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   sh_audio->samplesize=2;
 115
 116   return 1;
 117 }
 118
 119 static void uninit(sh_audio_t *sh_audio)
 120 {
 121 }
 122
 123 static int control(sh_audio_t *sh_audio,int cmd,void* arg, ...)
 124 {
 125   if(cmd==ADCTRL_SKIP_FRAME){
 126     demux_read_data(sh_audio->ds, sh_audio->a_in_buffer,sh_audio->ds->ss_mul);
 127     return CONTROL_TRUE;
 128   }
 129   return CONTROL_UNKNOWN;
 130 }
 131
 132 static void decode_nibbles(unsigned short *output,
 133   int output_size, int channels,
 134   int predictor_l, int index_l,
 135   int predictor_r, int index_r)
 136 {
 137   int step[2];
 138   int predictor[2];
 139   int index[2];
 140   int diff;
 141   int i;
 142   int sign;
 143   int delta;
 144   int channel_number = 0;
 145
 146   step[0] = adpcm_step[index_l];
 147   step[1] = adpcm_step[index_r];
 148   predictor[0] = predictor_l;
 149   predictor[1] = predictor_r;
 150   index[0] = index_l;
 151   index[1] = index_r;
 152
 153   for (i = 0; i < output_size; i++)
 154   {
 155     delta = output[i];
 156
 157     index[channel_number] += adpcm_index[delta];
 158     CLAMP_0_TO_88(index[channel_number]);
 159
 160     sign = delta & 8;
 161     delta = delta & 7;
 162
 163     diff = step[channel_number] >> 3;
 164     if (delta & 4) diff += step[channel_number];
 165     if (delta & 2) diff += step[channel_number] >> 1;
 166     if (delta & 1) diff += step[channel_number] >> 2;
 167
 168     if (sign)
 169       predictor[channel_number] -= diff;
 170     else
 171       predictor[channel_number] += diff;
 172
 173     CLAMP_S16(predictor[channel_number]);
 174     output[i] = predictor[channel_number];
 175     step[channel_number] = adpcm_step[index[channel_number]];
 176
 177     // toggle channel
 178     channel_number ^= channels - 1;
 179
 180   }
 181 }
 182
 183 static int qt_ima_adpcm_decode_block(unsigned short *output,
 184   unsigned char *input, int channels)
 185 {
 186   int initial_predictor_l = 0;
 187   int initial_predictor_r = 0;
 188   int initial_index_l = 0;
 189   int initial_index_r = 0;
 190   int i;
 191
 192   initial_predictor_l = BE_16(&input[0]);
 193   initial_index_l = initial_predictor_l;
 194
 195   // mask, sign-extend, and clamp the predictor portion
 196   initial_predictor_l &= 0xFF80;
 197   SE_16BIT(initial_predictor_l);
 198   CLAMP_S16(initial_predictor_l);
 199
 200   // mask and clamp the index portion
 201   initial_index_l &= 0x7F;
 202   CLAMP_0_TO_88(initial_index_l);
 203
 204   // handle stereo
 205   if (channels > 1)
 206   {
 207     initial_predictor_r = BE_16(&input[QT_IMA_ADPCM_BLOCK_SIZE]);
 208     initial_index_r = initial_predictor_r;
 209
 210     // mask, sign-extend, and clamp the predictor portion
 211     initial_predictor_r &= 0xFF80;
 212     SE_16BIT(initial_predictor_r);
 213     CLAMP_S16(initial_predictor_r);
 214
 215     // mask and clamp the index portion
 216     initial_index_r &= 0x7F;
 217     CLAMP_0_TO_88(initial_index_r);
 218   }
 219
 220   // break apart all of the nibbles in the block
 221   if (channels == 1)
 222     for (i = 0; i < QT_IMA_ADPCM_SAMPLES_PER_BLOCK / 2; i++)
 223     {
 224       output[i * 2 + 0] = input[2 + i] & 0x0F;
 225       output[i * 2 + 1] = input[2 + i] >> 4;
 226     }
 227   else
 228     for (i = 0; i < QT_IMA_ADPCM_SAMPLES_PER_BLOCK / 2; i++)
 229     {
 230       output[i * 4 + 0] = input[2 + i] & 0x0F;
 231       output[i * 4 + 1] = input[2 + QT_IMA_ADPCM_BLOCK_SIZE + i] & 0x0F;
 232       output[i * 4 + 2] = input[2 + i] >> 4;
 233       output[i * 4 + 3] = input[2 + QT_IMA_ADPCM_BLOCK_SIZE + i] >> 4;
 234     }
 235
 236   decode_nibbles(output,
 237     QT_IMA_ADPCM_SAMPLES_PER_BLOCK * channels, channels,
 238     initial_predictor_l, initial_index_l,
 239     initial_predictor_r, initial_index_r);
 240
 241   return QT_IMA_ADPCM_SAMPLES_PER_BLOCK * channels;
 242 }
 243
 244 static int ms_ima_adpcm_decode_block(unsigned short *output,
 245   unsigned char *input, int channels, int block_size)
 246 {
 247   int predictor_l = 0;
 248   int predictor_r = 0;
 249   int index_l = 0;
 250   int index_r = 0;
 251   int i;
 252   int channel_counter;
 253   int channel_index;
 254   int channel_index_l;
 255   int channel_index_r;
 256
 257   predictor_l = LE_16(&input[0]);
 258   SE_16BIT(predictor_l);
 259   index_l = input[2];
 260   if (channels == 2)
 261   {
 262     predictor_r = LE_16(&input[4]);
 263     SE_16BIT(predictor_r);
 264     index_r = input[6];
 265   }
 266
 267   if (channels == 1)
 268     for (i = 0;
 269       i < (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels); i++)
 270     {
 271       output[i * 2 + 0] = input[MS_IMA_ADPCM_PREAMBLE_SIZE + i] & 0x0F;
 272       output[i * 2 + 1] = input[MS_IMA_ADPCM_PREAMBLE_SIZE + i] >> 4;
 273     }
 274   else
 275   {
 276     // encoded as 8 nibbles (4 bytes) per channel; switch channel every
 277     // 4th byte
 278     channel_counter = 0;
 279     channel_index_l = 0;
 280     channel_index_r = 1;
 281     channel_index = channel_index_l;
 282     for (i = 0;
 283       i < (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels); i++)
 284     {
 285       output[channel_index + 0] =
 286         input[MS_IMA_ADPCM_PREAMBLE_SIZE * 2 + i] & 0x0F;
 287       output[channel_index + 2] =
 288         input[MS_IMA_ADPCM_PREAMBLE_SIZE * 2 + i] >> 4;
 289       channel_index += 4;
 290       channel_counter++;
 291       if (channel_counter == 4)
 292       {
 293         channel_index_l = channel_index;
 294         channel_index = channel_index_r;
 295       }
 296       else if (channel_counter == 8)
 297       {
 298         channel_index_r = channel_index;
 299         channel_index = channel_index_l;
 300         channel_counter = 0;
 301       }
 302     }
 303   }
 304
 305   decode_nibbles(output,
 306     (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels) * 2,
 307     channels,
 308     predictor_l, index_l,
 309     predictor_r, index_r);
 310
 311   return (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels) * 2;
 312 }
 313
 314 static int dk4_ima_adpcm_decode_block(unsigned short *output,
 315   unsigned char *input, int channels, int block_size)
 316 {
 317   int i;
 318   int output_ptr;
 319   int predictor_l = 0;
 320   int predictor_r = 0;
 321   int index_l = 0;
 322   int index_r = 0;
 323
 324   // the first predictor value goes straight to the output
 325   predictor_l = output[0] = LE_16(&input[0]);
 326   SE_16BIT(predictor_l);
 327   index_l = input[2];
 328   if (channels == 2)
 329   {
 330     predictor_r = output[1] = LE_16(&input[4]);
 331     SE_16BIT(predictor_r);
 332     index_r = input[6];
 333   }
 334
 335   output_ptr = channels;
 336   for (i = MS_IMA_ADPCM_PREAMBLE_SIZE * channels; i < block_size; i++)
 337   {
 338     output[output_ptr++] = input[i] >> 4;
 339     output[output_ptr++] = input[i] & 0x0F;
 340   }
 341
 342   decode_nibbles(&output[channels],
 343     (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels) * 2 - channels,
 344     channels,
 345     predictor_l, index_l,
 346     predictor_r, index_r);
 347
 348   return (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels) * 2 - channels;
 349 }
 350
 351 static int decode_audio(sh_audio_t *sh_audio,unsigned char *buf,int minlen,int maxlen)
 352 {
 353   if (demux_read_data(sh_audio->ds, sh_audio->a_in_buffer,
 354     sh_audio->ds->ss_mul) !=
 355     sh_audio->ds->ss_mul)
 356     return -1;
 357
 358   if ((sh_audio->format == 0x11) || (sh_audio->format == 0x1100736d))
 359   {
 360     return 2 * ms_ima_adpcm_decode_block(
 361       (unsigned short*)buf, sh_audio->a_in_buffer, sh_audio->wf->nChannels,
 362       sh_audio->ds->ss_mul);
 363   }
 364   else if (sh_audio->format == 0x61)
 365   {
 366     return 2 * dk4_ima_adpcm_decode_block(
 367       (unsigned short*)buf, sh_audio->a_in_buffer, sh_audio->wf->nChannels,
 368       sh_audio->ds->ss_mul);
 369   }
 370   else
 371   {
 372     return 2 * qt_ima_adpcm_decode_block(
 373       (unsigned short*)buf, sh_audio->a_in_buffer, sh_audio->wf->nChannels);
 374   }
 375 }