libmpcodecs/ad_imaadpcm.c

   1 /*
   2     IMA ADPCM Decoder for MPlayer
   3       by Mike Melanson
   4
   5     This file is in charge of decoding all of the various IMA ADPCM data
   6     formats that various entities have created. Details about the data
   7     formats can be found here:
   8       http://www.pcisys.net/~melanson/codecs/
   9
  10     So far, this file handles these formats:
  11       'ima4': IMA ADPCM found in QT files
  12         0x11: IMA ADPCM found in MS AVI/ASF/WAV files
  13         0x61: DK4 ADPCM found in certain AVI files on Sega Saturn CD-ROMs;
  14               note that this is a 'rogue' format number in that it was
  15               never officially registered with Microsoft
  16     0x1100736d: IMA ADPCM coded like in MS AVI/ASF/WAV found in QT files
  17 */
  18
  19 #include <stdio.h>
  20 #include <stdlib.h>
  21 #include <unistd.h>
  22
  23 #include "config.h"
  24 #include "bswap.h"
  25 #include "ad_internal.h"
  26
  27 #define MS_IMA_ADPCM_PREAMBLE_SIZE 4
  28
  29 #define QT_IMA_ADPCM_PREAMBLE_SIZE 2
  30 #define QT_IMA_ADPCM_BLOCK_SIZE 0x22
  31 #define QT_IMA_ADPCM_SAMPLES_PER_BLOCK 64
  32
  33 #define BE_16(x) (be2me_16(*(unsigned short *)(x)))
  34 #define BE_32(x) (be2me_32(*(unsigned int *)(x)))
  35 #define LE_16(x) (le2me_16(*(unsigned short *)(x)))
  36 #define LE_32(x) (le2me_32(*(unsigned int *)(x)))
  37
  38 // pertinent tables for IMA ADPCM
  39 static int adpcm_step[89] =
  40 {
  41   7, 8, 9, 10, 11, 12, 13, 14, 16, 17,
  42   19, 21, 23, 25, 28, 31, 34, 37, 41, 45,
  43   50, 55, 60, 66, 73, 80, 88, 97, 107, 118,
  44   130, 143, 157, 173, 190, 209, 230, 253, 279, 307,
  45   337, 371, 408, 449, 494, 544, 598, 658, 724, 796,
  46   876, 963, 1060, 1166, 1282, 1411, 1552, 1707, 1878, 2066,
  47   2272, 2499, 2749, 3024, 3327, 3660, 4026, 4428, 4871, 5358,
  48   5894, 6484, 7132, 7845, 8630, 9493, 10442, 11487, 12635, 13899,
  49   15289, 16818, 18500, 20350, 22385, 24623, 27086, 29794, 32767
  50 };
  51
  52 static int adpcm_index[16] =
  53 {
  54   -1, -1, -1, -1, 2, 4, 6, 8,
  55   -1, -1, -1, -1, 2, 4, 6, 8
  56 };
  57
  58 // useful macros
  59 // clamp a number between 0 and 88
  60 #define CLAMP_0_TO_88(x)  if (x < 0) x = 0; else if (x > 88) x = 88;
  61 // clamp a number within a signed 16-bit range
  62 #define CLAMP_S16(x)  if (x < -32768) x = -32768; \
  63   else if (x > 32767) x = 32767;
  64 // clamp a number above 16
  65 #define CLAMP_ABOVE_16(x)  if (x < 16) x = 16;
  66 // sign extend a 16-bit value
  67 #define SE_16BIT(x)  if (x & 0x8000) x -= 0x10000;
  68 // sign extend a 4-bit value
  69 #define SE_4BIT(x)  if (x & 0x8) x -= 0x10;
  70
  71 static ad_info_t info =
  72 {
  73         "IMA ADPCM audio decoder",
  74         "imaadpcm",
  75         "Nick Kurshev",
  76         "Mike Melanson",
  77         ""
  78 };
  79
  80 LIBAD_EXTERN(imaadpcm)
  81
  82 static int preinit(sh_audio_t *sh_audio)
  83 {
  84   // not exactly sure what this field is for
  85   sh_audio->audio_out_minsize = 8192;
  86
  87   // if format is "ima4", assume the audio is coming from a QT file which
  88   // indicates constant block size, whereas an AVI/ASF/WAV file will fill
  89   // in this field with 0x11
  90   if ((sh_audio->format == 0x11) || (sh_audio->format == 0x61) ||
  91       (sh_audio->format == 0x1100736d))
  92   {
  93     sh_audio->ds->ss_div = (sh_audio->wf->nBlockAlign -
  94       (MS_IMA_ADPCM_PREAMBLE_SIZE * sh_audio->wf->nChannels)) * 2;
  95     sh_audio->ds->ss_mul = sh_audio->wf->nBlockAlign;
  96   }
  97   else
  98   {
  99     sh_audio->ds->ss_div = QT_IMA_ADPCM_SAMPLES_PER_BLOCK;
 100     sh_audio->ds->ss_mul = QT_IMA_ADPCM_BLOCK_SIZE * sh_audio->wf->nChannels;
 101   }
 102   sh_audio->audio_in_minsize=sh_audio->ds->ss_mul;
 103   return 1;
 104 }
 105
 106 static int init(sh_audio_t *sh_audio)
 107 {
 108   /* IMA-ADPCM 4:1 audio codec:*/
 109   sh_audio->channels=sh_audio->wf->nChannels;
 110   sh_audio->samplerate=sh_audio->wf->nSamplesPerSec;
 111   /* decodes 34 byte -> 64 short*/
 112   sh_audio->i_bps =
 113     (sh_audio->ds->ss_mul * sh_audio->samplerate) / sh_audio->ds->ss_div;
 114
 115   return 1;
 116 }
 117
 118 static void uninit(sh_audio_t *sh_audio)
 119 {
 120 }
 121
 122 static int control(sh_audio_t *sh_audio,int cmd,void* arg, ...)
 123 {
 124   if(cmd==ADCTRL_SKIP_FRAME){
 125     demux_read_data(sh_audio->ds, sh_audio->a_in_buffer,sh_audio->ds->ss_mul);
 126     return CONTROL_TRUE;
 127   }
 128   return CONTROL_UNKNOWN;
 129 }
 130
 131 static void decode_nibbles(unsigned short *output,
 132   int output_size, int channels,
 133   int predictor_l, int index_l,
 134   int predictor_r, int index_r)
 135 {
 136   int step[2];
 137   int predictor[2];
 138   int index[2];
 139   int diff;
 140   int i;
 141   int sign;
 142   int delta;
 143   int channel_number = 0;
 144
 145   step[0] = adpcm_step[index_l];
 146   step[1] = adpcm_step[index_r];
 147   predictor[0] = predictor_l;
 148   predictor[1] = predictor_r;
 149   index[0] = index_l;
 150   index[1] = index_r;
 151
 152   for (i = 0; i < output_size; i++)
 153   {
 154     delta = output[i];
 155
 156     index[channel_number] += adpcm_index[delta];
 157     CLAMP_0_TO_88(index[channel_number]);
 158
 159     sign = delta & 8;
 160     delta = delta & 7;
 161
 162     diff = step[channel_number] >> 3;
 163     if (delta & 4) diff += step[channel_number];
 164     if (delta & 2) diff += step[channel_number] >> 1;
 165     if (delta & 1) diff += step[channel_number] >> 2;
 166
 167     if (sign)
 168       predictor[channel_number] -= diff;
 169     else
 170       predictor[channel_number] += diff;
 171
 172     CLAMP_S16(predictor[channel_number]);
 173     output[i] = predictor[channel_number];
 174     step[channel_number] = adpcm_step[index[channel_number]];
 175
 176     // toggle channel
 177     channel_number ^= channels - 1;
 178
 179   }
 180 }
 181
 182 static int qt_ima_adpcm_decode_block(unsigned short *output,
 183   unsigned char *input, int channels)
 184 {
 185   int initial_predictor_l = 0;
 186   int initial_predictor_r = 0;
 187   int initial_index_l = 0;
 188   int initial_index_r = 0;
 189   int i;
 190
 191   initial_predictor_l = BE_16(&input[0]);
 192   initial_index_l = initial_predictor_l;
 193
 194   // mask, sign-extend, and clamp the predictor portion
 195   initial_predictor_l &= 0xFF80;
 196   SE_16BIT(initial_predictor_l);
 197   CLAMP_S16(initial_predictor_l);
 198
 199   // mask and clamp the index portion
 200   initial_index_l &= 0x7F;
 201   CLAMP_0_TO_88(initial_index_l);
 202
 203   // handle stereo
 204   if (channels > 1)
 205   {
 206     initial_predictor_r = BE_16(&input[QT_IMA_ADPCM_BLOCK_SIZE]);
 207     initial_index_r = initial_predictor_r;
 208
 209     // mask, sign-extend, and clamp the predictor portion
 210     initial_predictor_r &= 0xFF80;
 211     SE_16BIT(initial_predictor_r);
 212     CLAMP_S16(initial_predictor_r);
 213
 214     // mask and clamp the index portion
 215     initial_index_r &= 0x7F;
 216     CLAMP_0_TO_88(initial_index_r);
 217   }
 218
 219   // break apart all of the nibbles in the block
 220   if (channels == 1)
 221     for (i = 0; i < QT_IMA_ADPCM_SAMPLES_PER_BLOCK / 2; i++)
 222     {
 223       output[i * 2 + 0] = input[2 + i] & 0x0F;
 224       output[i * 2 + 1] = input[2 + i] >> 4;
 225     }
 226   else
 227     for (i = 0; i < QT_IMA_ADPCM_SAMPLES_PER_BLOCK / 2 * 2; i++)
 228     {
 229       output[i * 4 + 0] = input[2 + i] & 0x0F;
 230       output[i * 4 + 1] = input[2 + QT_IMA_ADPCM_BLOCK_SIZE + i] & 0x0F;
 231       output[i * 4 + 2] = input[2 + i] >> 4;
 232       output[i * 4 + 3] = input[2 + QT_IMA_ADPCM_BLOCK_SIZE + i] >> 4;
 233     }
 234
 235   decode_nibbles(output,
 236     QT_IMA_ADPCM_SAMPLES_PER_BLOCK * channels, channels,
 237     initial_predictor_l, initial_index_l,
 238     initial_predictor_r, initial_index_r);
 239
 240   return QT_IMA_ADPCM_SAMPLES_PER_BLOCK * channels;
 241 }
 242
 243 static int ms_ima_adpcm_decode_block(unsigned short *output,
 244   unsigned char *input, int channels, int block_size)
 245 {
 246   int predictor_l = 0;
 247   int predictor_r = 0;
 248   int index_l = 0;
 249   int index_r = 0;
 250   int i;
 251   int channel_counter;
 252   int channel_index;
 253   int channel_index_l;
 254   int channel_index_r;
 255
 256   predictor_l = LE_16(&input[0]);
 257   SE_16BIT(predictor_l);
 258   index_l = input[2];
 259   if (channels == 2)
 260   {
 261     predictor_r = LE_16(&input[4]);
 262     SE_16BIT(predictor_r);
 263     index_r = input[6];
 264   }
 265
 266   if (channels == 1)
 267     for (i = 0;
 268       i < (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels); i++)
 269     {
 270       output[i * 2 + 0] = input[MS_IMA_ADPCM_PREAMBLE_SIZE + i] & 0x0F;
 271       output[i * 2 + 1] = input[MS_IMA_ADPCM_PREAMBLE_SIZE + i] >> 4;
 272     }
 273   else
 274   {
 275     // encoded as 8 nibbles (4 bytes) per channel; switch channel every
 276     // 4th byte
 277     channel_counter = 0;
 278     channel_index_l = 0;
 279     channel_index_r = 1;
 280     channel_index = channel_index_l;
 281     for (i = 0;
 282       i < (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels); i++)
 283     {
 284       output[channel_index + 0] =
 285         input[MS_IMA_ADPCM_PREAMBLE_SIZE * 2 + i] & 0x0F;
 286       output[channel_index + 2] =
 287         input[MS_IMA_ADPCM_PREAMBLE_SIZE * 2 + i] >> 4;
 288       channel_index += 4;
 289       channel_counter++;
 290       if (channel_counter == 4)
 291       {
 292         channel_index_l = channel_index;
 293         channel_index = channel_index_r;
 294       }
 295       else if (channel_counter == 8)
 296       {
 297         channel_index_r = channel_index;
 298         channel_index = channel_index_l;
 299         channel_counter = 0;
 300       }
 301     }
 302   }
 303
 304   decode_nibbles(output,
 305     (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels) * 2,
 306     channels,
 307     predictor_l, index_l,
 308     predictor_r, index_r);
 309
 310   return (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels) * 2;
 311 }
 312
 313 static int dk4_ima_adpcm_decode_block(unsigned short *output,
 314   unsigned char *input, int channels, int block_size)
 315 {
 316   int i;
 317   int output_ptr;
 318   int predictor_l = 0;
 319   int predictor_r = 0;
 320   int index_l = 0;
 321   int index_r = 0;
 322
 323   // the first predictor value goes straight to the output
 324   predictor_l = output[0] = LE_16(&input[0]);
 325   SE_16BIT(predictor_l);
 326   index_l = input[2];
 327   if (channels == 2)
 328   {
 329     predictor_r = output[1] = LE_16(&input[4]);
 330     SE_16BIT(predictor_r);
 331     index_r = input[6];
 332   }
 333
 334   output_ptr = channels;
 335   for (i = MS_IMA_ADPCM_PREAMBLE_SIZE * channels; i < block_size; i++)
 336   {
 337     output[output_ptr++] = input[i] >> 4;
 338     output[output_ptr++] = input[i] & 0x0F;
 339   }
 340
 341   decode_nibbles(&output[channels],
 342     (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels) * 2 - channels,
 343     channels,
 344     predictor_l, index_l,
 345     predictor_r, index_r);
 346
 347   return (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels) * 2 - channels;
 348 }
 349
 350 static int decode_audio(sh_audio_t *sh_audio,unsigned char *buf,int minlen,int maxlen)
 351 {
 352   if (demux_read_data(sh_audio->ds, sh_audio->a_in_buffer,
 353     sh_audio->ds->ss_mul) !=
 354     sh_audio->ds->ss_mul)
 355     return -1;
 356
 357   if ((sh_audio->format == 0x11) || (sh_audio->format == 0x1100736d))
 358   {
 359     return 2 * ms_ima_adpcm_decode_block(
 360       (unsigned short*)buf, sh_audio->a_in_buffer, sh_audio->wf->nChannels,
 361       sh_audio->ds->ss_mul);
 362   }
 363   else if (sh_audio->format == 0x61)
 364   {
 365     return 2 * dk4_ima_adpcm_decode_block(
 366       (unsigned short*)buf, sh_audio->a_in_buffer, sh_audio->wf->nChannels,
 367       sh_audio->ds->ss_mul);
 368   }
 369   else
 370   {
 371     return 2 * qt_ima_adpcm_decode_block(
 372       (unsigned short*)buf, sh_audio->a_in_buffer, sh_audio->wf->nChannels);
 373   }
 374 }