libmpcodecs/ad_mpg123.c

   1 /*
   2  * MPEG 1.0/2.0/2.5 audio layer I, II, III decoding with libmpg123
   3  *
   4  * Copyright (C) 2010-2012 Thomas Orgis <thomas@orgis.org>
   5  *
   6  * MPlayer is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License as published by
   8  * the Free Software Foundation; either version 2 of the License, or
   9  * (at your option) any later version.
  10  *
  11  * MPlayer 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
  14  * GNU General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU General Public License along
  17  * with MPlayer; if not, write to the Free Software Foundation, Inc.,
  18  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  19  */
  20
  21 #include <stdio.h>
  22 #include <stdlib.h>
  23 #include <unistd.h>
  24
  25 #include "config.h"
  26
  27 #include "ad_internal.h"
  28
  29 static const ad_info_t info = {
  30     "MPEG 1.0/2.0/2.5 layers I, II, III",
  31     "mpg123",
  32     "Thomas Orgis",
  33     "mpg123.org",
  34     "High-performance decoder using libmpg123."
  35 };
  36
  37 LIBAD_EXTERN(mpg123)
  38
  39 #include "libvo/fastmemcpy.h"
  40
  41 /* Reducing the ifdeffery to two main variants:
  42  *   1. most compatible to any libmpg123 version
  43  *   2. fastest variant with recent libmpg123 (>=1.14)
  44  * Running variant 2 on older libmpg123 versions may work in
  45  * principle, but is not supported.
  46  * So, please leave the check for MPG123_API_VERSION there, m-kay?
  47  */
  48 #include <mpg123.h>
  49
  50 /* Enable faster mode of operation with newer libmpg123, avoiding
  51  * unnecessary memcpy() calls. */
  52 #if (defined MPG123_API_VERSION) && (MPG123_API_VERSION >= 33)
  53 #define AD_MPG123_FRAMEWISE
  54 #endif
  55
  56 /* Switch for updating bitrate info of VBR files. Not essential. */
  57 #define AD_MPG123_MEAN_BITRATE
  58
  59 /* Funny thing, that. I assume I shall use it for selecting mpg123 channels.
  60  * Please correct me if I guessed wrong. */
  61 extern int fakemono;
  62
  63 struct ad_mpg123_context {
  64     mpg123_handle *handle;
  65 #ifdef AD_MPG123_MEAN_BITRATE
  66     /* Running mean for bit rate, stream length estimation. */
  67     float mean_rate;
  68     unsigned int mean_count;
  69     /* Time delay for updates. */
  70     short delay;
  71 #endif
  72     /* If the stream is actually VBR. */
  73     char vbr;
  74 };
  75
  76 /* This initializes libmpg123 and prepares the handle, including funky
  77  * parameters. */
  78 static int preinit(sh_audio_t *sh)
  79 {
  80     int err, flag;
  81     struct ad_mpg123_context *con;
  82     /* Assumption: You always call preinit + init + uninit, on every file.
  83      * But you stop at preinit in case it fails.
  84      * If that is not true, one must ensure not to call mpg123_init / exit
  85      * twice in a row. */
  86     if (mpg123_init() != MPG123_OK)
  87         return 0;
  88
  89     sh->context = malloc(sizeof(struct ad_mpg123_context));
  90     con = sh->context;
  91     /* Auto-choice of optimized decoder (first argument NULL). */
  92     con->handle = mpg123_new(NULL, &err);
  93     if (!con->handle)
  94         goto bad_end;
  95
  96     /* Guessing here: Default value triggers forced upmix of mono to stereo. */
  97     flag = fakemono == 0 ? MPG123_FORCE_STEREO :
  98            fakemono == 1 ? MPG123_MONO_LEFT    :
  99            fakemono == 2 ? MPG123_MONO_RIGHT   : 0;
 100     if (mpg123_param(con->handle, MPG123_ADD_FLAGS, flag, 0.0) != MPG123_OK)
 101         goto bad_end;
 102
 103     /* Basic settings.
 104      * Don't spill messages, enable better resync with non-seekable streams.
 105      * Give both flags individually without error checking to keep going with
 106      * old libmpg123. Generally, it is not fatal if the flags are not
 107      * honored */
 108     mpg123_param(con->handle, MPG123_ADD_FLAGS, MPG123_QUIET, 0.0);
 109     /* Do not bail out on malformed streams at all.
 110      * MPlayer does not handle a decoder throwing the towel on crappy input. */
 111     mpg123_param(con->handle, MPG123_RESYNC_LIMIT, -1, 0.0);
 112
 113     /* Open decisions: Configure libmpg123 to force encoding (or stay open about
 114      * library builds that support only float or int32 output), (de)configure
 115      * gapless decoding (won't work with seeking in MPlayer, though).
 116      * Don't forget to eventually enable ReplayGain/RVA support, too.
 117      * Let's try to run with the default for now. */
 118
 119     /* That would produce floating point output.
 120      * You can get 32 and 24 bit ints, even 8 bit via format matrix. */
 121     /* mpg123_param(con->handle, MPG123_ADD_FLAGS, MPG123_FORCE_FLOAT, 0.); */
 122
 123     /* Example for RVA choice (available since libmpg123 1.0.0):
 124     mpg123_param(con->handle, MPG123_RVA, MPG123_RVA_MIX, 0.0) */
 125
 126 #ifdef AD_MPG123_FRAMEWISE
 127     /* Prevent funky automatic resampling.
 128      * This way, we can be sure that one frame will never produce
 129      * more than 1152 stereo samples. */
 130     mpg123_param(con->handle, MPG123_REMOVE_FLAGS, MPG123_AUTO_RESAMPLE, 0.);
 131 #else
 132     /* Older mpg123 is vulnerable to concatenated streams when gapless cutting
 133      * is enabled (will only play the jingle of a badly constructed radio
 134      * stream). The versions using framewise decoding are fine with that. */
 135     mpg123_param(con->handle, MPG123_REMOVE_FLAGS, MPG123_GAPLESS, 0.);
 136 #endif
 137
 138     return 1;
 139
 140   bad_end:
 141     if (!con->handle)
 142         mp_msg(MSGT_DECAUDIO, MSGL_ERR, "mpg123 preinit error: %s\n",
 143                mpg123_plain_strerror(err));
 144     else
 145         mp_msg(MSGT_DECAUDIO, MSGL_ERR, "mpg123 preinit error: %s\n",
 146                mpg123_strerror(con->handle));
 147
 148     if (con->handle)
 149         mpg123_delete(con->handle);
 150     mpg123_exit();
 151     free(sh->context);
 152     sh->context = NULL;
 153     return 0;
 154 }
 155
 156 /* Compute bitrate from frame size. */
 157 static int compute_bitrate(struct mpg123_frameinfo *i)
 158 {
 159     static const int samples_per_frame[4][4] = {
 160         {-1, 384, 1152, 1152},  /* MPEG 1 */
 161         {-1, 384, 1152,  576},  /* MPEG 2 */
 162         {-1, 384, 1152,  576},  /* MPEG 2.5 */
 163         {-1,  -1,   -1,   -1},  /* Unknown */
 164     };
 165     return (int) ((i->framesize + 4) * 8 * i->rate * 0.001 /
 166                   samples_per_frame[i->version][i->layer] + 0.5);
 167 }
 168
 169 /* Opted against the header printout from old mp3lib, too much
 170  * irrelevant info. This is modelled after the mpg123 app's
 171  * standard output line.
 172  * If more verbosity is demanded, one can add more detail and
 173  * also throw in ID3v2 info which libmpg123 collects anyway. */
 174 static void print_header_compact(struct mpg123_frameinfo *i)
 175 {
 176     static const char *smodes[5] = {
 177         "stereo", "joint-stereo", "dual-channel", "mono", "invalid"
 178     };
 179     static const char *layers[4] = {
 180         "Unknown", "I", "II", "III"
 181     };
 182     static const char *versions[4] = {
 183         "1.0", "2.0", "2.5", "x.x"
 184     };
 185
 186     mp_msg(MSGT_DECAUDIO, MSGL_V, "MPEG %s layer %s, ",
 187            versions[i->version], layers[i->layer]);
 188     switch (i->vbr) {
 189     case MPG123_CBR:
 190         if (i->bitrate)
 191             mp_msg(MSGT_DECAUDIO, MSGL_V, "%d kbit/s", i->bitrate);
 192         else
 193             mp_msg(MSGT_DECAUDIO, MSGL_V, "%d kbit/s (free format)",
 194                    compute_bitrate(i));
 195         break;
 196     case MPG123_VBR:
 197         mp_msg(MSGT_DECAUDIO, MSGL_V, "VBR");
 198         break;
 199     case MPG123_ABR:
 200         mp_msg(MSGT_DECAUDIO, MSGL_V, "%d kbit/s ABR", i->abr_rate);
 201         break;
 202     default:
 203         mp_msg(MSGT_DECAUDIO, MSGL_V, "???");
 204     }
 205     mp_msg(MSGT_DECAUDIO, MSGL_V, ", %ld Hz %s\n", i->rate,
 206            smodes[i->mode]);
 207 }
 208
 209 /* This tries to extract a requested amount of decoded data.
 210  * Even when you request 0 bytes, it will feed enough input so that
 211  * the decoder _could_ have delivered something.
 212  * Returns byte count >= 0, -1 on error.
 213  *
 214  * Thoughts on exact pts keeping:
 215  * We have to assume that MPEG frames are cut in pieces by packet boundaries.
 216  * Also, it might be possible that the first packet does not contain enough
 217  * data to ensure initial stream sync... or re-sync on erroneous streams.
 218  * So we need something robust to relate the decoded byte count to the correct
 219  * time stamp. This is tricky, though. From the outside, you cannot tell if,
 220  * after having fed two packets until the first output arrives, one should
 221  * start counting from the first packet's pts or the second packet's.
 222  * So, let's just count from the last fed package's pts. If the packets are
 223  * exactly cut to MPEG frames, this will cause one frame mismatch in the
 224  * beginning (when mpg123 peeks ahead for the following header), but will
 225  * be corrected with the third frame already. One might add special code to
 226  * not increment the base pts past the first packet's after a resync before
 227  * the first decoded bytes arrived. */
 228 static int decode_a_bit(sh_audio_t *sh, unsigned char *buf, int count)
 229 {
 230     int ret = MPG123_OK;
 231     int got = 0;
 232     struct ad_mpg123_context *con = sh->context;
 233
 234     /* There will be one MPG123_NEW_FORMAT message on first open.
 235      * This will be handled in init(). */
 236     do {
 237         size_t got_now = 0;
 238
 239         /* Feed the decoder. This will only fire from the second round on. */
 240         if (ret == MPG123_NEED_MORE) {
 241             int incount;
 242             double pts;
 243             unsigned char *inbuf;
 244             /* Feed more input data. */
 245             incount = ds_get_packet_pts(sh->ds, &inbuf, &pts);
 246             if (incount <= 0)
 247                 break;          /* Apparently that's it. EOF. */
 248
 249             /* Next bytes from that presentation time. */
 250             if (pts != MP_NOPTS_VALUE) {
 251                 sh->pts       = pts;
 252                 sh->pts_bytes = 0;
 253             }
 254
 255 #ifdef AD_MPG123_FRAMEWISE
 256             /* Have to use mpg123_feed() to avoid decoding here. */
 257             ret = mpg123_feed(con->handle, inbuf, incount);
 258 #else
 259             /* Do not use mpg123_feed(), added in later libmpg123 versions. */
 260             ret = mpg123_decode(con->handle, inbuf, incount, NULL, 0, NULL);
 261 #endif
 262             if (ret == MPG123_ERR)
 263                 break;
 264         }
 265         /* Theoretically, mpg123 could return MPG123_DONE, so be prepared.
 266          * Should not happen in our usage, but it is a valid return code. */
 267         else if (ret == MPG123_ERR || ret == MPG123_DONE)
 268             break;
 269
 270         /* Try to decode a bit. This is the return value that counts
 271          * for the loop condition. */
 272 #ifdef AD_MPG123_FRAMEWISE
 273         if (!buf) { /* fake call just for feeding to get format */
 274             ret = mpg123_getformat(con->handle, NULL, NULL, NULL);
 275         } else { /* This is the decoding. One frame at a time. */
 276             ret = mpg123_replace_buffer(con->handle, buf, count);
 277             if (ret == MPG123_OK)
 278                 ret = mpg123_decode_frame(con->handle, NULL, NULL, &got_now);
 279         }
 280 #else
 281         ret = mpg123_decode(con->handle, NULL, 0, buf + got, count - got,
 282                             &got_now);
 283 #endif
 284
 285         got += got_now;
 286         sh->pts_bytes += got_now;
 287
 288 #ifdef AD_MPG123_FRAMEWISE
 289     } while (ret == MPG123_NEED_MORE || (got == 0 && count != 0));
 290 #else
 291     } while (ret == MPG123_NEED_MORE || got < count);
 292 #endif
 293
 294     if (ret == MPG123_ERR) {
 295         mp_msg(MSGT_DECAUDIO, MSGL_ERR, "mpg123 decoding failed: %s\n",
 296                mpg123_strerror(con->handle));
 297         mpg123_close(con->handle);
 298         return -1;
 299     }
 300
 301     return got;
 302 }
 303
 304 /* Close, reopen stream. Feed data until we know the format of the stream.
 305  * 1 on success, 0 on error */
 306 static int reopen_stream(sh_audio_t *sh)
 307 {
 308     struct ad_mpg123_context *con = (struct ad_mpg123_context*) sh->context;
 309
 310     mpg123_close(con->handle);
 311     /* No resetting of the context:
 312      * We do not want to loose the mean bitrate data. */
 313
 314     /* Open and make sure we have fed enough data to get stream properties. */
 315     if (MPG123_OK == mpg123_open_feed(con->handle) &&
 316         /* Feed data until mpg123 is ready (has found stream beginning). */
 317         !decode_a_bit(sh, NULL, 0)) {
 318         return 1;
 319     } else {
 320         mp_msg(MSGT_DECAUDIO, MSGL_ERR,
 321                "mpg123 failed to reopen stream: %s\n",
 322                mpg123_strerror(con->handle));
 323         mpg123_close(con->handle);
 324         return 0;
 325     }
 326 }
 327
 328 /* Now we really start accessing some data and determining file format.
 329  * Paranoia note: The mpg123_close() on errors is not really necessary,
 330  * But it ensures that we don't accidentally continue decoding with a
 331  * bad state (possibly interpreting the format badly or whatnot). */
 332 static int init(sh_audio_t *sh)
 333 {
 334     long rate    = 0;
 335     int channels = 0;
 336     int encoding = 0;
 337     mpg123_id3v2 *v2;
 338     struct mpg123_frameinfo finfo;
 339     struct ad_mpg123_context *con = sh->context;
 340
 341     /* We're open about any output format that libmpg123 will suggest.
 342      * Note that a standard build will always default to 16 bit signed and
 343      * the native sample rate of the file. */
 344     if (MPG123_OK == mpg123_format_all(con->handle) &&
 345         reopen_stream(sh) &&
 346         MPG123_OK == mpg123_getformat(con->handle, &rate, &channels, &encoding) &&
 347         /* Forbid the format to change later on. */
 348         MPG123_OK == mpg123_format_none(con->handle) &&
 349         MPG123_OK == mpg123_format(con->handle, rate, channels, encoding) &&
 350         /* Get MPEG header info. */
 351         MPG123_OK == mpg123_info(con->handle, &finfo) &&
 352         /* Since we queried format, mpg123 should have read past ID3v2 tags.
 353          * We need to decide if printing of UTF-8 encoded text info is wanted. */
 354         MPG123_OK == mpg123_id3(con->handle, NULL, &v2)) {
 355         /* If we are here, we passed all hurdles. Yay! Extract the info. */
 356         print_header_compact(&finfo);
 357         /* Do we want to print out the UTF-8 Id3v2 info?
 358         if (v2)
 359             print_id3v2(v2); */
 360
 361         /* Have kb/s, want B/s
 362          * For VBR, the first frame will be a bad estimate. */
 363         sh->i_bps = (finfo.bitrate ? finfo.bitrate : compute_bitrate(&finfo))
 364                     * 1000 / 8;
 365 #ifdef AD_MPG123_MEAN_BITRATE
 366         con->delay      = 1;
 367         con->mean_rate  = 0.;
 368         con->mean_count = 0;
 369 #endif
 370         con->vbr = (finfo.vbr != MPG123_CBR);
 371         sh->channels   = channels;
 372         sh->samplerate = rate;
 373         /* Without external force, mpg123 will always choose signed encoding,
 374          * and non-16-bit only on builds that don't support it.
 375          * Be reminded that it doesn't matter to the MPEG file what encoding
 376          * is produced from it. */
 377         switch (encoding) {
 378         case MPG123_ENC_SIGNED_8:
 379             sh->sample_format = AF_FORMAT_S8;
 380             sh->samplesize    = 1;
 381             break;
 382         case MPG123_ENC_SIGNED_16:
 383             sh->sample_format = AF_FORMAT_S16_NE;
 384             sh->samplesize    = 2;
 385             break;
 386         /* To stay compatible with the oldest libmpg123 headers, do not rely
 387          * on float and 32 bit encoding symbols being defined.
 388          * Those formats came later */
 389         case 0x1180: /* MPG123_ENC_SIGNED_32 */
 390             sh->sample_format = AF_FORMAT_S32_NE;
 391             sh->samplesize    = 4;
 392             break;
 393         case 0x200: /* MPG123_ENC_FLOAT_32 */
 394             sh->sample_format = AF_FORMAT_FLOAT_NE;
 395             sh->samplesize    = 4;
 396             break;
 397         default:
 398             mp_msg(MSGT_DECAUDIO, MSGL_ERR,
 399                    "Bad encoding from mpg123: %i.\n", encoding);
 400             mpg123_close(con->handle);
 401             return 0;
 402         }
 403 #ifdef AD_MPG123_FRAMEWISE
 404         /* Going to decode directly to MPlayer's memory. It is important
 405          * to have MPG123_AUTO_RESAMPLE disabled for the buffer size
 406          * being an all-time limit. */
 407         sh->audio_out_minsize = 1152 * 2 * sh->samplesize;
 408 #endif
 409
 410         return 1;
 411     } else {
 412         mp_msg(MSGT_DECAUDIO, MSGL_ERR, "mpg123 init error: %s\n",
 413                mpg123_strerror(con->handle));
 414         mpg123_close(con->handle);
 415         return 0;
 416     }
 417 }
 418
 419 static void uninit(sh_audio_t *sh)
 420 {
 421     struct ad_mpg123_context *con = (struct ad_mpg123_context*) sh->context;
 422
 423     mpg123_close(con->handle);
 424     mpg123_delete(con->handle);
 425     free(sh->context);
 426     sh->context = NULL;
 427     mpg123_exit();
 428 }
 429
 430 #ifdef AD_MPG123_MEAN_BITRATE
 431 /* Update mean bitrate. This could be dropped if accurate time display
 432  * on audio file playback is not desired. */
 433 static void update_info(sh_audio_t *sh)
 434 {
 435     struct ad_mpg123_context *con = sh->context;
 436     if (con->vbr && --con->delay < 1) {
 437         struct mpg123_frameinfo finfo;
 438         if (MPG123_OK == mpg123_info(con->handle, &finfo)) {
 439             if (++con->mean_count > ((unsigned int) -1) / 2)
 440                 con->mean_count = ((unsigned int) -1) / 4;
 441
 442             /* Might not be numerically optimal, but works fine enough. */
 443             con->mean_rate = ((con->mean_count - 1) * con->mean_rate +
 444                               finfo.bitrate) / con->mean_count;
 445             sh->i_bps = (int) (con->mean_rate * 1000 / 8);
 446
 447             con->delay = 10;
 448         }
 449     }
 450 }
 451 #endif
 452
 453 static int decode_audio(sh_audio_t *sh, unsigned char *buf, int minlen,
 454                         int maxlen)
 455 {
 456     int bytes;
 457
 458     bytes = decode_a_bit(sh, buf, maxlen);
 459     if (bytes == 0)
 460         return -1;              /* EOF */
 461
 462 #ifdef AD_MPG123_MEAN_BITRATE
 463     update_info(sh);
 464 #endif
 465     return bytes;
 466 }
 467
 468 static int control(sh_audio_t *sh, int cmd, void *arg, ...)
 469 {
 470     switch (cmd) {
 471     case ADCTRL_RESYNC_STREAM:
 472         /* Close/reopen the stream for mpg123 to make sure it doesn't
 473          * think that it still knows the exact stream position.
 474          * Otherwise, we would have funny effects from the gapless code.
 475          * Oh, and it helps to minimize artifacts from jumping in the stream. */
 476         if (reopen_stream(sh)) {
 477 #ifdef AD_MPG123_MEAN_BITRATE
 478             update_info(sh);
 479 #endif
 480             return CONTROL_TRUE;
 481         } else {
 482             /* MPlayer ignores this case! It just keeps on decoding.
 483              * So we have to make sure resync never fails ... */
 484             mp_msg(MSGT_DECAUDIO, MSGL_ERR,
 485                    "mpg123 cannot reopen stream for resync.\n");
 486             return CONTROL_FALSE;
 487         }
 488         break;
 489     }
 490     return CONTROL_UNKNOWN;
 491 }