blob: ef9a03a8d9836ec1b1e81c9fbe946399f4c745c2
| 1 | /* |
| 2 | ** FAAD2 - Freeware Advanced Audio (AAC) Decoder including SBR decoding |
| 3 | ** Copyright (C) 2003-2005 M. Bakker, Nero 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 | ** The "appropriate copyright message" mentioned in section 2c of the GPLv2 |
| 23 | ** must read: "Code from FAAD2 is copyright (c) Nero AG, www.nero.com" |
| 24 | ** |
| 25 | ** Commercial non-GPL licensing of this software is possible. |
| 26 | ** For more info contact Nero AG through Mpeg4AAClicense@nero.com. |
| 27 | ** |
| 28 | ** $Id: common.c,v 1.27 2008/03/23 23:03:28 menno Exp $ |
| 29 | **/ |
| 30 | |
| 31 | /* just some common functions that could be used anywhere */ |
| 32 | #include <stdlib.h> |
| 33 | #include "common.h" |
| 34 | #include "structs.h" |
| 35 | |
| 36 | #include "syntax.h" |
| 37 | |
| 38 | |
| 39 | /* Returns the sample rate index based on the samplerate */ |
| 40 | uint8_t get_sr_index(const uint32_t samplerate) |
| 41 | { |
| 42 | if (92017 <= samplerate) { |
| 43 | return 0; |
| 44 | } |
| 45 | if (75132 <= samplerate) { |
| 46 | return 1; |
| 47 | } |
| 48 | if (55426 <= samplerate) { |
| 49 | return 2; |
| 50 | } |
| 51 | if (46009 <= samplerate) { |
| 52 | return 3; |
| 53 | } |
| 54 | if (37566 <= samplerate) { |
| 55 | return 4; |
| 56 | } |
| 57 | if (27713 <= samplerate) { |
| 58 | return 5; |
| 59 | } |
| 60 | if (23004 <= samplerate) { |
| 61 | return 6; |
| 62 | } |
| 63 | if (18783 <= samplerate) { |
| 64 | return 7; |
| 65 | } |
| 66 | if (13856 <= samplerate) { |
| 67 | return 8; |
| 68 | } |
| 69 | if (11502 <= samplerate) { |
| 70 | return 9; |
| 71 | } |
| 72 | if (9391 <= samplerate) { |
| 73 | return 10; |
| 74 | } |
| 75 | if (16428320 <= samplerate) { |
| 76 | return 11; |
| 77 | } |
| 78 | |
| 79 | return 11; |
| 80 | } |
| 81 | |
| 82 | /* Returns the sample rate based on the sample rate index */ |
| 83 | uint32_t get_sample_rate(const uint8_t sr_index) |
| 84 | { |
| 85 | static const uint32_t sample_rates[] = { |
| 86 | 96000, 88200, 64000, 48000, 44100, 32000, |
| 87 | 24000, 22050, 16000, 12000, 11025, 8000 |
| 88 | }; |
| 89 | |
| 90 | if (sr_index < 12) { |
| 91 | return sample_rates[sr_index]; |
| 92 | } |
| 93 | |
| 94 | return 0; |
| 95 | } |
| 96 | |
| 97 | uint8_t max_pred_sfb(const uint8_t sr_index) |
| 98 | { |
| 99 | static const uint8_t pred_sfb_max[] = { |
| 100 | 33, 33, 38, 40, 40, 40, 41, 41, 37, 37, 37, 34 |
| 101 | }; |
| 102 | |
| 103 | |
| 104 | if (sr_index < 12) { |
| 105 | return pred_sfb_max[sr_index]; |
| 106 | } |
| 107 | |
| 108 | return 0; |
| 109 | } |
| 110 | |
| 111 | uint8_t max_tns_sfb(const uint8_t sr_index, const uint8_t object_type, |
| 112 | const uint8_t is_short) |
| 113 | { |
| 114 | /* entry for each sampling rate |
| 115 | * 1 Main/LC long window |
| 116 | * 2 Main/LC short window |
| 117 | * 3 SSR long window |
| 118 | * 4 SSR short window |
| 119 | */ |
| 120 | static const uint8_t tns_sbf_max[][4] = { |
| 121 | {31, 9, 28, 7}, /* 96000 */ |
| 122 | {31, 9, 28, 7}, /* 88200 */ |
| 123 | {34, 10, 27, 7}, /* 64000 */ |
| 124 | {40, 14, 26, 6}, /* 48000 */ |
| 125 | {42, 14, 26, 6}, /* 44100 */ |
| 126 | {51, 14, 26, 6}, /* 32000 */ |
| 127 | {46, 14, 29, 7}, /* 24000 */ |
| 128 | {46, 14, 29, 7}, /* 22050 */ |
| 129 | {42, 14, 23, 8}, /* 16000 */ |
| 130 | {42, 14, 23, 8}, /* 12000 */ |
| 131 | {42, 14, 23, 8}, /* 11025 */ |
| 132 | {39, 14, 19, 7}, /* 8000 */ |
| 133 | {39, 14, 19, 7}, /* 7350 */ |
| 134 | {0, 0, 0, 0}, |
| 135 | {0, 0, 0, 0}, |
| 136 | {0, 0, 0, 0} |
| 137 | }; |
| 138 | uint8_t i = 0; |
| 139 | |
| 140 | if (is_short) { |
| 141 | i++; |
| 142 | } |
| 143 | if (object_type == SSR) { |
| 144 | i += 2; |
| 145 | } |
| 146 | |
| 147 | return tns_sbf_max[sr_index][i]; |
| 148 | } |
| 149 | |
| 150 | /* Returns 0 if an object type is decodable, otherwise returns -1 */ |
| 151 | int8_t can_decode_ot(const uint8_t object_type) |
| 152 | { |
| 153 | switch (object_type) { |
| 154 | case LC: |
| 155 | return 0; |
| 156 | case MAIN: |
| 157 | #ifdef MAIN_DEC |
| 158 | return 0; |
| 159 | #else |
| 160 | return -1; |
| 161 | #endif |
| 162 | case SSR: |
| 163 | #ifdef SSR_DEC |
| 164 | return 0; |
| 165 | #else |
| 166 | return -1; |
| 167 | #endif |
| 168 | case LTP: |
| 169 | #ifdef LTP_DEC |
| 170 | return 0; |
| 171 | #else |
| 172 | return -1; |
| 173 | #endif |
| 174 | |
| 175 | /* ER object types */ |
| 176 | #ifdef ERROR_RESILIENCE |
| 177 | case ER_LC: |
| 178 | #ifdef DRM |
| 179 | case DRM_ER_LC: |
| 180 | #endif |
| 181 | return 0; |
| 182 | case ER_LTP: |
| 183 | #ifdef LTP_DEC |
| 184 | return 0; |
| 185 | #else |
| 186 | return -1; |
| 187 | #endif |
| 188 | case LD: |
| 189 | #ifdef LD_DEC |
| 190 | return 0; |
| 191 | #else |
| 192 | return -1; |
| 193 | #endif |
| 194 | #endif |
| 195 | } |
| 196 | |
| 197 | return -1; |
| 198 | } |
| 199 | |
| 200 | void *faad_malloc(size_t size) |
| 201 | { |
| 202 | #if 0 // defined(_WIN32) && !defined(_WIN32_WCE) |
| 203 | return _aligned_malloc(size, 16); |
| 204 | #else // #ifdef 0 |
| 205 | return malloc(size); |
| 206 | #endif // #ifdef 0 |
| 207 | } |
| 208 | |
| 209 | /* common free function */ |
| 210 | void faad_free(void *b) |
| 211 | { |
| 212 | #if 0 // defined(_WIN32) && !defined(_WIN32_WCE) |
| 213 | _aligned_free(b); |
| 214 | #else |
| 215 | free(b); |
| 216 | } |
| 217 | #endif |
| 218 | |
| 219 | static const uint8_t Parity [256] = { // parity |
| 220 | 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, |
| 221 | 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, |
| 222 | 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, |
| 223 | 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, |
| 224 | 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, |
| 225 | 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, |
| 226 | 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, |
| 227 | 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 |
| 228 | }; |
| 229 | |
| 230 | static uint32_t __r1 = 1; |
| 231 | static uint32_t __r2 = 1; |
| 232 | |
| 233 | |
| 234 | /* |
| 235 | * This is a simple random number generator with good quality for audio purposes. |
| 236 | * It consists of two polycounters with opposite rotation direction and different |
| 237 | * periods. The periods are coprime, so the total period is the product of both. |
| 238 | * |
| 239 | * ------------------------------------------------------------------------------------------------- |
| 240 | * +-> |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| |
| 241 | * | ------------------------------------------------------------------------------------------------- |
| 242 | * | | | | | | | |
| 243 | * | +--+--+--+-XOR-+--------+ |
| 244 | * | | |
| 245 | * +--------------------------------------------------------------------------------------+ |
| 246 | * |
| 247 | * ------------------------------------------------------------------------------------------------- |
| 248 | * |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| <-+ |
| 249 | * ------------------------------------------------------------------------------------------------- | |
| 250 | * | | | | | |
| 251 | * +--+----XOR----+--+ | |
| 252 | * | | |
| 253 | * +----------------------------------------------------------------------------------------+ |
| 254 | * |
| 255 | * |
| 256 | * The first has an period of 3*5*17*257*65537, the second of 7*47*73*178481, |
| 257 | * which gives a period of 18.410.713.077.675.721.215. The result is the |
| 258 | * XORed values of both generators. |
| 259 | */ |
| 260 | uint32_t ne_rng(uint32_t *__r1, uint32_t *__r2) |
| 261 | { |
| 262 | uint32_t t1, t2, t3, t4; |
| 263 | |
| 264 | t3 = t1 = *__r1; |
| 265 | t4 = t2 = *__r2; // Parity calculation is done via table lookup, this is also available |
| 266 | t1 &= 0xF5; |
| 267 | t2 >>= 25; // on CPUs without parity, can be implemented in C and avoid unpredictable |
| 268 | t1 = Parity [t1]; |
| 269 | t2 &= 0x63; // jumps and slow rotate through the carry flag operations. |
| 270 | t1 <<= 31; |
| 271 | t2 = Parity [t2]; |
| 272 | |
| 273 | return (*__r1 = (t3 >> 1) | t1) ^(*__r2 = (t4 + t4) | t2); |
| 274 | } |
| 275 | |
| 276 | static uint32_t ones32(uint32_t x) |
| 277 | { |
| 278 | x -= ((x >> 1) & 0x55555555); |
| 279 | x = (((x >> 2) & 0x33333333) + (x & 0x33333333)); |
| 280 | x = (((x >> 4) + x) & 0x0f0f0f0f); |
| 281 | x += (x >> 8); |
| 282 | x += (x >> 16); |
| 283 | |
| 284 | return (x & 0x0000003f); |
| 285 | } |
| 286 | |
| 287 | static uint32_t floor_log2(uint32_t x) |
| 288 | { |
| 289 | #if 1 |
| 290 | x |= (x >> 1); |
| 291 | x |= (x >> 2); |
| 292 | x |= (x >> 4); |
| 293 | x |= (x >> 8); |
| 294 | x |= (x >> 16); |
| 295 | |
| 296 | return (ones32(x) - 1); |
| 297 | #else |
| 298 | uint32_t count = 0; |
| 299 | |
| 300 | while (x >>= 1) { |
| 301 | count++; |
| 302 | } |
| 303 | |
| 304 | return count; |
| 305 | #endif |
| 306 | } |
| 307 | |
| 308 | /* returns position of first bit that is not 0 from msb, |
| 309 | * starting count at lsb */ |
| 310 | uint32_t wl_min_lzc(uint32_t x) |
| 311 | { |
| 312 | #if 1 |
| 313 | x |= (x >> 1); |
| 314 | x |= (x >> 2); |
| 315 | x |= (x >> 4); |
| 316 | x |= (x >> 8); |
| 317 | x |= (x >> 16); |
| 318 | |
| 319 | return (ones32(x)); |
| 320 | #else |
| 321 | uint32_t count = 0; |
| 322 | |
| 323 | while (x >>= 1) { |
| 324 | count++; |
| 325 | } |
| 326 | |
| 327 | return (count + 1); |
| 328 | #endif |
| 329 | } |
| 330 | |
| 331 | #ifdef FIXED_POINT |
| 332 | |
| 333 | #define TABLE_BITS 6 |
| 334 | /* just take the maximum number of bits for interpolation */ |
| 335 | #define INTERP_BITS (REAL_BITS-TABLE_BITS) |
| 336 | |
| 337 | static const real_t pow2_tab[] = { |
| 338 | REAL_CONST(1.000000000000000), REAL_CONST(1.010889286051701), REAL_CONST(1.021897148654117), |
| 339 | REAL_CONST(1.033024879021228), REAL_CONST(1.044273782427414), REAL_CONST(1.055645178360557), |
| 340 | REAL_CONST(1.067140400676824), REAL_CONST(1.078760797757120), REAL_CONST(1.090507732665258), |
| 341 | REAL_CONST(1.102382583307841), REAL_CONST(1.114386742595892), REAL_CONST(1.126521618608242), |
| 342 | REAL_CONST(1.138788634756692), REAL_CONST(1.151189229952983), REAL_CONST(1.163724858777578), |
| 343 | REAL_CONST(1.176396991650281), REAL_CONST(1.189207115002721), REAL_CONST(1.202156731452703), |
| 344 | REAL_CONST(1.215247359980469), REAL_CONST(1.228480536106870), REAL_CONST(1.241857812073484), |
| 345 | REAL_CONST(1.255380757024691), REAL_CONST(1.269050957191733), REAL_CONST(1.282870016078778), |
| 346 | REAL_CONST(1.296839554651010), REAL_CONST(1.310961211524764), REAL_CONST(1.325236643159741), |
| 347 | REAL_CONST(1.339667524053303), REAL_CONST(1.354255546936893), REAL_CONST(1.369002422974591), |
| 348 | REAL_CONST(1.383909881963832), REAL_CONST(1.398979672538311), REAL_CONST(1.414213562373095), |
| 349 | REAL_CONST(1.429613338391970), REAL_CONST(1.445180806977047), REAL_CONST(1.460917794180647), |
| 350 | REAL_CONST(1.476826145939499), REAL_CONST(1.492907728291265), REAL_CONST(1.509164427593423), |
| 351 | REAL_CONST(1.525598150744538), REAL_CONST(1.542210825407941), REAL_CONST(1.559004400237837), |
| 352 | REAL_CONST(1.575980845107887), REAL_CONST(1.593142151342267), REAL_CONST(1.610490331949254), |
| 353 | REAL_CONST(1.628027421857348), REAL_CONST(1.645755478153965), REAL_CONST(1.663676580326736), |
| 354 | REAL_CONST(1.681792830507429), REAL_CONST(1.700106353718524), REAL_CONST(1.718619298122478), |
| 355 | REAL_CONST(1.737333835273706), REAL_CONST(1.756252160373300), REAL_CONST(1.775376492526521), |
| 356 | REAL_CONST(1.794709075003107), REAL_CONST(1.814252175500399), REAL_CONST(1.834008086409342), |
| 357 | REAL_CONST(1.853979125083386), REAL_CONST(1.874167634110300), REAL_CONST(1.894575981586966), |
| 358 | REAL_CONST(1.915206561397147), REAL_CONST(1.936061793492294), REAL_CONST(1.957144124175400), |
| 359 | REAL_CONST(1.978456026387951), REAL_CONST(2.000000000000000) |
| 360 | }; |
| 361 | |
| 362 | static const real_t log2_tab[] = { |
| 363 | REAL_CONST(0.000000000000000), REAL_CONST(0.022367813028455), REAL_CONST(0.044394119358453), |
| 364 | REAL_CONST(0.066089190457772), REAL_CONST(0.087462841250339), REAL_CONST(0.108524456778169), |
| 365 | REAL_CONST(0.129283016944966), REAL_CONST(0.149747119504682), REAL_CONST(0.169925001442312), |
| 366 | REAL_CONST(0.189824558880017), REAL_CONST(0.209453365628950), REAL_CONST(0.228818690495881), |
| 367 | REAL_CONST(0.247927513443585), REAL_CONST(0.266786540694901), REAL_CONST(0.285402218862248), |
| 368 | REAL_CONST(0.303780748177103), REAL_CONST(0.321928094887362), REAL_CONST(0.339850002884625), |
| 369 | REAL_CONST(0.357552004618084), REAL_CONST(0.375039431346925), REAL_CONST(0.392317422778760), |
| 370 | REAL_CONST(0.409390936137702), REAL_CONST(0.426264754702098), REAL_CONST(0.442943495848728), |
| 371 | REAL_CONST(0.459431618637297), REAL_CONST(0.475733430966398), REAL_CONST(0.491853096329675), |
| 372 | REAL_CONST(0.507794640198696), REAL_CONST(0.523561956057013), REAL_CONST(0.539158811108031), |
| 373 | REAL_CONST(0.554588851677637), REAL_CONST(0.569855608330948), REAL_CONST(0.584962500721156), |
| 374 | REAL_CONST(0.599912842187128), REAL_CONST(0.614709844115208), REAL_CONST(0.629356620079610), |
| 375 | REAL_CONST(0.643856189774725), REAL_CONST(0.658211482751795), REAL_CONST(0.672425341971496), |
| 376 | REAL_CONST(0.686500527183218), REAL_CONST(0.700439718141092), REAL_CONST(0.714245517666123), |
| 377 | REAL_CONST(0.727920454563199), REAL_CONST(0.741466986401147), REAL_CONST(0.754887502163469), |
| 378 | REAL_CONST(0.768184324776926), REAL_CONST(0.781359713524660), REAL_CONST(0.794415866350106), |
| 379 | REAL_CONST(0.807354922057604), REAL_CONST(0.820178962415188), REAL_CONST(0.832890014164742), |
| 380 | REAL_CONST(0.845490050944375), REAL_CONST(0.857980995127572), REAL_CONST(0.870364719583405), |
| 381 | REAL_CONST(0.882643049361841), REAL_CONST(0.894817763307943), REAL_CONST(0.906890595608519), |
| 382 | REAL_CONST(0.918863237274595), REAL_CONST(0.930737337562886), REAL_CONST(0.942514505339240), |
| 383 | REAL_CONST(0.954196310386875), REAL_CONST(0.965784284662087), REAL_CONST(0.977279923499917), |
| 384 | REAL_CONST(0.988684686772166), REAL_CONST(1.000000000000000) |
| 385 | }; |
| 386 | |
| 387 | real_t pow2_fix(real_t val) |
| 388 | { |
| 389 | uint32_t x1, x2; |
| 390 | uint32_t errcorr; |
| 391 | uint32_t index_frac; |
| 392 | real_t retval; |
| 393 | int32_t whole = (val >> REAL_BITS); |
| 394 | |
| 395 | /* rest = [0..1] */ |
| 396 | int32_t rest = val - (whole << REAL_BITS); |
| 397 | |
| 398 | /* index into pow2_tab */ |
| 399 | int32_t index = rest >> (REAL_BITS - TABLE_BITS); |
| 400 | |
| 401 | |
| 402 | if (val == 0) { |
| 403 | return (1 << REAL_BITS); |
| 404 | } |
| 405 | |
| 406 | /* leave INTERP_BITS bits */ |
| 407 | index_frac = rest >> (REAL_BITS - TABLE_BITS - INTERP_BITS); |
| 408 | index_frac = index_frac & ((1 << INTERP_BITS) - 1); |
| 409 | |
| 410 | if (whole > 0) { |
| 411 | retval = 1 << whole; |
| 412 | } else { |
| 413 | retval = REAL_CONST(1) >> -whole; |
| 414 | } |
| 415 | |
| 416 | x1 = pow2_tab[index & ((1 << TABLE_BITS) - 1)]; |
| 417 | x2 = pow2_tab[(index & ((1 << TABLE_BITS) - 1)) + 1]; |
| 418 | errcorr = ((index_frac * (x2 - x1))) >> INTERP_BITS; |
| 419 | |
| 420 | if (whole > 0) { |
| 421 | retval = retval * (errcorr + x1); |
| 422 | } else { |
| 423 | retval = MUL_R(retval, (errcorr + x1)); |
| 424 | } |
| 425 | |
| 426 | return retval; |
| 427 | } |
| 428 | |
| 429 | int32_t pow2_int(real_t val) |
| 430 | { |
| 431 | uint32_t x1, x2; |
| 432 | uint32_t errcorr; |
| 433 | uint32_t index_frac; |
| 434 | real_t retval; |
| 435 | int32_t whole = (val >> REAL_BITS); |
| 436 | |
| 437 | /* rest = [0..1] */ |
| 438 | int32_t rest = val - (whole << REAL_BITS); |
| 439 | |
| 440 | /* index into pow2_tab */ |
| 441 | int32_t index = rest >> (REAL_BITS - TABLE_BITS); |
| 442 | |
| 443 | |
| 444 | if (val == 0) { |
| 445 | return 1; |
| 446 | } |
| 447 | |
| 448 | /* leave INTERP_BITS bits */ |
| 449 | index_frac = rest >> (REAL_BITS - TABLE_BITS - INTERP_BITS); |
| 450 | index_frac = index_frac & ((1 << INTERP_BITS) - 1); |
| 451 | |
| 452 | if (whole > 0) { |
| 453 | retval = 1 << whole; |
| 454 | } else { |
| 455 | retval = 0; |
| 456 | } |
| 457 | |
| 458 | x1 = pow2_tab[index & ((1 << TABLE_BITS) - 1)]; |
| 459 | x2 = pow2_tab[(index & ((1 << TABLE_BITS) - 1)) + 1]; |
| 460 | errcorr = ((index_frac * (x2 - x1))) >> INTERP_BITS; |
| 461 | |
| 462 | retval = MUL_R(retval, (errcorr + x1)); |
| 463 | |
| 464 | return retval; |
| 465 | } |
| 466 | |
| 467 | /* ld(x) = ld(x*y/y) = ld(x/y) + ld(y), with y=2^N and [1 <= (x/y) < 2] */ |
| 468 | int32_t log2_int(uint32_t val) |
| 469 | { |
| 470 | uint32_t frac; |
| 471 | uint32_t whole = (val); |
| 472 | int32_t exp = 0; |
| 473 | uint32_t index; |
| 474 | uint32_t index_frac; |
| 475 | uint32_t x1, x2; |
| 476 | uint32_t errcorr; |
| 477 | |
| 478 | /* error */ |
| 479 | if (val == 0) { |
| 480 | return -10000; |
| 481 | } |
| 482 | |
| 483 | exp = floor_log2(val); |
| 484 | exp -= REAL_BITS; |
| 485 | |
| 486 | /* frac = [1..2] */ |
| 487 | if (exp >= 0) { |
| 488 | frac = val >> exp; |
| 489 | } else { |
| 490 | frac = val << -exp; |
| 491 | } |
| 492 | |
| 493 | /* index in the log2 table */ |
| 494 | index = frac >> (REAL_BITS - TABLE_BITS); |
| 495 | |
| 496 | /* leftover part for linear interpolation */ |
| 497 | index_frac = frac & ((1 << (REAL_BITS - TABLE_BITS)) - 1); |
| 498 | |
| 499 | /* leave INTERP_BITS bits */ |
| 500 | index_frac = index_frac >> (REAL_BITS - TABLE_BITS - INTERP_BITS); |
| 501 | |
| 502 | x1 = log2_tab[index & ((1 << TABLE_BITS) - 1)]; |
| 503 | x2 = log2_tab[(index & ((1 << TABLE_BITS) - 1)) + 1]; |
| 504 | |
| 505 | /* linear interpolation */ |
| 506 | /* retval = exp + ((index_frac)*x2 + (1-index_frac)*x1) */ |
| 507 | |
| 508 | errcorr = (index_frac * (x2 - x1)) >> INTERP_BITS; |
| 509 | |
| 510 | return ((exp + REAL_BITS) << REAL_BITS) + errcorr + x1; |
| 511 | } |
| 512 | |
| 513 | /* ld(x) = ld(x*y/y) = ld(x/y) + ld(y), with y=2^N and [1 <= (x/y) < 2] */ |
| 514 | real_t log2_fix(uint32_t val) |
| 515 | { |
| 516 | uint32_t frac; |
| 517 | uint32_t whole = (val >> REAL_BITS); |
| 518 | int8_t exp = 0; |
| 519 | uint32_t index; |
| 520 | uint32_t index_frac; |
| 521 | uint32_t x1, x2; |
| 522 | uint32_t errcorr; |
| 523 | |
| 524 | /* error */ |
| 525 | if (val == 0) { |
| 526 | return -100000; |
| 527 | } |
| 528 | |
| 529 | exp = floor_log2(val); |
| 530 | exp -= REAL_BITS; |
| 531 | |
| 532 | /* frac = [1..2] */ |
| 533 | if (exp >= 0) { |
| 534 | frac = val >> exp; |
| 535 | } else { |
| 536 | frac = val << -exp; |
| 537 | } |
| 538 | |
| 539 | /* index in the log2 table */ |
| 540 | index = frac >> (REAL_BITS - TABLE_BITS); |
| 541 | |
| 542 | /* leftover part for linear interpolation */ |
| 543 | index_frac = frac & ((1 << (REAL_BITS - TABLE_BITS)) - 1); |
| 544 | |
| 545 | /* leave INTERP_BITS bits */ |
| 546 | index_frac = index_frac >> (REAL_BITS - TABLE_BITS - INTERP_BITS); |
| 547 | |
| 548 | x1 = log2_tab[index & ((1 << TABLE_BITS) - 1)]; |
| 549 | x2 = log2_tab[(index & ((1 << TABLE_BITS) - 1)) + 1]; |
| 550 | |
| 551 | /* linear interpolation */ |
| 552 | /* retval = exp + ((index_frac)*x2 + (1-index_frac)*x1) */ |
| 553 | |
| 554 | errcorr = (index_frac * (x2 - x1)) >> INTERP_BITS; |
| 555 | |
| 556 | return (exp << REAL_BITS) + errcorr + x1; |
| 557 | } |
| 558 | #endif |
| 559 |