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 #include <inttypes.h>
  23
  24 #include "config.h"
  25 #include "libavutil/common.h"
  26 #include "mpbswap.h"
  27 #include "ad_internal.h"
  28
  29 #define MS_IMA_ADPCM_PREAMBLE_SIZE 4
  30
  31 #define QT_IMA_ADPCM_PREAMBLE_SIZE 2
  32 #define QT_IMA_ADPCM_BLOCK_SIZE 0x22
  33 #define QT_IMA_ADPCM_SAMPLES_PER_BLOCK 64
  34
  35 #define BE_16(x) (be2me_16(*(unsigned short *)(x)))
  36 #define LE_16(x) (le2me_16(*(unsigned short *)(x)))
  37
  38 // pertinent tables for IMA ADPCM
  39 static const int16_t 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 const int8_t adpcm_index[8] =
  53 {
  54   -1, -1, -1, -1, 2, 4, 6, 8,
  55 };
  56
  57 // useful macros
  58 // clamp a number between 0 and 88
  59 #define CLAMP_0_TO_88(x) x = av_clip(x, 0, 88);
  60 // clamp a number within a signed 16-bit range
  61 #define CLAMP_S16(x) x = av_clip_int16(x);
  62 // clamp a number above 16
  63 #define CLAMP_ABOVE_16(x)  if (x < 16) x = 16;
  64
  65 static ad_info_t info =
  66 {
  67         "IMA ADPCM audio decoder",
  68         "imaadpcm",
  69         "Nick Kurshev",
  70         "Mike Melanson",
  71         ""
  72 };
  73
  74 LIBAD_EXTERN(imaadpcm)
  75
  76 static int preinit(sh_audio_t *sh_audio)
  77 {
  78   // not exactly sure what this field is for
  79   sh_audio->audio_out_minsize = 8192;
  80
  81   // if format is "ima4", assume the audio is coming from a QT file which
  82   // indicates constant block size, whereas an AVI/ASF/WAV file will fill
  83   // in this field with 0x11
  84   if ((sh_audio->format == 0x11) || (sh_audio->format == 0x61) ||
  85       (sh_audio->format == 0x1100736d))
  86   {
  87     sh_audio->ds->ss_div = (sh_audio->wf->nBlockAlign -
  88       (MS_IMA_ADPCM_PREAMBLE_SIZE * sh_audio->wf->nChannels)) * 2;
  89     sh_audio->ds->ss_mul = sh_audio->wf->nBlockAlign;
  90   }
  91   else
  92   {
  93     sh_audio->ds->ss_div = QT_IMA_ADPCM_SAMPLES_PER_BLOCK;
  94     sh_audio->ds->ss_mul = QT_IMA_ADPCM_BLOCK_SIZE * sh_audio->wf->nChannels;
  95   }
  96   sh_audio->audio_in_minsize=sh_audio->ds->ss_mul;
  97   return 1;
  98 }
  99
 100 static int init(sh_audio_t *sh_audio)
 101 {
 102   /* IMA-ADPCM 4:1 audio codec:*/
 103   sh_audio->channels=sh_audio->wf->nChannels;
 104   sh_audio->samplerate=sh_audio->wf->nSamplesPerSec;
 105   /* decodes 34 byte -> 64 short*/
 106   sh_audio->i_bps =
 107     (sh_audio->ds->ss_mul * sh_audio->samplerate) / sh_audio->ds->ss_div;
 108   sh_audio->samplesize=2;
 109
 110   return 1;
 111 }
 112
 113 static void uninit(sh_audio_t *sh_audio)
 114 {
 115 }
 116
 117 static int control(sh_audio_t *sh_audio,int cmd,void* arg, ...)
 118 {
 119   if(cmd==ADCTRL_SKIP_FRAME){
 120     demux_read_data(sh_audio->ds, sh_audio->a_in_buffer,sh_audio->ds->ss_mul);
 121     return CONTROL_TRUE;
 122   }
 123   return CONTROL_UNKNOWN;
 124 }
 125
 126 static void decode_nibbles(unsigned short *output,
 127   int output_size, int channels,
 128   int predictor[2], int index[2])
 129 {
 130   int step[2];
 131   int i;
 132   int sign;
 133   int delta;
 134   int channel_number = 0;
 135
 136   step[0] = adpcm_step[index[0]];
 137   step[1] = adpcm_step[index[1]];
 138
 139   for (i = 0; i < output_size; i++)
 140   {
 141     delta = output[i];
 142     sign = delta & 8;
 143     delta = delta & 7;
 144
 145     index[channel_number] += adpcm_index[delta];
 146     CLAMP_0_TO_88(index[channel_number]);
 147
 148     delta = 2 * delta + 1;
 149     if (sign) delta = -delta;
 150
 151     predictor[channel_number] += (delta * step[channel_number]) >> 3;
 152
 153     CLAMP_S16(predictor[channel_number]);
 154     output[i] = predictor[channel_number];
 155     step[channel_number] = adpcm_step[index[channel_number]];
 156
 157     // toggle channel
 158     channel_number ^= channels - 1;
 159
 160   }
 161 }
 162
 163 static int qt_ima_adpcm_decode_block(unsigned short *output,
 164   unsigned char *input, int channels, int block_size)
 165 {
 166   int initial_predictor[2];
 167   int initial_index[2];
 168   int i;
 169
 170   if (channels != 1) channels = 2;
 171   if (block_size < channels * QT_IMA_ADPCM_BLOCK_SIZE)
 172     return -1;
 173
 174   for (i = 0; i < channels; i++) {
 175     initial_index[i] = initial_predictor[i] = (int16_t)BE_16(&input[i * QT_IMA_ADPCM_BLOCK_SIZE]);
 176
 177     // mask, sign-extend, and clamp the predictor portion
 178     initial_predictor[i] &= ~0x7F;
 179     CLAMP_S16(initial_predictor[i]);
 180
 181     // mask and clamp the index portion
 182     initial_index[i] &= 0x7F;
 183     CLAMP_0_TO_88(initial_index[i]);
 184   }
 185
 186   // break apart all of the nibbles in the block
 187   if (channels == 1)
 188     for (i = 0; i < QT_IMA_ADPCM_SAMPLES_PER_BLOCK / 2; i++)
 189     {
 190       output[i * 2 + 0] = input[2 + i] & 0x0F;
 191       output[i * 2 + 1] = input[2 + i] >> 4;
 192     }
 193   else
 194     for (i = 0; i < QT_IMA_ADPCM_SAMPLES_PER_BLOCK / 2; i++)
 195     {
 196       output[i * 4 + 0] = input[2 + i] & 0x0F;
 197       output[i * 4 + 1] = input[2 + QT_IMA_ADPCM_BLOCK_SIZE + i] & 0x0F;
 198       output[i * 4 + 2] = input[2 + i] >> 4;
 199       output[i * 4 + 3] = input[2 + QT_IMA_ADPCM_BLOCK_SIZE + i] >> 4;
 200     }
 201
 202   decode_nibbles(output,
 203     QT_IMA_ADPCM_SAMPLES_PER_BLOCK * channels, channels,
 204     initial_predictor, initial_index);
 205
 206   return QT_IMA_ADPCM_SAMPLES_PER_BLOCK * channels;
 207 }
 208
 209 static int ms_ima_adpcm_decode_block(unsigned short *output,
 210   unsigned char *input, int channels, int block_size)
 211 {
 212   int predictor[2];
 213   int index[2];
 214   int i;
 215   int channel_counter;
 216   int channel_index;
 217   int channel_index_l;
 218   int channel_index_r;
 219
 220   if (channels != 1) channels = 2;
 221   if (block_size < MS_IMA_ADPCM_PREAMBLE_SIZE * channels)
 222     return -1;
 223
 224   for (i = 0; i < channels; i++) {
 225     predictor[i] = (int16_t)LE_16(&input[i * 4]);
 226     index[i] = input[i * 4 + 2];
 227   }
 228
 229   if (channels == 1)
 230     for (i = 0;
 231       i < (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels); i++)
 232     {
 233       output[i * 2 + 0] = input[MS_IMA_ADPCM_PREAMBLE_SIZE + i] & 0x0F;
 234       output[i * 2 + 1] = input[MS_IMA_ADPCM_PREAMBLE_SIZE + i] >> 4;
 235     }
 236   else
 237   {
 238     // encoded as 8 nibbles (4 bytes) per channel; switch channel every
 239     // 4th byte
 240     channel_counter = 0;
 241     channel_index_l = 0;
 242     channel_index_r = 1;
 243     channel_index = channel_index_l;
 244     for (i = 0;
 245       i < (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels); i++)
 246     {
 247       output[channel_index + 0] =
 248         input[MS_IMA_ADPCM_PREAMBLE_SIZE * 2 + i] & 0x0F;
 249       output[channel_index + 2] =
 250         input[MS_IMA_ADPCM_PREAMBLE_SIZE * 2 + i] >> 4;
 251       channel_index += 4;
 252       channel_counter++;
 253       if (channel_counter == 4)
 254       {
 255         channel_index_l = channel_index;
 256         channel_index = channel_index_r;
 257       }
 258       else if (channel_counter == 8)
 259       {
 260         channel_index_r = channel_index;
 261         channel_index = channel_index_l;
 262         channel_counter = 0;
 263       }
 264     }
 265   }
 266
 267   decode_nibbles(output,
 268     (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels) * 2,
 269     channels,
 270     predictor, index);
 271
 272   return (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels) * 2;
 273 }
 274
 275 static int dk4_ima_adpcm_decode_block(unsigned short *output,
 276   unsigned char *input, int channels, int block_size)
 277 {
 278   int i;
 279   int output_ptr;
 280   int predictor[2];
 281   int index[2];
 282
 283   if (channels != 1) channels = 2;
 284   if (block_size < MS_IMA_ADPCM_PREAMBLE_SIZE * channels)
 285     return -1;
 286
 287   for (i = 0; i < channels; i++) {
 288     // the first predictor value goes straight to the output
 289     predictor[i] = output[i] = (int16_t)LE_16(&input[i * 4]);
 290     index[i] = input[i * 4 + 2];
 291   }
 292
 293   output_ptr = channels;
 294   for (i = MS_IMA_ADPCM_PREAMBLE_SIZE * channels; i < block_size; i++)
 295   {
 296     output[output_ptr++] = input[i] >> 4;
 297     output[output_ptr++] = input[i] & 0x0F;
 298   }
 299
 300   decode_nibbles(&output[channels],
 301     (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels) * 2 - channels,
 302     channels,
 303     predictor, index);
 304
 305   return (block_size - MS_IMA_ADPCM_PREAMBLE_SIZE * channels) * 2 - channels;
 306 }
 307
 308 static int decode_audio(sh_audio_t *sh_audio,unsigned char *buf,int minlen,int maxlen)
 309 {
 310   int res = -1;
 311   int (*decode_func)(unsigned short *output, unsigned char *input, int channels, int block_size) = qt_ima_adpcm_decode_block;
 312   if (demux_read_data(sh_audio->ds, sh_audio->a_in_buffer,
 313     sh_audio->ds->ss_mul) !=
 314     sh_audio->ds->ss_mul)
 315     return -1;
 316
 317   if ((sh_audio->format == 0x11) || (sh_audio->format == 0x1100736d))
 318     decode_func = ms_ima_adpcm_decode_block;
 319   else if (sh_audio->format == 0x61)
 320     decode_func = dk4_ima_adpcm_decode_block;
 321
 322   res = decode_func((unsigned short*)buf, sh_audio->a_in_buffer,
 323                     sh_audio->wf->nChannels, sh_audio->ds->ss_mul);
 324   return res < 0 ? res : 2 * res;
 325 }