libfaad2/common.c

   1 /*
   2 ** FAAD2 - Freeware Advanced Audio (AAC) Decoder including SBR decoding
   3 ** Copyright (C) 2003-2004 M. Bakker, Ahead Software AG, http://www.nero.com
   4 **
   5 ** This program is free software; you can redistribute it and/or modify
   6 ** it under the terms of the GNU General Public License as published by
   7 ** the Free Software Foundation; either version 2 of the License, or
   8 ** (at your option) any later version.
   9 **
  10 ** This program is distributed in the hope that it will be useful,
  11 ** but WITHOUT ANY WARRANTY; without even the implied warranty of
  12 ** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13 ** GNU General Public License for more details.
  14 **
  15 ** You should have received a copy of the GNU General Public License
  16 ** along with this program; if not, write to the Free Software
  17 ** Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
  18 **
  19 ** Any non-GPL usage of this software or parts of this software is strictly
  20 ** forbidden.
  21 **
  22 ** Commercial non-GPL licensing of this software is possible.
  23 ** For more info contact Ahead Software through Mpeg4AAClicense@nero.com.
  24 **
  25 ** $Id: common.c,v 1.19 2004/06/30 12:45:56 menno Exp $
  26 **/
  27
  28 /* just some common functions that could be used anywhere */
  29
  30 #include "common.h"
  31 #include "structs.h"
  32
  33 #include <stdlib.h>
  34 #include "syntax.h"
  35
  36 #ifdef USE_SSE
  37 __declspec(naked) static int32_t __fastcall test_cpuid(void)
  38 {
  39     __asm
  40     {
  41         pushf
  42         pop eax
  43         mov ecx,eax
  44         xor eax,(1<<21)
  45         push eax
  46         popf
  47         pushf
  48         pop eax
  49         push ecx
  50         popf
  51         cmp eax,ecx
  52         mov eax,0
  53         setne al
  54         ret
  55     }
  56 }
  57
  58 __declspec(naked) static void __fastcall run_cpuid(int32_t param, int32_t out[4])
  59 {
  60     __asm
  61     {
  62         pushad
  63         push edx
  64         mov eax,ecx
  65         cpuid
  66         pop edi
  67         mov [edi+0],eax
  68         mov [edi+4],ebx
  69         mov [edi+8],ecx
  70         mov [edi+12],edx
  71         popad
  72         ret
  73     }
  74 }
  75
  76 uint8_t cpu_has_sse()
  77 {
  78     int32_t features[4];
  79
  80     if (test_cpuid())
  81     {
  82         run_cpuid(1, features);
  83     }
  84
  85     /* check for SSE */
  86     if (features[3] & 0x02000000)
  87         return 1;
  88
  89     return 0;
  90 }
  91 #else
  92 uint8_t cpu_has_sse()
  93 {
  94     return 0;
  95 }
  96 #endif
  97
  98 /* Returns the sample rate index based on the samplerate */
  99 uint8_t get_sr_index(const uint32_t samplerate)
 100 {
 101     if (92017 <= samplerate) return 0;
 102     if (75132 <= samplerate) return 1;
 103     if (55426 <= samplerate) return 2;
 104     if (46009 <= samplerate) return 3;
 105     if (37566 <= samplerate) return 4;
 106     if (27713 <= samplerate) return 5;
 107     if (23004 <= samplerate) return 6;
 108     if (18783 <= samplerate) return 7;
 109     if (13856 <= samplerate) return 8;
 110     if (11502 <= samplerate) return 9;
 111     if (9391 <= samplerate) return 10;
 112     if (16428320 <= samplerate) return 11;
 113
 114     return 11;
 115 }
 116
 117 /* Returns the sample rate based on the sample rate index */
 118 uint32_t get_sample_rate(const uint8_t sr_index)
 119 {
 120     static const uint32_t sample_rates[] =
 121     {
 122         96000, 88200, 64000, 48000, 44100, 32000,
 123         24000, 22050, 16000, 12000, 11025, 8000
 124     };
 125
 126     if (sr_index < 12)
 127         return sample_rates[sr_index];
 128
 129     return 0;
 130 }
 131
 132 uint8_t max_pred_sfb(const uint8_t sr_index)
 133 {
 134     static const uint8_t pred_sfb_max[] =
 135     {
 136         33, 33, 38, 40, 40, 40, 41, 41, 37, 37, 37, 34
 137     };
 138
 139
 140     if (sr_index < 12)
 141         return pred_sfb_max[sr_index];
 142
 143     return 0;
 144 }
 145
 146 uint8_t max_tns_sfb(const uint8_t sr_index, const uint8_t object_type,
 147                     const uint8_t is_short)
 148 {
 149     /* entry for each sampling rate
 150      * 1    Main/LC long window
 151      * 2    Main/LC short window
 152      * 3    SSR long window
 153      * 4    SSR short window
 154      */
 155     static const uint8_t tns_sbf_max[][4] =
 156     {
 157         {31,  9, 28, 7}, /* 96000 */
 158         {31,  9, 28, 7}, /* 88200 */
 159         {34, 10, 27, 7}, /* 64000 */
 160         {40, 14, 26, 6}, /* 48000 */
 161         {42, 14, 26, 6}, /* 44100 */
 162         {51, 14, 26, 6}, /* 32000 */
 163         {46, 14, 29, 7}, /* 24000 */
 164         {46, 14, 29, 7}, /* 22050 */
 165         {42, 14, 23, 8}, /* 16000 */
 166         {42, 14, 23, 8}, /* 12000 */
 167         {42, 14, 23, 8}, /* 11025 */
 168         {39, 14, 19, 7}, /*  8000 */
 169         {39, 14, 19, 7}, /*  7350 */
 170         {0,0,0,0},
 171         {0,0,0,0},
 172         {0,0,0,0}
 173     };
 174     uint8_t i = 0;
 175
 176     if (is_short) i++;
 177     if (object_type == SSR) i += 2;
 178
 179     return tns_sbf_max[sr_index][i];
 180 }
 181
 182 /* Returns 0 if an object type is decodable, otherwise returns -1 */
 183 int8_t can_decode_ot(const uint8_t object_type)
 184 {
 185     switch (object_type)
 186     {
 187     case LC:
 188         return 0;
 189     case MAIN:
 190 #ifdef MAIN_DEC
 191         return 0;
 192 #else
 193         return -1;
 194 #endif
 195     case SSR:
 196 #ifdef SSR_DEC
 197         return 0;
 198 #else
 199         return -1;
 200 #endif
 201     case LTP:
 202 #ifdef LTP_DEC
 203         return 0;
 204 #else
 205         return -1;
 206 #endif
 207
 208     /* ER object types */
 209 #ifdef ERROR_RESILIENCE
 210     case ER_LC:
 211 #ifdef DRM
 212     case DRM_ER_LC:
 213 #endif
 214         return 0;
 215     case ER_LTP:
 216 #ifdef LTP_DEC
 217         return 0;
 218 #else
 219         return -1;
 220 #endif
 221     case LD:
 222 #ifdef LD_DEC
 223         return 0;
 224 #else
 225         return -1;
 226 #endif
 227 #endif
 228     }
 229
 230     return -1;
 231 }
 232
 233 /* common malloc function */
 234 void *faad_malloc(size_t size)
 235 {
 236 #if 0 // defined(_WIN32) && !defined(_WIN32_WCE)
 237     return _aligned_malloc(size, 16);
 238 #else
 239     return malloc(size);
 240 #endif
 241 }
 242
 243 /* common free function */
 244 void faad_free(void *b)
 245 {
 246 #if 0 // defined(_WIN32) && !defined(_WIN32_WCE)
 247     _aligned_free(b);
 248 #else
 249     free(b);
 250 #endif
 251 }
 252
 253 static const  uint8_t    Parity [256] = {  // parity
 254         0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,
 255         1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,
 256         1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,
 257         0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,
 258         1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,
 259         0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,
 260         0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,
 261         1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0
 262 };
 263
 264 static uint32_t  __r1 = 1;
 265 static uint32_t  __r2 = 1;
 266
 267
 268 /*
 269  *  This is a simple random number generator with good quality for audio purposes.
 270  *  It consists of two polycounters with opposite rotation direction and different
 271  *  periods. The periods are coprime, so the total period is the product of both.
 272  *
 273  *     -------------------------------------------------------------------------------------------------
 274  * +-> |31:30:29:28:27:26:25:24:23:22:21:20:19:18:17:16:15:14:13:12:11:10: 9: 8: 7: 6: 5: 4: 3: 2: 1: 0|
 275  * |   -------------------------------------------------------------------------------------------------
 276  * |                                                                          |  |  |  |     |        |
 277  * |                                                                          +--+--+--+-XOR-+--------+
 278  * |                                                                                      |
 279  * +--------------------------------------------------------------------------------------+
 280  *
 281  *     -------------------------------------------------------------------------------------------------
 282  *     |31:30:29:28:27:26:25:24:23:22:21:20:19:18:17:16:15:14:13:12:11:10: 9: 8: 7: 6: 5: 4: 3: 2: 1: 0| <-+
 283  *     -------------------------------------------------------------------------------------------------   |
 284  *       |  |           |  |                                                                               |
 285  *       +--+----XOR----+--+                                                                               |
 286  *                |                                                                                        |
 287  *                +----------------------------------------------------------------------------------------+
 288  *
 289  *
 290  *  The first has an period of 3*5*17*257*65537, the second of 7*47*73*178481,
 291  *  which gives a period of 18.410.713.077.675.721.215. The result is the
 292  *  XORed values of both generators.
 293  */
 294 uint32_t random_int(void)
 295 {
 296         uint32_t  t1, t2, t3, t4;
 297
 298         t3   = t1 = __r1;   t4   = t2 = __r2;       // Parity calculation is done via table lookup, this is also available
 299         t1  &= 0xF5;        t2 >>= 25;              // on CPUs without parity, can be implemented in C and avoid unpredictable
 300         t1   = Parity [t1]; t2  &= 0x63;            // jumps and slow rotate through the carry flag operations.
 301         t1 <<= 31;          t2   = Parity [t2];
 302
 303         return (__r1 = (t3 >> 1) | t1 ) ^ (__r2 = (t4 + t4) | t2 );
 304 }
 305
 306 uint32_t ones32(uint32_t x)
 307 {
 308     x -= ((x >> 1) & 0x55555555);
 309     x = (((x >> 2) & 0x33333333) + (x & 0x33333333));
 310     x = (((x >> 4) + x) & 0x0f0f0f0f);
 311     x += (x >> 8);
 312     x += (x >> 16);
 313
 314     return (x & 0x0000003f);
 315 }
 316
 317 uint32_t floor_log2(uint32_t x)
 318 {
 319 #if 1
 320     x |= (x >> 1);
 321     x |= (x >> 2);
 322     x |= (x >> 4);
 323     x |= (x >> 8);
 324     x |= (x >> 16);
 325
 326     return (ones32(x) - 1);
 327 #else
 328     uint32_t count = 0;
 329
 330     while (x >>= 1)
 331         count++;
 332
 333     return count;
 334 #endif
 335 }
 336
 337 /* returns position of first bit that is not 0 from msb,
 338  * starting count at lsb */
 339 uint32_t wl_min_lzc(uint32_t x)
 340 {
 341 #if 1
 342     x |= (x >> 1);
 343     x |= (x >> 2);
 344     x |= (x >> 4);
 345     x |= (x >> 8);
 346     x |= (x >> 16);
 347
 348     return (ones32(x));
 349 #else
 350     uint32_t count = 0;
 351
 352     while (x >>= 1)
 353         count++;
 354
 355     return (count + 1);
 356 #endif
 357 }
 358
 359 #ifdef FIXED_POINT
 360
 361 #define TABLE_BITS 6
 362 /* just take the maximum number of bits for interpolation */
 363 #define INTERP_BITS (REAL_BITS-TABLE_BITS)
 364
 365 static const real_t pow2_tab[] = {
 366     REAL_CONST(1.000000000000000), REAL_CONST(1.010889286051701), REAL_CONST(1.021897148654117),
 367     REAL_CONST(1.033024879021228), REAL_CONST(1.044273782427414), REAL_CONST(1.055645178360557),
 368     REAL_CONST(1.067140400676824), REAL_CONST(1.078760797757120), REAL_CONST(1.090507732665258),
 369     REAL_CONST(1.102382583307841), REAL_CONST(1.114386742595892), REAL_CONST(1.126521618608242),
 370     REAL_CONST(1.138788634756692), REAL_CONST(1.151189229952983), REAL_CONST(1.163724858777578),
 371     REAL_CONST(1.176396991650281), REAL_CONST(1.189207115002721), REAL_CONST(1.202156731452703),
 372     REAL_CONST(1.215247359980469), REAL_CONST(1.228480536106870), REAL_CONST(1.241857812073484),
 373     REAL_CONST(1.255380757024691), REAL_CONST(1.269050957191733), REAL_CONST(1.282870016078778),
 374     REAL_CONST(1.296839554651010), REAL_CONST(1.310961211524764), REAL_CONST(1.325236643159741),
 375     REAL_CONST(1.339667524053303), REAL_CONST(1.354255546936893), REAL_CONST(1.369002422974591),
 376     REAL_CONST(1.383909881963832), REAL_CONST(1.398979672538311), REAL_CONST(1.414213562373095),
 377     REAL_CONST(1.429613338391970), REAL_CONST(1.445180806977047), REAL_CONST(1.460917794180647),
 378     REAL_CONST(1.476826145939499), REAL_CONST(1.492907728291265), REAL_CONST(1.509164427593423),
 379     REAL_CONST(1.525598150744538), REAL_CONST(1.542210825407941), REAL_CONST(1.559004400237837),
 380     REAL_CONST(1.575980845107887), REAL_CONST(1.593142151342267), REAL_CONST(1.610490331949254),
 381     REAL_CONST(1.628027421857348), REAL_CONST(1.645755478153965), REAL_CONST(1.663676580326736),
 382     REAL_CONST(1.681792830507429), REAL_CONST(1.700106353718524), REAL_CONST(1.718619298122478),
 383     REAL_CONST(1.737333835273706), REAL_CONST(1.756252160373300), REAL_CONST(1.775376492526521),
 384     REAL_CONST(1.794709075003107), REAL_CONST(1.814252175500399), REAL_CONST(1.834008086409342),
 385     REAL_CONST(1.853979125083386), REAL_CONST(1.874167634110300), REAL_CONST(1.894575981586966),
 386     REAL_CONST(1.915206561397147), REAL_CONST(1.936061793492294), REAL_CONST(1.957144124175400),
 387     REAL_CONST(1.978456026387951), REAL_CONST(2.000000000000000)
 388 };
 389
 390 static const real_t log2_tab[] = {
 391     REAL_CONST(0.000000000000000), REAL_CONST(0.022367813028455), REAL_CONST(0.044394119358453),
 392     REAL_CONST(0.066089190457772), REAL_CONST(0.087462841250339), REAL_CONST(0.108524456778169),
 393     REAL_CONST(0.129283016944966), REAL_CONST(0.149747119504682), REAL_CONST(0.169925001442312),
 394     REAL_CONST(0.189824558880017), REAL_CONST(0.209453365628950), REAL_CONST(0.228818690495881),
 395     REAL_CONST(0.247927513443585), REAL_CONST(0.266786540694901), REAL_CONST(0.285402218862248),
 396     REAL_CONST(0.303780748177103), REAL_CONST(0.321928094887362), REAL_CONST(0.339850002884625),
 397     REAL_CONST(0.357552004618084), REAL_CONST(0.375039431346925), REAL_CONST(0.392317422778760),
 398     REAL_CONST(0.409390936137702), REAL_CONST(0.426264754702098), REAL_CONST(0.442943495848728),
 399     REAL_CONST(0.459431618637297), REAL_CONST(0.475733430966398), REAL_CONST(0.491853096329675),
 400     REAL_CONST(0.507794640198696), REAL_CONST(0.523561956057013), REAL_CONST(0.539158811108031),
 401     REAL_CONST(0.554588851677637), REAL_CONST(0.569855608330948), REAL_CONST(0.584962500721156),
 402     REAL_CONST(0.599912842187128), REAL_CONST(0.614709844115208), REAL_CONST(0.629356620079610),
 403     REAL_CONST(0.643856189774725), REAL_CONST(0.658211482751795), REAL_CONST(0.672425341971496),
 404     REAL_CONST(0.686500527183218), REAL_CONST(0.700439718141092), REAL_CONST(0.714245517666123),
 405     REAL_CONST(0.727920454563199), REAL_CONST(0.741466986401147), REAL_CONST(0.754887502163469),
 406     REAL_CONST(0.768184324776926), REAL_CONST(0.781359713524660), REAL_CONST(0.794415866350106),
 407     REAL_CONST(0.807354922057604), REAL_CONST(0.820178962415188), REAL_CONST(0.832890014164742),
 408     REAL_CONST(0.845490050944375), REAL_CONST(0.857980995127572), REAL_CONST(0.870364719583405),
 409     REAL_CONST(0.882643049361841), REAL_CONST(0.894817763307943), REAL_CONST(0.906890595608519),
 410     REAL_CONST(0.918863237274595), REAL_CONST(0.930737337562886), REAL_CONST(0.942514505339240),
 411     REAL_CONST(0.954196310386875), REAL_CONST(0.965784284662087), REAL_CONST(0.977279923499917),
 412     REAL_CONST(0.988684686772166), REAL_CONST(1.000000000000000)
 413 };
 414
 415 real_t pow2_fix(real_t val)
 416 {
 417     uint32_t x1, x2;
 418     uint32_t errcorr;
 419     uint32_t index_frac;
 420     real_t retval;
 421     int32_t whole = (val >> REAL_BITS);
 422
 423     /* rest = [0..1] */
 424     int32_t rest = val - (whole << REAL_BITS);
 425
 426     /* index into pow2_tab */
 427     int32_t index = rest >> (REAL_BITS-TABLE_BITS);
 428
 429
 430     if (val == 0)
 431         return (1<<REAL_BITS);
 432
 433     /* leave INTERP_BITS bits */
 434     index_frac = rest >> (REAL_BITS-TABLE_BITS-INTERP_BITS);
 435     index_frac = index_frac & ((1<<INTERP_BITS)-1);
 436
 437     if (whole > 0)
 438     {
 439         retval = 1 << whole;
 440     } else {
 441         retval = REAL_CONST(1) >> -whole;
 442     }
 443
 444     x1 = pow2_tab[index & ((1<<TABLE_BITS)-1)];
 445     x2 = pow2_tab[(index & ((1<<TABLE_BITS)-1)) + 1];
 446     errcorr = ( (index_frac*(x2-x1))) >> INTERP_BITS;
 447
 448     if (whole > 0)
 449     {
 450         retval = retval * (errcorr + x1);
 451     } else {
 452         retval = MUL_R(retval, (errcorr + x1));
 453     }
 454
 455     return retval;
 456 }
 457
 458 int32_t pow2_int(real_t val)
 459 {
 460     uint32_t x1, x2;
 461     uint32_t errcorr;
 462     uint32_t index_frac;
 463     real_t retval;
 464     int32_t whole = (val >> REAL_BITS);
 465
 466     /* rest = [0..1] */
 467     int32_t rest = val - (whole << REAL_BITS);
 468
 469     /* index into pow2_tab */
 470     int32_t index = rest >> (REAL_BITS-TABLE_BITS);
 471
 472
 473     if (val == 0)
 474         return 1;
 475
 476     /* leave INTERP_BITS bits */
 477     index_frac = rest >> (REAL_BITS-TABLE_BITS-INTERP_BITS);
 478     index_frac = index_frac & ((1<<INTERP_BITS)-1);
 479
 480     if (whole > 0)
 481         retval = 1 << whole;
 482     else
 483         retval = 0;
 484
 485     x1 = pow2_tab[index & ((1<<TABLE_BITS)-1)];
 486     x2 = pow2_tab[(index & ((1<<TABLE_BITS)-1)) + 1];
 487     errcorr = ( (index_frac*(x2-x1))) >> INTERP_BITS;
 488
 489     retval = MUL_R(retval, (errcorr + x1));
 490
 491     return retval;
 492 }
 493
 494 /* ld(x) = ld(x*y/y) = ld(x/y) + ld(y), with y=2^N and [1 <= (x/y) < 2] */
 495 int32_t log2_int(uint32_t val)
 496 {
 497     uint32_t frac;
 498     uint32_t whole = (val);
 499     int8_t exp = 0;
 500     uint32_t index;
 501     uint32_t index_frac;
 502     uint32_t x1, x2;
 503     uint32_t errcorr;
 504
 505     /* error */
 506     if (val == 0)
 507         return -10000;
 508
 509     exp = floor_log2(val);
 510     exp -= REAL_BITS;
 511
 512     /* frac = [1..2] */
 513     if (exp >= 0)
 514         frac = val >> exp;
 515     else
 516         frac = val << -exp;
 517
 518     /* index in the log2 table */
 519     index = frac >> (REAL_BITS-TABLE_BITS);
 520
 521     /* leftover part for linear interpolation */
 522     index_frac = frac & ((1<<(REAL_BITS-TABLE_BITS))-1);
 523
 524     /* leave INTERP_BITS bits */
 525     index_frac = index_frac >> (REAL_BITS-TABLE_BITS-INTERP_BITS);
 526
 527     x1 = log2_tab[index & ((1<<TABLE_BITS)-1)];
 528     x2 = log2_tab[(index & ((1<<TABLE_BITS)-1)) + 1];
 529
 530     /* linear interpolation */
 531     /* retval = exp + ((index_frac)*x2 + (1-index_frac)*x1) */
 532
 533     errcorr = (index_frac * (x2-x1)) >> INTERP_BITS;
 534
 535     return ((exp+REAL_BITS) << REAL_BITS) + errcorr + x1;
 536 }
 537 #endif