav.
   5  *
   6  * Libav 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  * Libav 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 Libav; 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/common.h"
  22 #include "libavutil/dict.h"
  23 // #include "libavutil/error.h"
  24 #include "libavutil/log.h"
  25 #include "libavutil/mem.h"
  26 #include "libavutil/opt.h"
  27
  28 #include "avresample.h"
  29 #include "audio_data.h"
  30 #include "internal.h"
  31
  32 int avresample_open(AVAudioResampleContext *avr)
  33 {
  34     int ret;
  35
  36     /* set channel mixing parameters */
  37     avr->in_channels = av_get_channel_layout_nb_channels(avr->in_channel_layout);
  38     if (avr->in_channels <= 0 || avr->in_channels > AVRESAMPLE_MAX_CHANNELS) {
  39         av_log(avr, AV_LOG_ERROR, "Invalid input channel layout: %"PRIu64"\n",
  40                avr->in_channel_layout);
  41         return AVERROR(EINVAL);
  42     }
  43     avr->out_channels = av_get_channel_layout_nb_channels(avr->out_channel_layout);
  44     if (avr->out_channels <= 0 || avr->out_channels > AVRESAMPLE_MAX_CHANNELS) {
  45         av_log(avr, AV_LOG_ERROR, "Invalid output channel layout: %"PRIu64"\n",
  46                avr->out_channel_layout);
  47         return AVERROR(EINVAL);
  48     }
  49     avr->resample_channels = FFMIN(avr->in_channels, avr->out_channels);
  50     avr->downmix_needed    = avr->in_channels  > avr->out_channels;
  51     avr->upmix_needed      = avr->out_channels > avr->in_channels ||
  52                              (!avr->downmix_needed && (avr->mix_matrix ||
  53                               avr->in_channel_layout != avr->out_channel_layout));
  54     avr->mixing_needed     = avr->downmix_needed || avr->upmix_needed;
  55
  56     /* set resampling parameters */
  57     avr->resample_needed   = avr->in_sample_rate != avr->out_sample_rate ||
  58                              avr->force_resampling;
  59
  60     /* select internal sample format if not specified by the user */
  61     if (avr->internal_sample_fmt == AV_SAMPLE_FMT_NONE &&
  62         (avr->mixing_needed || avr->resample_needed)) {
  63         enum AVSampleFormat  in_fmt = av_get_planar_sample_fmt(avr->in_sample_fmt);
  64         enum AVSampleFormat out_fmt = av_get_planar_sample_fmt(avr->out_sample_fmt);
  65         int max_bps = FFMAX(av_get_bytes_per_sample(in_fmt),
  66                             av_get_bytes_per_sample(out_fmt));
  67         if (max_bps <= 2) {
  68             avr->internal_sample_fmt = AV_SAMPLE_FMT_S16P;
  69         } else if (avr->mixing_needed) {
  70             avr->internal_sample_fmt = AV_SAMPLE_FMT_FLTP;
  71         } else {
  72             if (max_bps <= 4) {
  73                 if (in_fmt  == AV_SAMPLE_FMT_S32P ||
  74                     out_fmt == AV_SAMPLE_FMT_S32P) {
  75                     if (in_fmt  == AV_SAMPLE_FMT_FLTP ||
  76                         out_fmt == AV_SAMPLE_FMT_FLTP) {
  77                         /* if one is s32 and the other is flt, use dbl */
  78                         avr->internal_sample_fmt = AV_SAMPLE_FMT_DBLP;
  79                     } else {
  80                         /* if one is s32 and the other is s32, s16, or u8, use s32 */
  81                         avr->internal_sample_fmt = AV_SAMPLE_FMT_S32P;
  82                     }
  83                 } else {
  84                     /* if one is flt and the other is flt, s16 or u8, use flt */
  85                     avr->internal_sample_fmt = AV_SAMPLE_FMT_FLTP;
  86                 }
  87             } else {
  88                 /* if either is dbl, use dbl */
  89                 avr->internal_sample_fmt = AV_SAMPLE_FMT_DBLP;
  90             }
  91         }
  92         av_log(avr, AV_LOG_DEBUG, "Using %s as internal sample format\n",
  93                av_get_sample_fmt_name(avr->internal_sample_fmt));
  94     }
  95
  96     /* set sample format conversion parameters */
  97     if (avr->in_channels == 1)
  98         avr->in_sample_fmt = av_get_planar_sample_fmt(avr->in_sample_fmt);
  99     if (avr->out_channels == 1)
 100         avr->out_sample_fmt = av_get_planar_sample_fmt(avr->out_sample_fmt);
 101     avr->in_convert_needed = (avr->resample_needed || avr->mixing_needed) &&
 102                               avr->in_sample_fmt != avr->internal_sample_fmt;
 103     if (avr->resample_needed || avr->mixing_needed)
 104         avr->out_convert_needed = avr->internal_sample_fmt != avr->out_sample_fmt;
 105     else
 106         avr->out_convert_needed = avr->in_sample_fmt != avr->out_sample_fmt;
 107
 108     /* allocate buffers */
 109     if (avr->mixing_needed || avr->in_convert_needed) {
 110         avr->in_buffer = ff_audio_data_alloc(FFMAX(avr->in_channels, avr->out_channels),
 111                                              0, avr->internal_sample_fmt,
 112                                              "in_buffer");
 113         if (!avr->in_buffer) {
 114             ret = AVERROR(EINVAL);
 115             goto error;
 116         }
 117     }
 118     if (avr->resample_needed) {
 119         avr->resample_out_buffer = ff_audio_data_alloc(avr->out_channels,
 120                                                        0, avr->internal_sample_fmt,
 121                                                        "resample_out_buffer");
 122         if (!avr->resample_out_buffer) {
 123             ret = AVERROR(EINVAL);
 124             goto error;
 125         }
 126     }
 127     if (avr->out_convert_needed) {
 128         avr->out_buffer = ff_audio_data_alloc(avr->out_channels, 0,
 129                                               avr->out_sample_fmt, "out_buffer");
 130         if (!avr->out_buffer) {
 131             ret = AVERROR(EINVAL);
 132             goto error;
 133         }
 134     }
 135     avr->out_fifo = av_audio_fifo_alloc(avr->out_sample_fmt, avr->out_channels,
 136                                         1024);
 137     if (!avr->out_fifo) {
 138         ret = AVERROR(ENOMEM);
 139         goto error;
 140     }
 141
 142     /* setup contexts */
 143     if (avr->in_convert_needed) {
 144         avr->ac_in = ff_audio_convert_alloc(avr, avr->internal_sample_fmt,
 145                                             avr->in_sample_fmt, avr->in_channels,
 146                                             avr->in_sample_rate);
 147         if (!avr->ac_in) {
 148             ret = AVERROR(ENOMEM);
 149             goto error;
 150         }
 151     }
 152     if (avr->out_convert_needed) {
 153         enum AVSampleFormat src_fmt;
 154         if (avr->in_convert_needed)
 155             src_fmt = avr->internal_sample_fmt;
 156         else
 157             src_fmt = avr->in_sample_fmt;
 158         avr->ac_out = ff_audio_convert_alloc(avr, avr->out_sample_fmt, src_fmt,
 159                                              avr->out_channels,
 160                                              avr->out_sample_rate);
 161         if (!avr->ac_out) {
 162             ret = AVERROR(ENOMEM);
 163             goto error;
 164         }
 165     }
 166     if (avr->resample_needed) {
 167         avr->resample = ff_audio_resample_init(avr);
 168         if (!avr->resample) {
 169             ret = AVERROR(ENOMEM);
 170             goto error;
 171         }
 172     }
 173     if (avr->mixing_needed) {
 174         avr->am = ff_audio_mix_alloc(avr);
 175         if (!avr->am) {
 176             ret = AVERROR(ENOMEM);
 177             goto error;
 178         }
 179     }
 180
 181     return 0;
 182
 183 error:
 184     avresample_close(avr);
 185     return ret;
 186 }
 187
 188 void avresample_close(AVAudioResampleContext *avr)
 189 {
 190     ff_audio_data_free(&avr->in_buffer);
 191     ff_audio_data_free(&avr->resample_out_buffer);
 192     ff_audio_data_free(&avr->out_buffer);
 193     av_audio_fifo_free(avr->out_fifo);
 194     avr->out_fifo = NULL;
 195     ff_audio_convert_free(&avr->ac_in);
 196     ff_audio_convert_free(&avr->ac_out);
 197     ff_audio_resample_free(&avr->resample);
 198     ff_audio_mix_free(&avr->am);
 199     av_freep(&avr->mix_matrix);
 200 }
 201
 202 void avresample_free(AVAudioResampleContext **avr)
 203 {
 204     if (!*avr)
 205         return;
 206     avresample_close(*avr);
 207     av_opt_free(*avr);
 208     av_freep(avr);
 209 }
 210
 211 static int handle_buffered_output(AVAudioResampleContext *avr,
 212                                   AudioData *output, AudioData *converted)
 213 {
 214     int ret;
 215
 216     if (!output || av_audio_fifo_size(avr->out_fifo) > 0 ||
 217         (converted && output->allocated_samples < converted->nb_samples)) {
 218         if (converted) {
 219             /* if there are any samples in the output FIFO or if the
 220                user-supplied output buffer is not large enough for all samples,
 221                we add to the output FIFO */
 222             av_dlog(avr, "[FIFO] add %s to out_fifo\n", converted->name);
 223             ret = ff_audio_data_add_to_fifo(avr->out_fifo, converted, 0,
 224                                             converted->nb_samples);
 225             if (ret < 0)
 226                 return ret;
 227         }
 228
 229         /* if the user specified an output buffer, read samples from the output
 230            FIFO to the user output */
 231         if (output && output->allocated_samples > 0) {
 232             av_dlog(avr, "[FIFO] read from out_fifo to output\n");
 233             av_dlog(avr, "[end conversion]\n");
 234             return ff_audio_data_read_from_fifo(avr->out_fifo, output,
 235                                                 output->allocated_samples);
 236         }
 237     } else if (converted) {
 238         /* copy directly to output if it is large enough or there is not any
 239            data in the output FIFO */
 240         av_dlog(avr, "[copy] %s to output\n", converted->name);
 241         output->nb_samples = 0;
 242         ret = ff_audio_data_copy(output, converted);
 243         if (ret < 0)
 244             return ret;
 245         av_dlog(avr, "[end conversion]\n");
 246         return output->nb_samples;
 247     }
 248     av_dlog(avr, "[end conversion]\n");
 249     return 0;
 250 }
 251
 252 int attribute_align_arg avresample_convert(AVAudioResampleContext *avr,
 253                                            uint8_t **output, int out_plane_size,
 254                                            int out_samples, uint8_t **input,
 255                                            int in_plane_size, int in_samples)
 256 {
 257     AudioData input_buffer;
 258     AudioData output_buffer;
 259     AudioData *current_buffer;
 260     int ret, direct_output;
 261
 262     /* reset internal buffers */
 263     if (avr->in_buffer) {
 264         avr->in_buffer->nb_samples = 0;
 265         ff_audio_data_set_channels(avr->in_buffer,
 266                                    avr->in_buffer->allocated_channels);
 267     }
 268     if (avr->resample_out_buffer) {
 269         avr->resample_out_buffer->nb_samples = 0;
 270         ff_audio_data_set_channels(avr->resample_out_buffer,
 271                                    avr->resample_out_buffer->allocated_channels);
 272     }
 273     if (avr->out_buffer) {
 274         avr->out_buffer->nb_samples = 0;
 275         ff_audio_data_set_channels(avr->out_buffer,
 276                                    avr->out_buffer->allocated_channels);
 277     }
 278
 279     av_dlog(avr, "[start conversion]\n");
 280
 281     /* initialize output_buffer with output data */
 282     direct_output = output && av_audio_fifo_size(avr->out_fifo) == 0;
 283     if (output) {
 284         ret = ff_audio_data_init(&output_buffer, output, out_plane_size,
 285                                  avr->out_channels, out_samples,
 286                                  avr->out_sample_fmt, 0, "output");
 287         if (ret < 0)
 288             return ret;
 289         output_buffer.nb_samples = 0;
 290     }
 291
 292     if (input) {
 293         /* initialize input_buffer with input data */
 294         ret = ff_audio_data_init(&input_buffer, input, in_plane_size,
 295                                  avr->in_channels, in_samples,
 296                                  avr->in_sample_fmt, 1, "input");
 297         if (ret < 0)
 298             return ret;
 299         current_buffer = &input_buffer;
 300
 301         if (avr->upmix_needed && !avr->in_convert_needed && !avr->resample_needed &&
 302             !avr->out_convert_needed && direct_output && out_samples >= in_samples) {
 303             /* in some rare cases we can copy input to output and upmix
 304                directly in the output buffer */
 305             av_dlog(avr, "[copy] %s to output\n", current_buffer->name);
 306             ret = ff_audio_data_copy(&output_buffer, current_buffer);
 307             if (ret < 0)
 308                 return ret;
 309             current_buffer = &output_buffer;
 310         } else if (avr->mixing_needed || avr->in_convert_needed) {
 311             /* if needed, copy or convert input to in_buffer, and downmix if
 312                applicable */
 313             if (avr->in_convert_needed) {
 314                 ret = ff_audio_data_realloc(avr->in_buffer,
 315                                             current_buffer->nb_samples);
 316                 if (ret < 0)
 317                     return ret;
 318                 av_dlog(avr, "[convert] %s to in_buffer\n", current_buffer->name);
 319                 ret = ff_audio_convert(avr->ac_in, avr->in_buffer,
 320                                        current_buffer);
 321                 if (ret < 0)
 322                     return ret;
 323             } else {
 324                 av_dlog(avr, "[copy] %s to in_buffer\n", current_buffer->name);
 325                 ret = ff_audio_data_copy(avr->in_buffer, current_buffer);
 326                 if (ret < 0)
 327                     return ret;
 328             }
 329             ff_audio_data_set_channels(avr->in_buffer, avr->in_channels);
 330             if (avr->downmix_needed) {
 331                 av_dlog(avr, "[downmix] in_buffer\n");
 332                 ret = ff_audio_mix(avr->am, avr->in_buffer);
 333                 if (ret < 0)
 334                     return ret;
 335             }
 336             current_buffer = avr->in_buffer;
 337         }
 338     } else {
 339         /* flush resampling buffer and/or output FIFO if input is NULL */
 340         if (!avr->resample_needed)
 341             return handle_buffered_output(avr, output ? &output_buffer : NULL,
 342                                           NULL);
 343         current_buffer = NULL;
 344     }
 345
 346     if (avr->resample_needed) {
 347         AudioData *resample_out;
 348
 349         if (!avr->out_convert_needed && direct_output && out_samples > 0)
 350             resample_out = &output_buffer;
 351         else
 352             resample_out = avr->resample_out_buffer;
 353         av_dlog(avr, "[resample] %s to %s\n", current_buffer->name,
 354                 resample_out->name);
 355         ret = ff_audio_resample(avr->resample, resample_out,
 356                                 current_buffer);
 357         if (ret < 0)
 358             return ret;
 359
 360         /* if resampling did not produce any samples, just return 0 */
 361         if (resample_out->nb_samples == 0) {
 362             av_dlog(avr, "[end conversion]\n");
 363             return 0;
 364         }
 365
 366         current_buffer = resample_out;
 367     }
 368
 369     if (avr->upmix_needed) {
 370         av_dlog(avr, "[upmix] %s\n", current_buffer->name);
 371         ret = ff_audio_mix(avr->am, current_buffer);
 372         if (ret < 0)
 373             return ret;
 374     }
 375
 376     /* if we resampled or upmixed directly to output, return here */
 377     if (current_buffer == &output_buffer) {
 378         av_dlog(avr, "[end conversion]\n");
 379         return current_buffer->nb_samples;
 380     }
 381
 382     if (avr->out_convert_needed) {
 383         if (direct_output && out_samples >= current_buffer->nb_samples) {
 384             /* convert directly to output */
 385             av_dlog(avr, "[convert] %s to output\n", current_buffer->name);
 386             ret = ff_audio_convert(avr->ac_out, &output_buffer, current_buffer);
 387             if (ret < 0)
 388                 return ret;
 389
 390             av_dlog(avr, "[end conversion]\n");
 391             return output_buffer.nb_samples;
 392         } else {
 393             ret = ff_audio_data_realloc(avr->out_buffer,
 394                                         current_buffer->nb_samples);
 395             if (ret < 0)
 396                 return ret;
 397             av_dlog(avr, "[convert] %s to out_buffer\n", current_buffer->name);
 398             ret = ff_audio_convert(avr->ac_out, avr->out_buffer,
 399                                    current_buffer);
 400             if (ret < 0)
 401                 return ret;
 402             current_buffer = avr->out_buffer;
 403         }
 404     }
 405
 406     return handle_buffered_output(avr, output ? &output_buffer : NULL,
 407                                   current_buffer);
 408 }
 409
 410 int avresample_get_matrix(AVAudioResampleContext *avr, double *matrix,
 411                           int stride)
 412 {
 413     int in_channels, out_channels, i, o;
 414
 415     if (avr->am)
 416         return ff_audio_mix_get_matrix(avr->am, matrix, stride);
 417
 418     in_channels  = av_get_channel_layout_nb_channels(avr->in_channel_layout);
 419     out_channels = av_get_channel_layout_nb_channels(avr->out_channel_layout);
 420
 421     if ( in_channels <= 0 ||  in_channels > AVRESAMPLE_MAX_CHANNELS ||
 422         out_channels <= 0 || out_channels > AVRESAMPLE_MAX_CHANNELS) {
 423         av_log(avr, AV_LOG_ERROR, "Invalid channel layouts\n");
 424         return AVERROR(EINVAL);
 425     }
 426
 427     if (!avr->mix_matrix) {
 428         av_log(avr, AV_LOG_ERROR, "matrix is not set\n");
 429         return AVERROR(EINVAL);
 430     }
 431
 432     for (o = 0; o < out_channels; o++)
 433         for (i = 0; i < in_channels; i++)
 434             matrix[o * stride + i] = avr->mix_matrix[o * in_channels + i];
 435
 436     return 0;
 437 }
 438
 439 int avresample_set_matrix(AVAudioResampleContext *avr, const double *matrix,
 440                           int stride)
 441 {
 442     int in_channels, out_channels, i, o;
 443
 444     if (avr->am)
 445         return ff_audio_mix_set_matrix(avr->am, matrix, stride);
 446
 447     in_channels  = av_get_channel_layout_nb_channels(avr->in_channel_layout);
 448     out_channels = av_get_channel_layout_nb_channels(avr->out_channel_layout);
 449
 450     if ( in_channels <= 0 ||  in_channels > AVRESAMPLE_MAX_CHANNELS ||
 451         out_channels <= 0 || out_channels > AVRESAMPLE_MAX_CHANNELS) {
 452         av_log(avr, AV_LOG_ERROR, "Invalid channel layouts\n");
 453         return AVERROR(EINVAL);
 454     }
 455
 456     if (avr->mix_matrix)
 457         av_freep(&avr->mix_matrix);
 458     avr->mix_matrix = av_malloc(in_channels * out_channels *
 459                                 sizeof(*avr->mix_matrix));
 460     if (!avr->mix_matrix)
 461         return AVERROR(ENOMEM);
 462
 463     for (o = 0; o < out_channels; o++)
 464         for (i = 0; i < in_channels; i++)
 465             avr->mix_matrix[o * in_channels + i] = matrix[o * stride + i];
 466
 467     return 0;
 468 }
 469
 470 int avresample_available(AVAudioResampleContext *avr)
 471 {
 472     return av_audio_fifo_size(avr->out_fifo);
 473 }
 474
 475 int avresample_read(AVAudioResampleContext *avr, uint8_t **output, int nb_samples)
 476 {
 477     if (!output)
 478         return av_audio_fifo_drain(avr->out_fifo, nb_samples);
 479     return av_audio_fifo_read(avr->out_fifo, (void**)output, nb_samples);
 480 }
 481
 482 unsigned avresample_version(void)
 483 {
 484     return LIBAVRESAMPLE_VERSION_INT;
 485 }
 486
 487 const char *avresample_license(void)
 488 {
 489 #define LICENSE_PREFIX "libavresample license: "
 490     return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
 491 }
 492
 493 const char *avresample_configuration(void)
 494 {
 495     return FFMPEG_CONFIGURATION;
 496 }