FFmpeg: libavresample/utils.c Source File

00001 /*
00002  * Copyright (c) 2012 Justin Ruggles <justin.ruggles@gmail.com>
00003  *
00004  * This file is part of Libav.
00005  *
00006  * Libav is free software; you can redistribute it and/or
00007  * modify it under the terms of the GNU Lesser General Public
00008  * License as published by the Free Software Foundation; either
00009  * version 2.1 of the License, or (at your option) any later version.
00010  *
00011  * Libav is distributed in the hope that it will be useful,
00012  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00013  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00014  * Lesser General Public License for more details.
00015  *
00016  * You should have received a copy of the GNU Lesser General Public
00017  * License along with Libav; if not, write to the Free Software
00018  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00019  */
00020 
00021 #include "libavutil/common.h"
00022 #include "libavutil/dict.h"
00023 // #include "libavutil/error.h"
00024 #include "libavutil/log.h"
00025 #include "libavutil/mem.h"
00026 #include "libavutil/opt.h"
00027 
00028 #include "avresample.h"
00029 #include "audio_data.h"
00030 #include "internal.h"
00031 
00032 int avresample_open(AVAudioResampleContext *avr)
00033 {
00034     int ret;
00035 
00036     /* set channel mixing parameters */
00037     avr->in_channels = av_get_channel_layout_nb_channels(avr->in_channel_layout);
00038     if (avr->in_channels <= 0 || avr->in_channels > AVRESAMPLE_MAX_CHANNELS) {
00039         av_log(avr, AV_LOG_ERROR, "Invalid input channel layout: %"PRIu64"\n",
00040                avr->in_channel_layout);
00041         return AVERROR(EINVAL);
00042     }
00043     avr->out_channels = av_get_channel_layout_nb_channels(avr->out_channel_layout);
00044     if (avr->out_channels <= 0 || avr->out_channels > AVRESAMPLE_MAX_CHANNELS) {
00045         av_log(avr, AV_LOG_ERROR, "Invalid output channel layout: %"PRIu64"\n",
00046                avr->out_channel_layout);
00047         return AVERROR(EINVAL);
00048     }
00049     avr->resample_channels = FFMIN(avr->in_channels, avr->out_channels);
00050     avr->downmix_needed    = avr->in_channels  > avr->out_channels;
00051     avr->upmix_needed      = avr->out_channels > avr->in_channels ||
00052                              (!avr->downmix_needed && (avr->mix_matrix ||
00053                               avr->in_channel_layout != avr->out_channel_layout));
00054     avr->mixing_needed     = avr->downmix_needed || avr->upmix_needed;
00055 
00056     /* set resampling parameters */
00057     avr->resample_needed   = avr->in_sample_rate != avr->out_sample_rate ||
00058                              avr->force_resampling;
00059 
00060     /* select internal sample format if not specified by the user */
00061     if (avr->internal_sample_fmt == AV_SAMPLE_FMT_NONE &&
00062         (avr->mixing_needed || avr->resample_needed)) {
00063         enum AVSampleFormat  in_fmt = av_get_planar_sample_fmt(avr->in_sample_fmt);
00064         enum AVSampleFormat out_fmt = av_get_planar_sample_fmt(avr->out_sample_fmt);
00065         int max_bps = FFMAX(av_get_bytes_per_sample(in_fmt),
00066                             av_get_bytes_per_sample(out_fmt));
00067         if (max_bps <= 2) {
00068             avr->internal_sample_fmt = AV_SAMPLE_FMT_S16P;
00069         } else if (avr->mixing_needed) {
00070             avr->internal_sample_fmt = AV_SAMPLE_FMT_FLTP;
00071         } else {
00072             if (max_bps <= 4) {
00073                 if (in_fmt  == AV_SAMPLE_FMT_S32P ||
00074                     out_fmt == AV_SAMPLE_FMT_S32P) {
00075                     if (in_fmt  == AV_SAMPLE_FMT_FLTP ||
00076                         out_fmt == AV_SAMPLE_FMT_FLTP) {
00077                         /* if one is s32 and the other is flt, use dbl */
00078                         avr->internal_sample_fmt = AV_SAMPLE_FMT_DBLP;
00079                     } else {
00080                         /* if one is s32 and the other is s32, s16, or u8, use s32 */
00081                         avr->internal_sample_fmt = AV_SAMPLE_FMT_S32P;
00082                     }
00083                 } else {
00084                     /* if one is flt and the other is flt, s16 or u8, use flt */
00085                     avr->internal_sample_fmt = AV_SAMPLE_FMT_FLTP;
00086                 }
00087             } else {
00088                 /* if either is dbl, use dbl */
00089                 avr->internal_sample_fmt = AV_SAMPLE_FMT_DBLP;
00090             }
00091         }
00092         av_log(avr, AV_LOG_DEBUG, "Using %s as internal sample format\n",
00093                av_get_sample_fmt_name(avr->internal_sample_fmt));
00094     }
00095 
00096     /* set sample format conversion parameters */
00097     if (avr->in_channels == 1)
00098         avr->in_sample_fmt = av_get_planar_sample_fmt(avr->in_sample_fmt);
00099     if (avr->out_channels == 1)
00100         avr->out_sample_fmt = av_get_planar_sample_fmt(avr->out_sample_fmt);
00101     avr->in_convert_needed = (avr->resample_needed || avr->mixing_needed) &&
00102                               avr->in_sample_fmt != avr->internal_sample_fmt;
00103     if (avr->resample_needed || avr->mixing_needed)
00104         avr->out_convert_needed = avr->internal_sample_fmt != avr->out_sample_fmt;
00105     else
00106         avr->out_convert_needed = avr->in_sample_fmt != avr->out_sample_fmt;
00107 
00108     /* allocate buffers */
00109     if (avr->mixing_needed || avr->in_convert_needed) {
00110         avr->in_buffer = ff_audio_data_alloc(FFMAX(avr->in_channels, avr->out_channels),
00111                                              0, avr->internal_sample_fmt,
00112                                              "in_buffer");
00113         if (!avr->in_buffer) {
00114             ret = AVERROR(EINVAL);
00115             goto error;
00116         }
00117     }
00118     if (avr->resample_needed) {
00119         avr->resample_out_buffer = ff_audio_data_alloc(avr->out_channels,
00120                                                        0, avr->internal_sample_fmt,
00121                                                        "resample_out_buffer");
00122         if (!avr->resample_out_buffer) {
00123             ret = AVERROR(EINVAL);
00124             goto error;
00125         }
00126     }
00127     if (avr->out_convert_needed) {
00128         avr->out_buffer = ff_audio_data_alloc(avr->out_channels, 0,
00129                                               avr->out_sample_fmt, "out_buffer");
00130         if (!avr->out_buffer) {
00131             ret = AVERROR(EINVAL);
00132             goto error;
00133         }
00134     }
00135     avr->out_fifo = av_audio_fifo_alloc(avr->out_sample_fmt, avr->out_channels,
00136                                         1024);
00137     if (!avr->out_fifo) {
00138         ret = AVERROR(ENOMEM);
00139         goto error;
00140     }
00141 
00142     /* setup contexts */
00143     if (avr->in_convert_needed) {
00144         avr->ac_in = ff_audio_convert_alloc(avr, avr->internal_sample_fmt,
00145                                             avr->in_sample_fmt, avr->in_channels);
00146         if (!avr->ac_in) {
00147             ret = AVERROR(ENOMEM);
00148             goto error;
00149         }
00150     }
00151     if (avr->out_convert_needed) {
00152         enum AVSampleFormat src_fmt;
00153         if (avr->in_convert_needed)
00154             src_fmt = avr->internal_sample_fmt;
00155         else
00156             src_fmt = avr->in_sample_fmt;
00157         avr->ac_out = ff_audio_convert_alloc(avr, avr->out_sample_fmt, src_fmt,
00158                                              avr->out_channels);
00159         if (!avr->ac_out) {
00160             ret = AVERROR(ENOMEM);
00161             goto error;
00162         }
00163     }
00164     if (avr->resample_needed) {
00165         avr->resample = ff_audio_resample_init(avr);
00166         if (!avr->resample) {
00167             ret = AVERROR(ENOMEM);
00168             goto error;
00169         }
00170     }
00171     if (avr->mixing_needed) {
00172         avr->am = ff_audio_mix_alloc(avr);
00173         if (!avr->am) {
00174             ret = AVERROR(ENOMEM);
00175             goto error;
00176         }
00177     }
00178 
00179     return 0;
00180 
00181 error:
00182     avresample_close(avr);
00183     return ret;
00184 }
00185 
00186 void avresample_close(AVAudioResampleContext *avr)
00187 {
00188     ff_audio_data_free(&avr->in_buffer);
00189     ff_audio_data_free(&avr->resample_out_buffer);
00190     ff_audio_data_free(&avr->out_buffer);
00191     av_audio_fifo_free(avr->out_fifo);
00192     avr->out_fifo = NULL;
00193     av_freep(&avr->ac_in);
00194     av_freep(&avr->ac_out);
00195     ff_audio_resample_free(&avr->resample);
00196     ff_audio_mix_free(&avr->am);
00197     av_freep(&avr->mix_matrix);
00198 }
00199 
00200 void avresample_free(AVAudioResampleContext **avr)
00201 {
00202     if (!*avr)
00203         return;
00204     avresample_close(*avr);
00205     av_opt_free(*avr);
00206     av_freep(avr);
00207 }
00208 
00209 static int handle_buffered_output(AVAudioResampleContext *avr,
00210                                   AudioData *output, AudioData *converted)
00211 {
00212     int ret;
00213 
00214     if (!output || av_audio_fifo_size(avr->out_fifo) > 0 ||
00215         (converted && output->allocated_samples < converted->nb_samples)) {
00216         if (converted) {
00217             /* if there are any samples in the output FIFO or if the
00218                user-supplied output buffer is not large enough for all samples,
00219                we add to the output FIFO */
00220             av_dlog(avr, "[FIFO] add %s to out_fifo\n", converted->name);
00221             ret = ff_audio_data_add_to_fifo(avr->out_fifo, converted, 0,
00222                                             converted->nb_samples);
00223             if (ret < 0)
00224                 return ret;
00225         }
00226 
00227         /* if the user specified an output buffer, read samples from the output
00228            FIFO to the user output */
00229         if (output && output->allocated_samples > 0) {
00230             av_dlog(avr, "[FIFO] read from out_fifo to output\n");
00231             av_dlog(avr, "[end conversion]\n");
00232             return ff_audio_data_read_from_fifo(avr->out_fifo, output,
00233                                                 output->allocated_samples);
00234         }
00235     } else if (converted) {
00236         /* copy directly to output if it is large enough or there is not any
00237            data in the output FIFO */
00238         av_dlog(avr, "[copy] %s to output\n", converted->name);
00239         output->nb_samples = 0;
00240         ret = ff_audio_data_copy(output, converted);
00241         if (ret < 0)
00242             return ret;
00243         av_dlog(avr, "[end conversion]\n");
00244         return output->nb_samples;
00245     }
00246     av_dlog(avr, "[end conversion]\n");
00247     return 0;
00248 }
00249 
00250 int attribute_align_arg avresample_convert(AVAudioResampleContext *avr,
00251                                            uint8_t **output, int out_plane_size,
00252                                            int out_samples, uint8_t **input,
00253                                            int in_plane_size, int in_samples)
00254 {
00255     AudioData input_buffer;
00256     AudioData output_buffer;
00257     AudioData *current_buffer;
00258     int ret, direct_output;
00259 
00260     /* reset internal buffers */
00261     if (avr->in_buffer) {
00262         avr->in_buffer->nb_samples = 0;
00263         ff_audio_data_set_channels(avr->in_buffer,
00264                                    avr->in_buffer->allocated_channels);
00265     }
00266     if (avr->resample_out_buffer) {
00267         avr->resample_out_buffer->nb_samples = 0;
00268         ff_audio_data_set_channels(avr->resample_out_buffer,
00269                                    avr->resample_out_buffer->allocated_channels);
00270     }
00271     if (avr->out_buffer) {
00272         avr->out_buffer->nb_samples = 0;
00273         ff_audio_data_set_channels(avr->out_buffer,
00274                                    avr->out_buffer->allocated_channels);
00275     }
00276 
00277     av_dlog(avr, "[start conversion]\n");
00278 
00279     /* initialize output_buffer with output data */
00280     direct_output = output && av_audio_fifo_size(avr->out_fifo) == 0;
00281     if (output) {
00282         ret = ff_audio_data_init(&output_buffer, output, out_plane_size,
00283                                  avr->out_channels, out_samples,
00284                                  avr->out_sample_fmt, 0, "output");
00285         if (ret < 0)
00286             return ret;
00287         output_buffer.nb_samples = 0;
00288     }
00289 
00290     if (input) {
00291         /* initialize input_buffer with input data */
00292         ret = ff_audio_data_init(&input_buffer, input, in_plane_size,
00293                                  avr->in_channels, in_samples,
00294                                  avr->in_sample_fmt, 1, "input");
00295         if (ret < 0)
00296             return ret;
00297         current_buffer = &input_buffer;
00298 
00299         if (avr->upmix_needed && !avr->in_convert_needed && !avr->resample_needed &&
00300             !avr->out_convert_needed && direct_output && out_samples >= in_samples) {
00301             /* in some rare cases we can copy input to output and upmix
00302                directly in the output buffer */
00303             av_dlog(avr, "[copy] %s to output\n", current_buffer->name);
00304             ret = ff_audio_data_copy(&output_buffer, current_buffer);
00305             if (ret < 0)
00306                 return ret;
00307             current_buffer = &output_buffer;
00308         } else if (avr->mixing_needed || avr->in_convert_needed) {
00309             /* if needed, copy or convert input to in_buffer, and downmix if
00310                applicable */
00311             if (avr->in_convert_needed) {
00312                 ret = ff_audio_data_realloc(avr->in_buffer,
00313                                             current_buffer->nb_samples);
00314                 if (ret < 0)
00315                     return ret;
00316                 av_dlog(avr, "[convert] %s to in_buffer\n", current_buffer->name);
00317                 ret = ff_audio_convert(avr->ac_in, avr->in_buffer,
00318                                        current_buffer);
00319                 if (ret < 0)
00320                     return ret;
00321             } else {
00322                 av_dlog(avr, "[copy] %s to in_buffer\n", current_buffer->name);
00323                 ret = ff_audio_data_copy(avr->in_buffer, current_buffer);
00324                 if (ret < 0)
00325                     return ret;
00326             }
00327             ff_audio_data_set_channels(avr->in_buffer, avr->in_channels);
00328             if (avr->downmix_needed) {
00329                 av_dlog(avr, "[downmix] in_buffer\n");
00330                 ret = ff_audio_mix(avr->am, avr->in_buffer);
00331                 if (ret < 0)
00332                     return ret;
00333             }
00334             current_buffer = avr->in_buffer;
00335         }
00336     } else {
00337         /* flush resampling buffer and/or output FIFO if input is NULL */
00338         if (!avr->resample_needed)
00339             return handle_buffered_output(avr, output ? &output_buffer : NULL,
00340                                           NULL);
00341         current_buffer = NULL;
00342     }
00343 
00344     if (avr->resample_needed) {
00345         AudioData *resample_out;
00346 
00347         if (!avr->out_convert_needed && direct_output && out_samples > 0)
00348             resample_out = &output_buffer;
00349         else
00350             resample_out = avr->resample_out_buffer;
00351         av_dlog(avr, "[resample] %s to %s\n", current_buffer->name,
00352                 resample_out->name);
00353         ret = ff_audio_resample(avr->resample, resample_out,
00354                                 current_buffer);
00355         if (ret < 0)
00356             return ret;
00357 
00358         /* if resampling did not produce any samples, just return 0 */
00359         if (resample_out->nb_samples == 0) {
00360             av_dlog(avr, "[end conversion]\n");
00361             return 0;
00362         }
00363 
00364         current_buffer = resample_out;
00365     }
00366 
00367     if (avr->upmix_needed) {
00368         av_dlog(avr, "[upmix] %s\n", current_buffer->name);
00369         ret = ff_audio_mix(avr->am, current_buffer);
00370         if (ret < 0)
00371             return ret;
00372     }
00373 
00374     /* if we resampled or upmixed directly to output, return here */
00375     if (current_buffer == &output_buffer) {
00376         av_dlog(avr, "[end conversion]\n");
00377         return current_buffer->nb_samples;
00378     }
00379 
00380     if (avr->out_convert_needed) {
00381         if (direct_output && out_samples >= current_buffer->nb_samples) {
00382             /* convert directly to output */
00383             av_dlog(avr, "[convert] %s to output\n", current_buffer->name);
00384             ret = ff_audio_convert(avr->ac_out, &output_buffer, current_buffer);
00385             if (ret < 0)
00386                 return ret;
00387 
00388             av_dlog(avr, "[end conversion]\n");
00389             return output_buffer.nb_samples;
00390         } else {
00391             ret = ff_audio_data_realloc(avr->out_buffer,
00392                                         current_buffer->nb_samples);
00393             if (ret < 0)
00394                 return ret;
00395             av_dlog(avr, "[convert] %s to out_buffer\n", current_buffer->name);
00396             ret = ff_audio_convert(avr->ac_out, avr->out_buffer,
00397                                    current_buffer);
00398             if (ret < 0)
00399                 return ret;
00400             current_buffer = avr->out_buffer;
00401         }
00402     }
00403 
00404     return handle_buffered_output(avr, output ? &output_buffer : NULL,
00405                                   current_buffer);
00406 }
00407 
00408 int avresample_get_matrix(AVAudioResampleContext *avr, double *matrix,
00409                           int stride)
00410 {
00411     int in_channels, out_channels, i, o;
00412 
00413     if (avr->am)
00414         return ff_audio_mix_get_matrix(avr->am, matrix, stride);
00415 
00416     in_channels  = av_get_channel_layout_nb_channels(avr->in_channel_layout);
00417     out_channels = av_get_channel_layout_nb_channels(avr->out_channel_layout);
00418 
00419     if ( in_channels <= 0 ||  in_channels > AVRESAMPLE_MAX_CHANNELS ||
00420         out_channels <= 0 || out_channels > AVRESAMPLE_MAX_CHANNELS) {
00421         av_log(avr, AV_LOG_ERROR, "Invalid channel layouts\n");
00422         return AVERROR(EINVAL);
00423     }
00424 
00425     if (!avr->mix_matrix) {
00426         av_log(avr, AV_LOG_ERROR, "matrix is not set\n");
00427         return AVERROR(EINVAL);
00428     }
00429 
00430     for (o = 0; o < out_channels; o++)
00431         for (i = 0; i < in_channels; i++)
00432             matrix[o * stride + i] = avr->mix_matrix[o * in_channels + i];
00433 
00434     return 0;
00435 }
00436 
00437 int avresample_set_matrix(AVAudioResampleContext *avr, const double *matrix,
00438                           int stride)
00439 {
00440     int in_channels, out_channels, i, o;
00441 
00442     if (avr->am)
00443         return ff_audio_mix_set_matrix(avr->am, matrix, stride);
00444 
00445     in_channels  = av_get_channel_layout_nb_channels(avr->in_channel_layout);
00446     out_channels = av_get_channel_layout_nb_channels(avr->out_channel_layout);
00447 
00448     if ( in_channels <= 0 ||  in_channels > AVRESAMPLE_MAX_CHANNELS ||
00449         out_channels <= 0 || out_channels > AVRESAMPLE_MAX_CHANNELS) {
00450         av_log(avr, AV_LOG_ERROR, "Invalid channel layouts\n");
00451         return AVERROR(EINVAL);
00452     }
00453 
00454     if (avr->mix_matrix)
00455         av_freep(&avr->mix_matrix);
00456     avr->mix_matrix = av_malloc(in_channels * out_channels *
00457                                 sizeof(*avr->mix_matrix));
00458     if (!avr->mix_matrix)
00459         return AVERROR(ENOMEM);
00460 
00461     for (o = 0; o < out_channels; o++)
00462         for (i = 0; i < in_channels; i++)
00463             avr->mix_matrix[o * in_channels + i] = matrix[o * stride + i];
00464 
00465     return 0;
00466 }
00467 
00468 int avresample_available(AVAudioResampleContext *avr)
00469 {
00470     return av_audio_fifo_size(avr->out_fifo);
00471 }
00472 
00473 int avresample_read(AVAudioResampleContext *avr, uint8_t **output, int nb_samples)
00474 {
00475     if (!output)
00476         return av_audio_fifo_drain(avr->out_fifo, nb_samples);
00477     return av_audio_fifo_read(avr->out_fifo, (void**)output, nb_samples);
00478 }
00479 
00480 unsigned avresample_version(void)
00481 {
00482     return LIBAVRESAMPLE_VERSION_INT;
00483 }
00484 
00485 const char *avresample_license(void)
00486 {
00487 #define LICENSE_PREFIX "libavresample license: "
00488     return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
00489 }
00490 
00491 const char *avresample_configuration(void)
00492 {
00493     return FFMPEG_CONFIGURATION;
00494 }