1 /* ****************************************************************** 2 FSE : Finite State Entropy codec 3 Public Prototypes declaration 4 Copyright (C) 2013-2016, Yann Collet. 5 6 BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) 7 8 Redistribution and use in source and binary forms, with or without 9 modification, are permitted provided that the following conditions are 10 met: 11 12 * Redistributions of source code must retain the above copyright 13 notice, this list of conditions and the following disclaimer. 14 * Redistributions in binary form must reproduce the above 15 copyright notice, this list of conditions and the following disclaimer 16 in the documentation and/or other materials provided with the 17 distribution. 18 19 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 20 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 21 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 22 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 23 OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 24 SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 25 LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 26 DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 27 THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 28 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 29 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 30 31 You can contact the author at : 32 - Source repository : https://github.com/Cyan4973/FiniteStateEntropy 33 ****************************************************************** */ 34 35 #if defined (__cplusplus) 36 extern "C" { 37 #endif 38 39 #ifndef FSE_H 40 #define FSE_H 41 42 43 /*-***************************************** 44 * Dependencies 45 ******************************************/ 46 #include <stddef.h> /* size_t, ptrdiff_t */ 47 48 49 /*-***************************************** 50 * FSE_PUBLIC_API : control library symbols visibility 51 ******************************************/ 52 #if defined(FSE_DLL_EXPORT) && (FSE_DLL_EXPORT==1) && defined(__GNUC__) && (__GNUC__ >= 4) 53 # define FSE_PUBLIC_API __attribute__ ((visibility ("default"))) 54 #elif defined(FSE_DLL_EXPORT) && (FSE_DLL_EXPORT==1) /* Visual expected */ 55 # define FSE_PUBLIC_API __declspec(dllexport) 56 #elif defined(FSE_DLL_IMPORT) && (FSE_DLL_IMPORT==1) 57 # define FSE_PUBLIC_API __declspec(dllimport) /* It isn't required but allows to generate better code, saving a function pointer load from the IAT and an indirect jump.*/ 58 #else 59 # define FSE_PUBLIC_API 60 #endif 61 62 /*------ Version ------*/ 63 #define FSE_VERSION_MAJOR 0 64 #define FSE_VERSION_MINOR 9 65 #define FSE_VERSION_RELEASE 0 66 67 #define FSE_LIB_VERSION FSE_VERSION_MAJOR.FSE_VERSION_MINOR.FSE_VERSION_RELEASE 68 #define FSE_QUOTE(str) #str 69 #define FSE_EXPAND_AND_QUOTE(str) FSE_QUOTE(str) 70 #define FSE_VERSION_STRING FSE_EXPAND_AND_QUOTE(FSE_LIB_VERSION) 71 72 #define FSE_VERSION_NUMBER (FSE_VERSION_MAJOR *100*100 + FSE_VERSION_MINOR *100 + FSE_VERSION_RELEASE) 73 FSE_PUBLIC_API unsigned FSE_versionNumber(void); /**< library version number; to be used when checking dll version */ 74 75 76 /*-**************************************** 77 * FSE simple functions 78 ******************************************/ 79 /*! FSE_compress() : 80 Compress content of buffer 'src', of size 'srcSize', into destination buffer 'dst'. 81 'dst' buffer must be already allocated. Compression runs faster is dstCapacity >= FSE_compressBound(srcSize). 82 @return : size of compressed data (<= dstCapacity). 83 Special values : if return == 0, srcData is not compressible => Nothing is stored within dst !!! 84 if return == 1, srcData is a single byte symbol * srcSize times. Use RLE compression instead. 85 if FSE_isError(return), compression failed (more details using FSE_getErrorName()) 86 */ 87 FSE_PUBLIC_API size_t FSE_compress(void* dst, size_t dstCapacity, 88 const void* src, size_t srcSize); 89 90 /*! FSE_decompress(): 91 Decompress FSE data from buffer 'cSrc', of size 'cSrcSize', 92 into already allocated destination buffer 'dst', of size 'dstCapacity'. 93 @return : size of regenerated data (<= maxDstSize), 94 or an error code, which can be tested using FSE_isError() . 95 96 ** Important ** : FSE_decompress() does not decompress non-compressible nor RLE data !!! 97 Why ? : making this distinction requires a header. 98 Header management is intentionally delegated to the user layer, which can better manage special cases. 99 */ 100 FSE_PUBLIC_API size_t FSE_decompress(void* dst, size_t dstCapacity, 101 const void* cSrc, size_t cSrcSize); 102 103 104 /*-***************************************** 105 * Tool functions 106 ******************************************/ 107 FSE_PUBLIC_API size_t FSE_compressBound(size_t size); /* maximum compressed size */ 108 109 /* Error Management */ 110 FSE_PUBLIC_API unsigned FSE_isError(size_t code); /* tells if a return value is an error code */ 111 FSE_PUBLIC_API const char* FSE_getErrorName(size_t code); /* provides error code string (useful for debugging) */ 112 113 114 /*-***************************************** 115 * FSE advanced functions 116 ******************************************/ 117 /*! FSE_compress2() : 118 Same as FSE_compress(), but allows the selection of 'maxSymbolValue' and 'tableLog' 119 Both parameters can be defined as '0' to mean : use default value 120 @return : size of compressed data 121 Special values : if return == 0, srcData is not compressible => Nothing is stored within cSrc !!! 122 if return == 1, srcData is a single byte symbol * srcSize times. Use RLE compression. 123 if FSE_isError(return), it's an error code. 124 */ 125 FSE_PUBLIC_API size_t FSE_compress2 (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog); 126 127 128 /*-***************************************** 129 * FSE detailed API 130 ******************************************/ 131 /*! 132 FSE_compress() does the following: 133 1. count symbol occurrence from source[] into table count[] (see hist.h) 134 2. normalize counters so that sum(count[]) == Power_of_2 (2^tableLog) 135 3. save normalized counters to memory buffer using writeNCount() 136 4. build encoding table 'CTable' from normalized counters 137 5. encode the data stream using encoding table 'CTable' 138 139 FSE_decompress() does the following: 140 1. read normalized counters with readNCount() 141 2. build decoding table 'DTable' from normalized counters 142 3. decode the data stream using decoding table 'DTable' 143 144 The following API allows targeting specific sub-functions for advanced tasks. 145 For example, it's possible to compress several blocks using the same 'CTable', 146 or to save and provide normalized distribution using external method. 147 */ 148 149 /* *** COMPRESSION *** */ 150 151 /*! FSE_optimalTableLog(): 152 dynamically downsize 'tableLog' when conditions are met. 153 It saves CPU time, by using smaller tables, while preserving or even improving compression ratio. 154 @return : recommended tableLog (necessarily <= 'maxTableLog') */ 155 FSE_PUBLIC_API unsigned FSE_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue); 156 157 /*! FSE_normalizeCount(): 158 normalize counts so that sum(count[]) == Power_of_2 (2^tableLog) 159 'normalizedCounter' is a table of short, of minimum size (maxSymbolValue+1). 160 @return : tableLog, 161 or an errorCode, which can be tested using FSE_isError() */ 162 FSE_PUBLIC_API size_t FSE_normalizeCount(short* normalizedCounter, unsigned tableLog, 163 const unsigned* count, size_t srcSize, unsigned maxSymbolValue); 164 165 /*! FSE_NCountWriteBound(): 166 Provides the maximum possible size of an FSE normalized table, given 'maxSymbolValue' and 'tableLog'. 167 Typically useful for allocation purpose. */ 168 FSE_PUBLIC_API size_t FSE_NCountWriteBound(unsigned maxSymbolValue, unsigned tableLog); 169 170 /*! FSE_writeNCount(): 171 Compactly save 'normalizedCounter' into 'buffer'. 172 @return : size of the compressed table, 173 or an errorCode, which can be tested using FSE_isError(). */ 174 FSE_PUBLIC_API size_t FSE_writeNCount (void* buffer, size_t bufferSize, 175 const short* normalizedCounter, 176 unsigned maxSymbolValue, unsigned tableLog); 177 178 /*! Constructor and Destructor of FSE_CTable. 179 Note that FSE_CTable size depends on 'tableLog' and 'maxSymbolValue' */ 180 typedef unsigned FSE_CTable; /* don't allocate that. It's only meant to be more restrictive than void* */ 181 FSE_PUBLIC_API FSE_CTable* FSE_createCTable (unsigned maxSymbolValue, unsigned tableLog); 182 FSE_PUBLIC_API void FSE_freeCTable (FSE_CTable* ct); 183 184 /*! FSE_buildCTable(): 185 Builds `ct`, which must be already allocated, using FSE_createCTable(). 186 @return : 0, or an errorCode, which can be tested using FSE_isError() */ 187 FSE_PUBLIC_API size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog); 188 189 /*! FSE_compress_usingCTable(): 190 Compress `src` using `ct` into `dst` which must be already allocated. 191 @return : size of compressed data (<= `dstCapacity`), 192 or 0 if compressed data could not fit into `dst`, 193 or an errorCode, which can be tested using FSE_isError() */ 194 FSE_PUBLIC_API size_t FSE_compress_usingCTable (void* dst, size_t dstCapacity, const void* src, size_t srcSize, const FSE_CTable* ct); 195 196 /*! 197 Tutorial : 198 ---------- 199 The first step is to count all symbols. FSE_count() does this job very fast. 200 Result will be saved into 'count', a table of unsigned int, which must be already allocated, and have 'maxSymbolValuePtr[0]+1' cells. 201 'src' is a table of bytes of size 'srcSize'. All values within 'src' MUST be <= maxSymbolValuePtr[0] 202 maxSymbolValuePtr[0] will be updated, with its real value (necessarily <= original value) 203 FSE_count() will return the number of occurrence of the most frequent symbol. 204 This can be used to know if there is a single symbol within 'src', and to quickly evaluate its compressibility. 205 If there is an error, the function will return an ErrorCode (which can be tested using FSE_isError()). 206 207 The next step is to normalize the frequencies. 208 FSE_normalizeCount() will ensure that sum of frequencies is == 2 ^'tableLog'. 209 It also guarantees a minimum of 1 to any Symbol with frequency >= 1. 210 You can use 'tableLog'==0 to mean "use default tableLog value". 211 If you are unsure of which tableLog value to use, you can ask FSE_optimalTableLog(), 212 which will provide the optimal valid tableLog given sourceSize, maxSymbolValue, and a user-defined maximum (0 means "default"). 213 214 The result of FSE_normalizeCount() will be saved into a table, 215 called 'normalizedCounter', which is a table of signed short. 216 'normalizedCounter' must be already allocated, and have at least 'maxSymbolValue+1' cells. 217 The return value is tableLog if everything proceeded as expected. 218 It is 0 if there is a single symbol within distribution. 219 If there is an error (ex: invalid tableLog value), the function will return an ErrorCode (which can be tested using FSE_isError()). 220 221 'normalizedCounter' can be saved in a compact manner to a memory area using FSE_writeNCount(). 222 'buffer' must be already allocated. 223 For guaranteed success, buffer size must be at least FSE_headerBound(). 224 The result of the function is the number of bytes written into 'buffer'. 225 If there is an error, the function will return an ErrorCode (which can be tested using FSE_isError(); ex : buffer size too small). 226 227 'normalizedCounter' can then be used to create the compression table 'CTable'. 228 The space required by 'CTable' must be already allocated, using FSE_createCTable(). 229 You can then use FSE_buildCTable() to fill 'CTable'. 230 If there is an error, both functions will return an ErrorCode (which can be tested using FSE_isError()). 231 232 'CTable' can then be used to compress 'src', with FSE_compress_usingCTable(). 233 Similar to FSE_count(), the convention is that 'src' is assumed to be a table of char of size 'srcSize' 234 The function returns the size of compressed data (without header), necessarily <= `dstCapacity`. 235 If it returns '0', compressed data could not fit into 'dst'. 236 If there is an error, the function will return an ErrorCode (which can be tested using FSE_isError()). 237 */ 238 239 240 /* *** DECOMPRESSION *** */ 241 242 /*! FSE_readNCount(): 243 Read compactly saved 'normalizedCounter' from 'rBuffer'. 244 @return : size read from 'rBuffer', 245 or an errorCode, which can be tested using FSE_isError(). 246 maxSymbolValuePtr[0] and tableLogPtr[0] will also be updated with their respective values */ 247 FSE_PUBLIC_API size_t FSE_readNCount (short* normalizedCounter, 248 unsigned* maxSymbolValuePtr, unsigned* tableLogPtr, 249 const void* rBuffer, size_t rBuffSize); 250 251 /*! Constructor and Destructor of FSE_DTable. 252 Note that its size depends on 'tableLog' */ 253 typedef unsigned FSE_DTable; /* don't allocate that. It's just a way to be more restrictive than void* */ 254 FSE_PUBLIC_API FSE_DTable* FSE_createDTable(unsigned tableLog); 255 FSE_PUBLIC_API void FSE_freeDTable(FSE_DTable* dt); 256 257 /*! FSE_buildDTable(): 258 Builds 'dt', which must be already allocated, using FSE_createDTable(). 259 return : 0, or an errorCode, which can be tested using FSE_isError() */ 260 FSE_PUBLIC_API size_t FSE_buildDTable (FSE_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog); 261 262 /*! FSE_decompress_usingDTable(): 263 Decompress compressed source `cSrc` of size `cSrcSize` using `dt` 264 into `dst` which must be already allocated. 265 @return : size of regenerated data (necessarily <= `dstCapacity`), 266 or an errorCode, which can be tested using FSE_isError() */ 267 FSE_PUBLIC_API size_t FSE_decompress_usingDTable(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, const FSE_DTable* dt); 268 269 /*! 270 Tutorial : 271 ---------- 272 (Note : these functions only decompress FSE-compressed blocks. 273 If block is uncompressed, use memcpy() instead 274 If block is a single repeated byte, use memset() instead ) 275 276 The first step is to obtain the normalized frequencies of symbols. 277 This can be performed by FSE_readNCount() if it was saved using FSE_writeNCount(). 278 'normalizedCounter' must be already allocated, and have at least 'maxSymbolValuePtr[0]+1' cells of signed short. 279 In practice, that means it's necessary to know 'maxSymbolValue' beforehand, 280 or size the table to handle worst case situations (typically 256). 281 FSE_readNCount() will provide 'tableLog' and 'maxSymbolValue'. 282 The result of FSE_readNCount() is the number of bytes read from 'rBuffer'. 283 Note that 'rBufferSize' must be at least 4 bytes, even if useful information is less than that. 284 If there is an error, the function will return an error code, which can be tested using FSE_isError(). 285 286 The next step is to build the decompression tables 'FSE_DTable' from 'normalizedCounter'. 287 This is performed by the function FSE_buildDTable(). 288 The space required by 'FSE_DTable' must be already allocated using FSE_createDTable(). 289 If there is an error, the function will return an error code, which can be tested using FSE_isError(). 290 291 `FSE_DTable` can then be used to decompress `cSrc`, with FSE_decompress_usingDTable(). 292 `cSrcSize` must be strictly correct, otherwise decompression will fail. 293 FSE_decompress_usingDTable() result will tell how many bytes were regenerated (<=`dstCapacity`). 294 If there is an error, the function will return an error code, which can be tested using FSE_isError(). (ex: dst buffer too small) 295 */ 296 297 #endif /* FSE_H */ 298 299 #if defined(FSE_STATIC_LINKING_ONLY) && !defined(FSE_H_FSE_STATIC_LINKING_ONLY) 300 #define FSE_H_FSE_STATIC_LINKING_ONLY 301 302 /* *** Dependency *** */ 303 #include "bitstream.h" 304 305 306 /* ***************************************** 307 * Static allocation 308 *******************************************/ 309 /* FSE buffer bounds */ 310 #define FSE_NCOUNTBOUND 512 311 #define FSE_BLOCKBOUND(size) (size + (size>>7)) 312 #define FSE_COMPRESSBOUND(size) (FSE_NCOUNTBOUND + FSE_BLOCKBOUND(size)) /* Macro version, useful for static allocation */ 313 314 /* It is possible to statically allocate FSE CTable/DTable as a table of FSE_CTable/FSE_DTable using below macros */ 315 #define FSE_CTABLE_SIZE_U32(maxTableLog, maxSymbolValue) (1 + (1<<(maxTableLog-1)) + ((maxSymbolValue+1)*2)) 316 #define FSE_DTABLE_SIZE_U32(maxTableLog) (1 + (1<<maxTableLog)) 317 318 /* or use the size to malloc() space directly. Pay attention to alignment restrictions though */ 319 #define FSE_CTABLE_SIZE(maxTableLog, maxSymbolValue) (FSE_CTABLE_SIZE_U32(maxTableLog, maxSymbolValue) * sizeof(FSE_CTable)) 320 #define FSE_DTABLE_SIZE(maxTableLog) (FSE_DTABLE_SIZE_U32(maxTableLog) * sizeof(FSE_DTable)) 321 322 323 /* ***************************************** 324 * FSE advanced API 325 ***************************************** */ 326 327 unsigned FSE_optimalTableLog_internal(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue, unsigned minus); 328 /**< same as FSE_optimalTableLog(), which used `minus==2` */ 329 330 /* FSE_compress_wksp() : 331 * Same as FSE_compress2(), but using an externally allocated scratch buffer (`workSpace`). 332 * FSE_WKSP_SIZE_U32() provides the minimum size required for `workSpace` as a table of FSE_CTable. 333 */ 334 #define FSE_WKSP_SIZE_U32(maxTableLog, maxSymbolValue) ( FSE_CTABLE_SIZE_U32(maxTableLog, maxSymbolValue) + ((maxTableLog > 12) ? (1 << (maxTableLog - 2)) : 1024) ) 335 size_t FSE_compress_wksp (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize); 336 337 size_t FSE_buildCTable_raw (FSE_CTable* ct, unsigned nbBits); 338 /**< build a fake FSE_CTable, designed for a flat distribution, where each symbol uses nbBits */ 339 340 size_t FSE_buildCTable_rle (FSE_CTable* ct, unsigned char symbolValue); 341 /**< build a fake FSE_CTable, designed to compress always the same symbolValue */ 342 343 /* FSE_buildCTable_wksp() : 344 * Same as FSE_buildCTable(), but using an externally allocated scratch buffer (`workSpace`). 345 * `wkspSize` must be >= `(1<<tableLog)`. 346 */ 347 size_t FSE_buildCTable_wksp(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize); 348 349 size_t FSE_buildDTable_raw (FSE_DTable* dt, unsigned nbBits); 350 /**< build a fake FSE_DTable, designed to read a flat distribution where each symbol uses nbBits */ 351 352 size_t FSE_buildDTable_rle (FSE_DTable* dt, unsigned char symbolValue); 353 /**< build a fake FSE_DTable, designed to always generate the same symbolValue */ 354 355 size_t FSE_decompress_wksp(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, FSE_DTable* workSpace, unsigned maxLog); 356 /**< same as FSE_decompress(), using an externally allocated `workSpace` produced with `FSE_DTABLE_SIZE_U32(maxLog)` */ 357 358 typedef enum { 359 FSE_repeat_none, /**< Cannot use the previous table */ 360 FSE_repeat_check, /**< Can use the previous table but it must be checked */ 361 FSE_repeat_valid /**< Can use the previous table and it is assumed to be valid */ 362 } FSE_repeat; 363 364 /* ***************************************** 365 * FSE symbol compression API 366 *******************************************/ 367 /*! 368 This API consists of small unitary functions, which highly benefit from being inlined. 369 Hence their body are included in next section. 370 */ 371 typedef struct { 372 ptrdiff_t value; 373 const void* stateTable; 374 const void* symbolTT; 375 unsigned stateLog; 376 } FSE_CState_t; 377 378 static void FSE_initCState(FSE_CState_t* CStatePtr, const FSE_CTable* ct); 379 380 static void FSE_encodeSymbol(BIT_CStream_t* bitC, FSE_CState_t* CStatePtr, unsigned symbol); 381 382 static void FSE_flushCState(BIT_CStream_t* bitC, const FSE_CState_t* CStatePtr); 383 384 /**< 385 These functions are inner components of FSE_compress_usingCTable(). 386 They allow the creation of custom streams, mixing multiple tables and bit sources. 387 388 A key property to keep in mind is that encoding and decoding are done **in reverse direction**. 389 So the first symbol you will encode is the last you will decode, like a LIFO stack. 390 391 You will need a few variables to track your CStream. They are : 392 393 FSE_CTable ct; // Provided by FSE_buildCTable() 394 BIT_CStream_t bitStream; // bitStream tracking structure 395 FSE_CState_t state; // State tracking structure (can have several) 396 397 398 The first thing to do is to init bitStream and state. 399 size_t errorCode = BIT_initCStream(&bitStream, dstBuffer, maxDstSize); 400 FSE_initCState(&state, ct); 401 402 Note that BIT_initCStream() can produce an error code, so its result should be tested, using FSE_isError(); 403 You can then encode your input data, byte after byte. 404 FSE_encodeSymbol() outputs a maximum of 'tableLog' bits at a time. 405 Remember decoding will be done in reverse direction. 406 FSE_encodeByte(&bitStream, &state, symbol); 407 408 At any time, you can also add any bit sequence. 409 Note : maximum allowed nbBits is 25, for compatibility with 32-bits decoders 410 BIT_addBits(&bitStream, bitField, nbBits); 411 412 The above methods don't commit data to memory, they just store it into local register, for speed. 413 Local register size is 64-bits on 64-bits systems, 32-bits on 32-bits systems (size_t). 414 Writing data to memory is a manual operation, performed by the flushBits function. 415 BIT_flushBits(&bitStream); 416 417 Your last FSE encoding operation shall be to flush your last state value(s). 418 FSE_flushState(&bitStream, &state); 419 420 Finally, you must close the bitStream. 421 The function returns the size of CStream in bytes. 422 If data couldn't fit into dstBuffer, it will return a 0 ( == not compressible) 423 If there is an error, it returns an errorCode (which can be tested using FSE_isError()). 424 size_t size = BIT_closeCStream(&bitStream); 425 */ 426 427 428 /* ***************************************** 429 * FSE symbol decompression API 430 *******************************************/ 431 typedef struct { 432 size_t state; 433 const void* table; /* precise table may vary, depending on U16 */ 434 } FSE_DState_t; 435 436 437 static void FSE_initDState(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD, const FSE_DTable* dt); 438 439 static unsigned char FSE_decodeSymbol(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD); 440 441 static unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr); 442 443 /**< 444 Let's now decompose FSE_decompress_usingDTable() into its unitary components. 445 You will decode FSE-encoded symbols from the bitStream, 446 and also any other bitFields you put in, **in reverse order**. 447 448 You will need a few variables to track your bitStream. They are : 449 450 BIT_DStream_t DStream; // Stream context 451 FSE_DState_t DState; // State context. Multiple ones are possible 452 FSE_DTable* DTablePtr; // Decoding table, provided by FSE_buildDTable() 453 454 The first thing to do is to init the bitStream. 455 errorCode = BIT_initDStream(&DStream, srcBuffer, srcSize); 456 457 You should then retrieve your initial state(s) 458 (in reverse flushing order if you have several ones) : 459 errorCode = FSE_initDState(&DState, &DStream, DTablePtr); 460 461 You can then decode your data, symbol after symbol. 462 For information the maximum number of bits read by FSE_decodeSymbol() is 'tableLog'. 463 Keep in mind that symbols are decoded in reverse order, like a LIFO stack (last in, first out). 464 unsigned char symbol = FSE_decodeSymbol(&DState, &DStream); 465 466 You can retrieve any bitfield you eventually stored into the bitStream (in reverse order) 467 Note : maximum allowed nbBits is 25, for 32-bits compatibility 468 size_t bitField = BIT_readBits(&DStream, nbBits); 469 470 All above operations only read from local register (which size depends on size_t). 471 Refueling the register from memory is manually performed by the reload method. 472 endSignal = FSE_reloadDStream(&DStream); 473 474 BIT_reloadDStream() result tells if there is still some more data to read from DStream. 475 BIT_DStream_unfinished : there is still some data left into the DStream. 476 BIT_DStream_endOfBuffer : Dstream reached end of buffer. Its container may no longer be completely filled. 477 BIT_DStream_completed : Dstream reached its exact end, corresponding in general to decompression completed. 478 BIT_DStream_tooFar : Dstream went too far. Decompression result is corrupted. 479 480 When reaching end of buffer (BIT_DStream_endOfBuffer), progress slowly, notably if you decode multiple symbols per loop, 481 to properly detect the exact end of stream. 482 After each decoded symbol, check if DStream is fully consumed using this simple test : 483 BIT_reloadDStream(&DStream) >= BIT_DStream_completed 484 485 When it's done, verify decompression is fully completed, by checking both DStream and the relevant states. 486 Checking if DStream has reached its end is performed by : 487 BIT_endOfDStream(&DStream); 488 Check also the states. There might be some symbols left there, if some high probability ones (>50%) are possible. 489 FSE_endOfDState(&DState); 490 */ 491 492 493 /* ***************************************** 494 * FSE unsafe API 495 *******************************************/ 496 static unsigned char FSE_decodeSymbolFast(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD); 497 /* faster, but works only if nbBits is always >= 1 (otherwise, result will be corrupted) */ 498 499 500 /* ***************************************** 501 * Implementation of inlined functions 502 *******************************************/ 503 typedef struct { 504 int deltaFindState; 505 U32 deltaNbBits; 506 } FSE_symbolCompressionTransform; /* total 8 bytes */ 507 508 MEM_STATIC void FSE_initCState(FSE_CState_t* statePtr, const FSE_CTable* ct) 509 { 510 const void* ptr = ct; 511 const U16* u16ptr = (const U16*) ptr; 512 const U32 tableLog = MEM_read16(ptr); 513 statePtr->value = (ptrdiff_t)1<<tableLog; 514 statePtr->stateTable = u16ptr+2; 515 statePtr->symbolTT = ct + 1 + (tableLog ? (1<<(tableLog-1)) : 1); 516 statePtr->stateLog = tableLog; 517 } 518 519 520 /*! FSE_initCState2() : 521 * Same as FSE_initCState(), but the first symbol to include (which will be the last to be read) 522 * uses the smallest state value possible, saving the cost of this symbol */ 523 MEM_STATIC void FSE_initCState2(FSE_CState_t* statePtr, const FSE_CTable* ct, U32 symbol) 524 { 525 FSE_initCState(statePtr, ct); 526 { const FSE_symbolCompressionTransform symbolTT = ((const FSE_symbolCompressionTransform*)(statePtr->symbolTT))[symbol]; 527 const U16* stateTable = (const U16*)(statePtr->stateTable); 528 U32 nbBitsOut = (U32)((symbolTT.deltaNbBits + (1<<15)) >> 16); 529 statePtr->value = (nbBitsOut << 16) - symbolTT.deltaNbBits; 530 statePtr->value = stateTable[(statePtr->value >> nbBitsOut) + symbolTT.deltaFindState]; 531 } 532 } 533 534 MEM_STATIC void FSE_encodeSymbol(BIT_CStream_t* bitC, FSE_CState_t* statePtr, unsigned symbol) 535 { 536 FSE_symbolCompressionTransform const symbolTT = ((const FSE_symbolCompressionTransform*)(statePtr->symbolTT))[symbol]; 537 const U16* const stateTable = (const U16*)(statePtr->stateTable); 538 U32 const nbBitsOut = (U32)((statePtr->value + symbolTT.deltaNbBits) >> 16); 539 BIT_addBits(bitC, statePtr->value, nbBitsOut); 540 statePtr->value = stateTable[ (statePtr->value >> nbBitsOut) + symbolTT.deltaFindState]; 541 } 542 543 MEM_STATIC void FSE_flushCState(BIT_CStream_t* bitC, const FSE_CState_t* statePtr) 544 { 545 BIT_addBits(bitC, statePtr->value, statePtr->stateLog); 546 BIT_flushBits(bitC); 547 } 548 549 550 /* FSE_getMaxNbBits() : 551 * Approximate maximum cost of a symbol, in bits. 552 * Fractional get rounded up (i.e : a symbol with a normalized frequency of 3 gives the same result as a frequency of 2) 553 * note 1 : assume symbolValue is valid (<= maxSymbolValue) 554 * note 2 : if freq[symbolValue]==0, @return a fake cost of tableLog+1 bits */ 555 MEM_STATIC U32 FSE_getMaxNbBits(const void* symbolTTPtr, U32 symbolValue) 556 { 557 const FSE_symbolCompressionTransform* symbolTT = (const FSE_symbolCompressionTransform*) symbolTTPtr; 558 return (symbolTT[symbolValue].deltaNbBits + ((1<<16)-1)) >> 16; 559 } 560 561 /* FSE_bitCost() : 562 * Approximate symbol cost, as fractional value, using fixed-point format (accuracyLog fractional bits) 563 * note 1 : assume symbolValue is valid (<= maxSymbolValue) 564 * note 2 : if freq[symbolValue]==0, @return a fake cost of tableLog+1 bits */ 565 MEM_STATIC U32 FSE_bitCost(const void* symbolTTPtr, U32 tableLog, U32 symbolValue, U32 accuracyLog) 566 { 567 const FSE_symbolCompressionTransform* symbolTT = (const FSE_symbolCompressionTransform*) symbolTTPtr; 568 U32 const minNbBits = symbolTT[symbolValue].deltaNbBits >> 16; 569 U32 const threshold = (minNbBits+1) << 16; 570 assert(tableLog < 16); 571 assert(accuracyLog < 31-tableLog); /* ensure enough room for renormalization double shift */ 572 { U32 const tableSize = 1 << tableLog; 573 U32 const deltaFromThreshold = threshold - (symbolTT[symbolValue].deltaNbBits + tableSize); 574 U32 const normalizedDeltaFromThreshold = (deltaFromThreshold << accuracyLog) >> tableLog; /* linear interpolation (very approximate) */ 575 U32 const bitMultiplier = 1 << accuracyLog; 576 assert(symbolTT[symbolValue].deltaNbBits + tableSize <= threshold); 577 assert(normalizedDeltaFromThreshold <= bitMultiplier); 578 return (minNbBits+1)*bitMultiplier - normalizedDeltaFromThreshold; 579 } 580 } 581 582 583 /* ====== Decompression ====== */ 584 585 typedef struct { 586 U16 tableLog; 587 U16 fastMode; 588 } FSE_DTableHeader; /* sizeof U32 */ 589 590 typedef struct 591 { 592 unsigned short newState; 593 unsigned char symbol; 594 unsigned char nbBits; 595 } FSE_decode_t; /* size == U32 */ 596 597 MEM_STATIC void FSE_initDState(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD, const FSE_DTable* dt) 598 { 599 const void* ptr = dt; 600 const FSE_DTableHeader* const DTableH = (const FSE_DTableHeader*)ptr; 601 DStatePtr->state = BIT_readBits(bitD, DTableH->tableLog); 602 BIT_reloadDStream(bitD); 603 DStatePtr->table = dt + 1; 604 } 605 606 MEM_STATIC BYTE FSE_peekSymbol(const FSE_DState_t* DStatePtr) 607 { 608 FSE_decode_t const DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state]; 609 return DInfo.symbol; 610 } 611 612 MEM_STATIC void FSE_updateState(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD) 613 { 614 FSE_decode_t const DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state]; 615 U32 const nbBits = DInfo.nbBits; 616 size_t const lowBits = BIT_readBits(bitD, nbBits); 617 DStatePtr->state = DInfo.newState + lowBits; 618 } 619 620 MEM_STATIC BYTE FSE_decodeSymbol(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD) 621 { 622 FSE_decode_t const DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state]; 623 U32 const nbBits = DInfo.nbBits; 624 BYTE const symbol = DInfo.symbol; 625 size_t const lowBits = BIT_readBits(bitD, nbBits); 626 627 DStatePtr->state = DInfo.newState + lowBits; 628 return symbol; 629 } 630 631 /*! FSE_decodeSymbolFast() : 632 unsafe, only works if no symbol has a probability > 50% */ 633 MEM_STATIC BYTE FSE_decodeSymbolFast(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD) 634 { 635 FSE_decode_t const DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state]; 636 U32 const nbBits = DInfo.nbBits; 637 BYTE const symbol = DInfo.symbol; 638 size_t const lowBits = BIT_readBitsFast(bitD, nbBits); 639 640 DStatePtr->state = DInfo.newState + lowBits; 641 return symbol; 642 } 643 644 MEM_STATIC unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr) 645 { 646 return DStatePtr->state == 0; 647 } 648 649 650 651 #ifndef FSE_COMMONDEFS_ONLY 652 653 /* ************************************************************** 654 * Tuning parameters 655 ****************************************************************/ 656 /*!MEMORY_USAGE : 657 * Memory usage formula : N->2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; etc.) 658 * Increasing memory usage improves compression ratio 659 * Reduced memory usage can improve speed, due to cache effect 660 * Recommended max value is 14, for 16KB, which nicely fits into Intel x86 L1 cache */ 661 #ifndef FSE_MAX_MEMORY_USAGE 662 # define FSE_MAX_MEMORY_USAGE 14 663 #endif 664 #ifndef FSE_DEFAULT_MEMORY_USAGE 665 # define FSE_DEFAULT_MEMORY_USAGE 13 666 #endif 667 668 /*!FSE_MAX_SYMBOL_VALUE : 669 * Maximum symbol value authorized. 670 * Required for proper stack allocation */ 671 #ifndef FSE_MAX_SYMBOL_VALUE 672 # define FSE_MAX_SYMBOL_VALUE 255 673 #endif 674 675 /* ************************************************************** 676 * template functions type & suffix 677 ****************************************************************/ 678 #define FSE_FUNCTION_TYPE BYTE 679 #define FSE_FUNCTION_EXTENSION 680 #define FSE_DECODE_TYPE FSE_decode_t 681 682 683 #endif /* !FSE_COMMONDEFS_ONLY */ 684 685 686 /* *************************************************************** 687 * Constants 688 *****************************************************************/ 689 #define FSE_MAX_TABLELOG (FSE_MAX_MEMORY_USAGE-2) 690 #define FSE_MAX_TABLESIZE (1U<<FSE_MAX_TABLELOG) 691 #define FSE_MAXTABLESIZE_MASK (FSE_MAX_TABLESIZE-1) 692 #define FSE_DEFAULT_TABLELOG (FSE_DEFAULT_MEMORY_USAGE-2) 693 #define FSE_MIN_TABLELOG 5 694 695 #define FSE_TABLELOG_ABSOLUTE_MAX 15 696 #if FSE_MAX_TABLELOG > FSE_TABLELOG_ABSOLUTE_MAX 697 # error "FSE_MAX_TABLELOG > FSE_TABLELOG_ABSOLUTE_MAX is not supported" 698 #endif 699 700 #define FSE_TABLESTEP(tableSize) ((tableSize>>1) + (tableSize>>3) + 3) 701 702 703 #endif /* FSE_STATIC_LINKING_ONLY */ 704 705 706 #if defined (__cplusplus) 707 } 708 #endif 709