Book a Demo!
CoCalc Logo Icon
StoreFeaturesDocsShareSupportNewsAboutPoliciesSign UpSign In
freebsd
GitHub Repository: freebsd/freebsd-src
Path: blob/main/sys/contrib/zstd/lib/common/fse.h
48378 views
1
/* ******************************************************************
2
* FSE : Finite State Entropy codec
3
* Public Prototypes declaration
4
* Copyright (c) Yann Collet, Facebook, Inc.
5
*
6
* You can contact the author at :
7
* - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
8
*
9
* This source code is licensed under both the BSD-style license (found in the
10
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
11
* in the COPYING file in the root directory of this source tree).
12
* You may select, at your option, one of the above-listed licenses.
13
****************************************************************** */
14
15
#if defined (__cplusplus)
16
extern "C" {
17
#endif
18
19
#ifndef FSE_H
20
#define FSE_H
21
22
23
/*-*****************************************
24
* Dependencies
25
******************************************/
26
#include "zstd_deps.h" /* size_t, ptrdiff_t */
27
28
29
/*-*****************************************
30
* FSE_PUBLIC_API : control library symbols visibility
31
******************************************/
32
#if defined(FSE_DLL_EXPORT) && (FSE_DLL_EXPORT==1) && defined(__GNUC__) && (__GNUC__ >= 4)
33
# define FSE_PUBLIC_API __attribute__ ((visibility ("default")))
34
#elif defined(FSE_DLL_EXPORT) && (FSE_DLL_EXPORT==1) /* Visual expected */
35
# define FSE_PUBLIC_API __declspec(dllexport)
36
#elif defined(FSE_DLL_IMPORT) && (FSE_DLL_IMPORT==1)
37
# 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.*/
38
#else
39
# define FSE_PUBLIC_API
40
#endif
41
42
/*------ Version ------*/
43
#define FSE_VERSION_MAJOR 0
44
#define FSE_VERSION_MINOR 9
45
#define FSE_VERSION_RELEASE 0
46
47
#define FSE_LIB_VERSION FSE_VERSION_MAJOR.FSE_VERSION_MINOR.FSE_VERSION_RELEASE
48
#define FSE_QUOTE(str) #str
49
#define FSE_EXPAND_AND_QUOTE(str) FSE_QUOTE(str)
50
#define FSE_VERSION_STRING FSE_EXPAND_AND_QUOTE(FSE_LIB_VERSION)
51
52
#define FSE_VERSION_NUMBER (FSE_VERSION_MAJOR *100*100 + FSE_VERSION_MINOR *100 + FSE_VERSION_RELEASE)
53
FSE_PUBLIC_API unsigned FSE_versionNumber(void); /**< library version number; to be used when checking dll version */
54
55
56
/*-****************************************
57
* FSE simple functions
58
******************************************/
59
/*! FSE_compress() :
60
Compress content of buffer 'src', of size 'srcSize', into destination buffer 'dst'.
61
'dst' buffer must be already allocated. Compression runs faster is dstCapacity >= FSE_compressBound(srcSize).
62
@return : size of compressed data (<= dstCapacity).
63
Special values : if return == 0, srcData is not compressible => Nothing is stored within dst !!!
64
if return == 1, srcData is a single byte symbol * srcSize times. Use RLE compression instead.
65
if FSE_isError(return), compression failed (more details using FSE_getErrorName())
66
*/
67
FSE_PUBLIC_API size_t FSE_compress(void* dst, size_t dstCapacity,
68
const void* src, size_t srcSize);
69
70
/*! FSE_decompress():
71
Decompress FSE data from buffer 'cSrc', of size 'cSrcSize',
72
into already allocated destination buffer 'dst', of size 'dstCapacity'.
73
@return : size of regenerated data (<= maxDstSize),
74
or an error code, which can be tested using FSE_isError() .
75
76
** Important ** : FSE_decompress() does not decompress non-compressible nor RLE data !!!
77
Why ? : making this distinction requires a header.
78
Header management is intentionally delegated to the user layer, which can better manage special cases.
79
*/
80
FSE_PUBLIC_API size_t FSE_decompress(void* dst, size_t dstCapacity,
81
const void* cSrc, size_t cSrcSize);
82
83
84
/*-*****************************************
85
* Tool functions
86
******************************************/
87
FSE_PUBLIC_API size_t FSE_compressBound(size_t size); /* maximum compressed size */
88
89
/* Error Management */
90
FSE_PUBLIC_API unsigned FSE_isError(size_t code); /* tells if a return value is an error code */
91
FSE_PUBLIC_API const char* FSE_getErrorName(size_t code); /* provides error code string (useful for debugging) */
92
93
94
/*-*****************************************
95
* FSE advanced functions
96
******************************************/
97
/*! FSE_compress2() :
98
Same as FSE_compress(), but allows the selection of 'maxSymbolValue' and 'tableLog'
99
Both parameters can be defined as '0' to mean : use default value
100
@return : size of compressed data
101
Special values : if return == 0, srcData is not compressible => Nothing is stored within cSrc !!!
102
if return == 1, srcData is a single byte symbol * srcSize times. Use RLE compression.
103
if FSE_isError(return), it's an error code.
104
*/
105
FSE_PUBLIC_API size_t FSE_compress2 (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog);
106
107
108
/*-*****************************************
109
* FSE detailed API
110
******************************************/
111
/*!
112
FSE_compress() does the following:
113
1. count symbol occurrence from source[] into table count[] (see hist.h)
114
2. normalize counters so that sum(count[]) == Power_of_2 (2^tableLog)
115
3. save normalized counters to memory buffer using writeNCount()
116
4. build encoding table 'CTable' from normalized counters
117
5. encode the data stream using encoding table 'CTable'
118
119
FSE_decompress() does the following:
120
1. read normalized counters with readNCount()
121
2. build decoding table 'DTable' from normalized counters
122
3. decode the data stream using decoding table 'DTable'
123
124
The following API allows targeting specific sub-functions for advanced tasks.
125
For example, it's possible to compress several blocks using the same 'CTable',
126
or to save and provide normalized distribution using external method.
127
*/
128
129
/* *** COMPRESSION *** */
130
131
/*! FSE_optimalTableLog():
132
dynamically downsize 'tableLog' when conditions are met.
133
It saves CPU time, by using smaller tables, while preserving or even improving compression ratio.
134
@return : recommended tableLog (necessarily <= 'maxTableLog') */
135
FSE_PUBLIC_API unsigned FSE_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue);
136
137
/*! FSE_normalizeCount():
138
normalize counts so that sum(count[]) == Power_of_2 (2^tableLog)
139
'normalizedCounter' is a table of short, of minimum size (maxSymbolValue+1).
140
useLowProbCount is a boolean parameter which trades off compressed size for
141
faster header decoding. When it is set to 1, the compressed data will be slightly
142
smaller. And when it is set to 0, FSE_readNCount() and FSE_buildDTable() will be
143
faster. If you are compressing a small amount of data (< 2 KB) then useLowProbCount=0
144
is a good default, since header deserialization makes a big speed difference.
145
Otherwise, useLowProbCount=1 is a good default, since the speed difference is small.
146
@return : tableLog,
147
or an errorCode, which can be tested using FSE_isError() */
148
FSE_PUBLIC_API size_t FSE_normalizeCount(short* normalizedCounter, unsigned tableLog,
149
const unsigned* count, size_t srcSize, unsigned maxSymbolValue, unsigned useLowProbCount);
150
151
/*! FSE_NCountWriteBound():
152
Provides the maximum possible size of an FSE normalized table, given 'maxSymbolValue' and 'tableLog'.
153
Typically useful for allocation purpose. */
154
FSE_PUBLIC_API size_t FSE_NCountWriteBound(unsigned maxSymbolValue, unsigned tableLog);
155
156
/*! FSE_writeNCount():
157
Compactly save 'normalizedCounter' into 'buffer'.
158
@return : size of the compressed table,
159
or an errorCode, which can be tested using FSE_isError(). */
160
FSE_PUBLIC_API size_t FSE_writeNCount (void* buffer, size_t bufferSize,
161
const short* normalizedCounter,
162
unsigned maxSymbolValue, unsigned tableLog);
163
164
/*! Constructor and Destructor of FSE_CTable.
165
Note that FSE_CTable size depends on 'tableLog' and 'maxSymbolValue' */
166
typedef unsigned FSE_CTable; /* don't allocate that. It's only meant to be more restrictive than void* */
167
FSE_PUBLIC_API FSE_CTable* FSE_createCTable (unsigned maxSymbolValue, unsigned tableLog);
168
FSE_PUBLIC_API void FSE_freeCTable (FSE_CTable* ct);
169
170
/*! FSE_buildCTable():
171
Builds `ct`, which must be already allocated, using FSE_createCTable().
172
@return : 0, or an errorCode, which can be tested using FSE_isError() */
173
FSE_PUBLIC_API size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
174
175
/*! FSE_compress_usingCTable():
176
Compress `src` using `ct` into `dst` which must be already allocated.
177
@return : size of compressed data (<= `dstCapacity`),
178
or 0 if compressed data could not fit into `dst`,
179
or an errorCode, which can be tested using FSE_isError() */
180
FSE_PUBLIC_API size_t FSE_compress_usingCTable (void* dst, size_t dstCapacity, const void* src, size_t srcSize, const FSE_CTable* ct);
181
182
/*!
183
Tutorial :
184
----------
185
The first step is to count all symbols. FSE_count() does this job very fast.
186
Result will be saved into 'count', a table of unsigned int, which must be already allocated, and have 'maxSymbolValuePtr[0]+1' cells.
187
'src' is a table of bytes of size 'srcSize'. All values within 'src' MUST be <= maxSymbolValuePtr[0]
188
maxSymbolValuePtr[0] will be updated, with its real value (necessarily <= original value)
189
FSE_count() will return the number of occurrence of the most frequent symbol.
190
This can be used to know if there is a single symbol within 'src', and to quickly evaluate its compressibility.
191
If there is an error, the function will return an ErrorCode (which can be tested using FSE_isError()).
192
193
The next step is to normalize the frequencies.
194
FSE_normalizeCount() will ensure that sum of frequencies is == 2 ^'tableLog'.
195
It also guarantees a minimum of 1 to any Symbol with frequency >= 1.
196
You can use 'tableLog'==0 to mean "use default tableLog value".
197
If you are unsure of which tableLog value to use, you can ask FSE_optimalTableLog(),
198
which will provide the optimal valid tableLog given sourceSize, maxSymbolValue, and a user-defined maximum (0 means "default").
199
200
The result of FSE_normalizeCount() will be saved into a table,
201
called 'normalizedCounter', which is a table of signed short.
202
'normalizedCounter' must be already allocated, and have at least 'maxSymbolValue+1' cells.
203
The return value is tableLog if everything proceeded as expected.
204
It is 0 if there is a single symbol within distribution.
205
If there is an error (ex: invalid tableLog value), the function will return an ErrorCode (which can be tested using FSE_isError()).
206
207
'normalizedCounter' can be saved in a compact manner to a memory area using FSE_writeNCount().
208
'buffer' must be already allocated.
209
For guaranteed success, buffer size must be at least FSE_headerBound().
210
The result of the function is the number of bytes written into 'buffer'.
211
If there is an error, the function will return an ErrorCode (which can be tested using FSE_isError(); ex : buffer size too small).
212
213
'normalizedCounter' can then be used to create the compression table 'CTable'.
214
The space required by 'CTable' must be already allocated, using FSE_createCTable().
215
You can then use FSE_buildCTable() to fill 'CTable'.
216
If there is an error, both functions will return an ErrorCode (which can be tested using FSE_isError()).
217
218
'CTable' can then be used to compress 'src', with FSE_compress_usingCTable().
219
Similar to FSE_count(), the convention is that 'src' is assumed to be a table of char of size 'srcSize'
220
The function returns the size of compressed data (without header), necessarily <= `dstCapacity`.
221
If it returns '0', compressed data could not fit into 'dst'.
222
If there is an error, the function will return an ErrorCode (which can be tested using FSE_isError()).
223
*/
224
225
226
/* *** DECOMPRESSION *** */
227
228
/*! FSE_readNCount():
229
Read compactly saved 'normalizedCounter' from 'rBuffer'.
230
@return : size read from 'rBuffer',
231
or an errorCode, which can be tested using FSE_isError().
232
maxSymbolValuePtr[0] and tableLogPtr[0] will also be updated with their respective values */
233
FSE_PUBLIC_API size_t FSE_readNCount (short* normalizedCounter,
234
unsigned* maxSymbolValuePtr, unsigned* tableLogPtr,
235
const void* rBuffer, size_t rBuffSize);
236
237
/*! FSE_readNCount_bmi2():
238
* Same as FSE_readNCount() but pass bmi2=1 when your CPU supports BMI2 and 0 otherwise.
239
*/
240
FSE_PUBLIC_API size_t FSE_readNCount_bmi2(short* normalizedCounter,
241
unsigned* maxSymbolValuePtr, unsigned* tableLogPtr,
242
const void* rBuffer, size_t rBuffSize, int bmi2);
243
244
/*! Constructor and Destructor of FSE_DTable.
245
Note that its size depends on 'tableLog' */
246
typedef unsigned FSE_DTable; /* don't allocate that. It's just a way to be more restrictive than void* */
247
FSE_PUBLIC_API FSE_DTable* FSE_createDTable(unsigned tableLog);
248
FSE_PUBLIC_API void FSE_freeDTable(FSE_DTable* dt);
249
250
/*! FSE_buildDTable():
251
Builds 'dt', which must be already allocated, using FSE_createDTable().
252
return : 0, or an errorCode, which can be tested using FSE_isError() */
253
FSE_PUBLIC_API size_t FSE_buildDTable (FSE_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
254
255
/*! FSE_decompress_usingDTable():
256
Decompress compressed source `cSrc` of size `cSrcSize` using `dt`
257
into `dst` which must be already allocated.
258
@return : size of regenerated data (necessarily <= `dstCapacity`),
259
or an errorCode, which can be tested using FSE_isError() */
260
FSE_PUBLIC_API size_t FSE_decompress_usingDTable(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, const FSE_DTable* dt);
261
262
/*!
263
Tutorial :
264
----------
265
(Note : these functions only decompress FSE-compressed blocks.
266
If block is uncompressed, use memcpy() instead
267
If block is a single repeated byte, use memset() instead )
268
269
The first step is to obtain the normalized frequencies of symbols.
270
This can be performed by FSE_readNCount() if it was saved using FSE_writeNCount().
271
'normalizedCounter' must be already allocated, and have at least 'maxSymbolValuePtr[0]+1' cells of signed short.
272
In practice, that means it's necessary to know 'maxSymbolValue' beforehand,
273
or size the table to handle worst case situations (typically 256).
274
FSE_readNCount() will provide 'tableLog' and 'maxSymbolValue'.
275
The result of FSE_readNCount() is the number of bytes read from 'rBuffer'.
276
Note that 'rBufferSize' must be at least 4 bytes, even if useful information is less than that.
277
If there is an error, the function will return an error code, which can be tested using FSE_isError().
278
279
The next step is to build the decompression tables 'FSE_DTable' from 'normalizedCounter'.
280
This is performed by the function FSE_buildDTable().
281
The space required by 'FSE_DTable' must be already allocated using FSE_createDTable().
282
If there is an error, the function will return an error code, which can be tested using FSE_isError().
283
284
`FSE_DTable` can then be used to decompress `cSrc`, with FSE_decompress_usingDTable().
285
`cSrcSize` must be strictly correct, otherwise decompression will fail.
286
FSE_decompress_usingDTable() result will tell how many bytes were regenerated (<=`dstCapacity`).
287
If there is an error, the function will return an error code, which can be tested using FSE_isError(). (ex: dst buffer too small)
288
*/
289
290
#endif /* FSE_H */
291
292
#if defined(FSE_STATIC_LINKING_ONLY) && !defined(FSE_H_FSE_STATIC_LINKING_ONLY)
293
#define FSE_H_FSE_STATIC_LINKING_ONLY
294
295
/* *** Dependency *** */
296
#include "bitstream.h"
297
298
299
/* *****************************************
300
* Static allocation
301
*******************************************/
302
/* FSE buffer bounds */
303
#define FSE_NCOUNTBOUND 512
304
#define FSE_BLOCKBOUND(size) ((size) + ((size)>>7) + 4 /* fse states */ + sizeof(size_t) /* bitContainer */)
305
#define FSE_COMPRESSBOUND(size) (FSE_NCOUNTBOUND + FSE_BLOCKBOUND(size)) /* Macro version, useful for static allocation */
306
307
/* It is possible to statically allocate FSE CTable/DTable as a table of FSE_CTable/FSE_DTable using below macros */
308
#define FSE_CTABLE_SIZE_U32(maxTableLog, maxSymbolValue) (1 + (1<<((maxTableLog)-1)) + (((maxSymbolValue)+1)*2))
309
#define FSE_DTABLE_SIZE_U32(maxTableLog) (1 + (1<<(maxTableLog)))
310
311
/* or use the size to malloc() space directly. Pay attention to alignment restrictions though */
312
#define FSE_CTABLE_SIZE(maxTableLog, maxSymbolValue) (FSE_CTABLE_SIZE_U32(maxTableLog, maxSymbolValue) * sizeof(FSE_CTable))
313
#define FSE_DTABLE_SIZE(maxTableLog) (FSE_DTABLE_SIZE_U32(maxTableLog) * sizeof(FSE_DTable))
314
315
316
/* *****************************************
317
* FSE advanced API
318
***************************************** */
319
320
unsigned FSE_optimalTableLog_internal(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue, unsigned minus);
321
/**< same as FSE_optimalTableLog(), which used `minus==2` */
322
323
/* FSE_compress_wksp() :
324
* Same as FSE_compress2(), but using an externally allocated scratch buffer (`workSpace`).
325
* FSE_COMPRESS_WKSP_SIZE_U32() provides the minimum size required for `workSpace` as a table of FSE_CTable.
326
*/
327
#define FSE_COMPRESS_WKSP_SIZE_U32(maxTableLog, maxSymbolValue) ( FSE_CTABLE_SIZE_U32(maxTableLog, maxSymbolValue) + ((maxTableLog > 12) ? (1 << (maxTableLog - 2)) : 1024) )
328
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);
329
330
size_t FSE_buildCTable_raw (FSE_CTable* ct, unsigned nbBits);
331
/**< build a fake FSE_CTable, designed for a flat distribution, where each symbol uses nbBits */
332
333
size_t FSE_buildCTable_rle (FSE_CTable* ct, unsigned char symbolValue);
334
/**< build a fake FSE_CTable, designed to compress always the same symbolValue */
335
336
/* FSE_buildCTable_wksp() :
337
* Same as FSE_buildCTable(), but using an externally allocated scratch buffer (`workSpace`).
338
* `wkspSize` must be >= `FSE_BUILD_CTABLE_WORKSPACE_SIZE_U32(maxSymbolValue, tableLog)` of `unsigned`.
339
* See FSE_buildCTable_wksp() for breakdown of workspace usage.
340
*/
341
#define FSE_BUILD_CTABLE_WORKSPACE_SIZE_U32(maxSymbolValue, tableLog) (((maxSymbolValue + 2) + (1ull << (tableLog)))/2 + sizeof(U64)/sizeof(U32) /* additional 8 bytes for potential table overwrite */)
342
#define FSE_BUILD_CTABLE_WORKSPACE_SIZE(maxSymbolValue, tableLog) (sizeof(unsigned) * FSE_BUILD_CTABLE_WORKSPACE_SIZE_U32(maxSymbolValue, tableLog))
343
size_t FSE_buildCTable_wksp(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize);
344
345
#define FSE_BUILD_DTABLE_WKSP_SIZE(maxTableLog, maxSymbolValue) (sizeof(short) * (maxSymbolValue + 1) + (1ULL << maxTableLog) + 8)
346
#define FSE_BUILD_DTABLE_WKSP_SIZE_U32(maxTableLog, maxSymbolValue) ((FSE_BUILD_DTABLE_WKSP_SIZE(maxTableLog, maxSymbolValue) + sizeof(unsigned) - 1) / sizeof(unsigned))
347
FSE_PUBLIC_API size_t FSE_buildDTable_wksp(FSE_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize);
348
/**< Same as FSE_buildDTable(), using an externally allocated `workspace` produced with `FSE_BUILD_DTABLE_WKSP_SIZE_U32(maxSymbolValue)` */
349
350
size_t FSE_buildDTable_raw (FSE_DTable* dt, unsigned nbBits);
351
/**< build a fake FSE_DTable, designed to read a flat distribution where each symbol uses nbBits */
352
353
size_t FSE_buildDTable_rle (FSE_DTable* dt, unsigned char symbolValue);
354
/**< build a fake FSE_DTable, designed to always generate the same symbolValue */
355
356
#define FSE_DECOMPRESS_WKSP_SIZE_U32(maxTableLog, maxSymbolValue) (FSE_DTABLE_SIZE_U32(maxTableLog) + FSE_BUILD_DTABLE_WKSP_SIZE_U32(maxTableLog, maxSymbolValue) + (FSE_MAX_SYMBOL_VALUE + 1) / 2 + 1)
357
#define FSE_DECOMPRESS_WKSP_SIZE(maxTableLog, maxSymbolValue) (FSE_DECOMPRESS_WKSP_SIZE_U32(maxTableLog, maxSymbolValue) * sizeof(unsigned))
358
size_t FSE_decompress_wksp(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, unsigned maxLog, void* workSpace, size_t wkspSize);
359
/**< same as FSE_decompress(), using an externally allocated `workSpace` produced with `FSE_DECOMPRESS_WKSP_SIZE_U32(maxLog, maxSymbolValue)` */
360
361
size_t FSE_decompress_wksp_bmi2(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, unsigned maxLog, void* workSpace, size_t wkspSize, int bmi2);
362
/**< Same as FSE_decompress_wksp() but with dynamic BMI2 support. Pass 1 if your CPU supports BMI2 or 0 if it doesn't. */
363
364
typedef enum {
365
FSE_repeat_none, /**< Cannot use the previous table */
366
FSE_repeat_check, /**< Can use the previous table but it must be checked */
367
FSE_repeat_valid /**< Can use the previous table and it is assumed to be valid */
368
} FSE_repeat;
369
370
/* *****************************************
371
* FSE symbol compression API
372
*******************************************/
373
/*!
374
This API consists of small unitary functions, which highly benefit from being inlined.
375
Hence their body are included in next section.
376
*/
377
typedef struct {
378
ptrdiff_t value;
379
const void* stateTable;
380
const void* symbolTT;
381
unsigned stateLog;
382
} FSE_CState_t;
383
384
static void FSE_initCState(FSE_CState_t* CStatePtr, const FSE_CTable* ct);
385
386
static void FSE_encodeSymbol(BIT_CStream_t* bitC, FSE_CState_t* CStatePtr, unsigned symbol);
387
388
static void FSE_flushCState(BIT_CStream_t* bitC, const FSE_CState_t* CStatePtr);
389
390
/**<
391
These functions are inner components of FSE_compress_usingCTable().
392
They allow the creation of custom streams, mixing multiple tables and bit sources.
393
394
A key property to keep in mind is that encoding and decoding are done **in reverse direction**.
395
So the first symbol you will encode is the last you will decode, like a LIFO stack.
396
397
You will need a few variables to track your CStream. They are :
398
399
FSE_CTable ct; // Provided by FSE_buildCTable()
400
BIT_CStream_t bitStream; // bitStream tracking structure
401
FSE_CState_t state; // State tracking structure (can have several)
402
403
404
The first thing to do is to init bitStream and state.
405
size_t errorCode = BIT_initCStream(&bitStream, dstBuffer, maxDstSize);
406
FSE_initCState(&state, ct);
407
408
Note that BIT_initCStream() can produce an error code, so its result should be tested, using FSE_isError();
409
You can then encode your input data, byte after byte.
410
FSE_encodeSymbol() outputs a maximum of 'tableLog' bits at a time.
411
Remember decoding will be done in reverse direction.
412
FSE_encodeByte(&bitStream, &state, symbol);
413
414
At any time, you can also add any bit sequence.
415
Note : maximum allowed nbBits is 25, for compatibility with 32-bits decoders
416
BIT_addBits(&bitStream, bitField, nbBits);
417
418
The above methods don't commit data to memory, they just store it into local register, for speed.
419
Local register size is 64-bits on 64-bits systems, 32-bits on 32-bits systems (size_t).
420
Writing data to memory is a manual operation, performed by the flushBits function.
421
BIT_flushBits(&bitStream);
422
423
Your last FSE encoding operation shall be to flush your last state value(s).
424
FSE_flushState(&bitStream, &state);
425
426
Finally, you must close the bitStream.
427
The function returns the size of CStream in bytes.
428
If data couldn't fit into dstBuffer, it will return a 0 ( == not compressible)
429
If there is an error, it returns an errorCode (which can be tested using FSE_isError()).
430
size_t size = BIT_closeCStream(&bitStream);
431
*/
432
433
434
/* *****************************************
435
* FSE symbol decompression API
436
*******************************************/
437
typedef struct {
438
size_t state;
439
const void* table; /* precise table may vary, depending on U16 */
440
} FSE_DState_t;
441
442
443
static void FSE_initDState(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD, const FSE_DTable* dt);
444
445
static unsigned char FSE_decodeSymbol(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD);
446
447
static unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr);
448
449
/**<
450
Let's now decompose FSE_decompress_usingDTable() into its unitary components.
451
You will decode FSE-encoded symbols from the bitStream,
452
and also any other bitFields you put in, **in reverse order**.
453
454
You will need a few variables to track your bitStream. They are :
455
456
BIT_DStream_t DStream; // Stream context
457
FSE_DState_t DState; // State context. Multiple ones are possible
458
FSE_DTable* DTablePtr; // Decoding table, provided by FSE_buildDTable()
459
460
The first thing to do is to init the bitStream.
461
errorCode = BIT_initDStream(&DStream, srcBuffer, srcSize);
462
463
You should then retrieve your initial state(s)
464
(in reverse flushing order if you have several ones) :
465
errorCode = FSE_initDState(&DState, &DStream, DTablePtr);
466
467
You can then decode your data, symbol after symbol.
468
For information the maximum number of bits read by FSE_decodeSymbol() is 'tableLog'.
469
Keep in mind that symbols are decoded in reverse order, like a LIFO stack (last in, first out).
470
unsigned char symbol = FSE_decodeSymbol(&DState, &DStream);
471
472
You can retrieve any bitfield you eventually stored into the bitStream (in reverse order)
473
Note : maximum allowed nbBits is 25, for 32-bits compatibility
474
size_t bitField = BIT_readBits(&DStream, nbBits);
475
476
All above operations only read from local register (which size depends on size_t).
477
Refueling the register from memory is manually performed by the reload method.
478
endSignal = FSE_reloadDStream(&DStream);
479
480
BIT_reloadDStream() result tells if there is still some more data to read from DStream.
481
BIT_DStream_unfinished : there is still some data left into the DStream.
482
BIT_DStream_endOfBuffer : Dstream reached end of buffer. Its container may no longer be completely filled.
483
BIT_DStream_completed : Dstream reached its exact end, corresponding in general to decompression completed.
484
BIT_DStream_tooFar : Dstream went too far. Decompression result is corrupted.
485
486
When reaching end of buffer (BIT_DStream_endOfBuffer), progress slowly, notably if you decode multiple symbols per loop,
487
to properly detect the exact end of stream.
488
After each decoded symbol, check if DStream is fully consumed using this simple test :
489
BIT_reloadDStream(&DStream) >= BIT_DStream_completed
490
491
When it's done, verify decompression is fully completed, by checking both DStream and the relevant states.
492
Checking if DStream has reached its end is performed by :
493
BIT_endOfDStream(&DStream);
494
Check also the states. There might be some symbols left there, if some high probability ones (>50%) are possible.
495
FSE_endOfDState(&DState);
496
*/
497
498
499
/* *****************************************
500
* FSE unsafe API
501
*******************************************/
502
static unsigned char FSE_decodeSymbolFast(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD);
503
/* faster, but works only if nbBits is always >= 1 (otherwise, result will be corrupted) */
504
505
506
/* *****************************************
507
* Implementation of inlined functions
508
*******************************************/
509
typedef struct {
510
int deltaFindState;
511
U32 deltaNbBits;
512
} FSE_symbolCompressionTransform; /* total 8 bytes */
513
514
MEM_STATIC void FSE_initCState(FSE_CState_t* statePtr, const FSE_CTable* ct)
515
{
516
const void* ptr = ct;
517
const U16* u16ptr = (const U16*) ptr;
518
const U32 tableLog = MEM_read16(ptr);
519
statePtr->value = (ptrdiff_t)1<<tableLog;
520
statePtr->stateTable = u16ptr+2;
521
statePtr->symbolTT = ct + 1 + (tableLog ? (1<<(tableLog-1)) : 1);
522
statePtr->stateLog = tableLog;
523
}
524
525
526
/*! FSE_initCState2() :
527
* Same as FSE_initCState(), but the first symbol to include (which will be the last to be read)
528
* uses the smallest state value possible, saving the cost of this symbol */
529
MEM_STATIC void FSE_initCState2(FSE_CState_t* statePtr, const FSE_CTable* ct, U32 symbol)
530
{
531
FSE_initCState(statePtr, ct);
532
{ const FSE_symbolCompressionTransform symbolTT = ((const FSE_symbolCompressionTransform*)(statePtr->symbolTT))[symbol];
533
const U16* stateTable = (const U16*)(statePtr->stateTable);
534
U32 nbBitsOut = (U32)((symbolTT.deltaNbBits + (1<<15)) >> 16);
535
statePtr->value = (nbBitsOut << 16) - symbolTT.deltaNbBits;
536
statePtr->value = stateTable[(statePtr->value >> nbBitsOut) + symbolTT.deltaFindState];
537
}
538
}
539
540
MEM_STATIC void FSE_encodeSymbol(BIT_CStream_t* bitC, FSE_CState_t* statePtr, unsigned symbol)
541
{
542
FSE_symbolCompressionTransform const symbolTT = ((const FSE_symbolCompressionTransform*)(statePtr->symbolTT))[symbol];
543
const U16* const stateTable = (const U16*)(statePtr->stateTable);
544
U32 const nbBitsOut = (U32)((statePtr->value + symbolTT.deltaNbBits) >> 16);
545
BIT_addBits(bitC, statePtr->value, nbBitsOut);
546
statePtr->value = stateTable[ (statePtr->value >> nbBitsOut) + symbolTT.deltaFindState];
547
}
548
549
MEM_STATIC void FSE_flushCState(BIT_CStream_t* bitC, const FSE_CState_t* statePtr)
550
{
551
BIT_addBits(bitC, statePtr->value, statePtr->stateLog);
552
BIT_flushBits(bitC);
553
}
554
555
556
/* FSE_getMaxNbBits() :
557
* Approximate maximum cost of a symbol, in bits.
558
* Fractional get rounded up (i.e : a symbol with a normalized frequency of 3 gives the same result as a frequency of 2)
559
* note 1 : assume symbolValue is valid (<= maxSymbolValue)
560
* note 2 : if freq[symbolValue]==0, @return a fake cost of tableLog+1 bits */
561
MEM_STATIC U32 FSE_getMaxNbBits(const void* symbolTTPtr, U32 symbolValue)
562
{
563
const FSE_symbolCompressionTransform* symbolTT = (const FSE_symbolCompressionTransform*) symbolTTPtr;
564
return (symbolTT[symbolValue].deltaNbBits + ((1<<16)-1)) >> 16;
565
}
566
567
/* FSE_bitCost() :
568
* Approximate symbol cost, as fractional value, using fixed-point format (accuracyLog fractional bits)
569
* note 1 : assume symbolValue is valid (<= maxSymbolValue)
570
* note 2 : if freq[symbolValue]==0, @return a fake cost of tableLog+1 bits */
571
MEM_STATIC U32 FSE_bitCost(const void* symbolTTPtr, U32 tableLog, U32 symbolValue, U32 accuracyLog)
572
{
573
const FSE_symbolCompressionTransform* symbolTT = (const FSE_symbolCompressionTransform*) symbolTTPtr;
574
U32 const minNbBits = symbolTT[symbolValue].deltaNbBits >> 16;
575
U32 const threshold = (minNbBits+1) << 16;
576
assert(tableLog < 16);
577
assert(accuracyLog < 31-tableLog); /* ensure enough room for renormalization double shift */
578
{ U32 const tableSize = 1 << tableLog;
579
U32 const deltaFromThreshold = threshold - (symbolTT[symbolValue].deltaNbBits + tableSize);
580
U32 const normalizedDeltaFromThreshold = (deltaFromThreshold << accuracyLog) >> tableLog; /* linear interpolation (very approximate) */
581
U32 const bitMultiplier = 1 << accuracyLog;
582
assert(symbolTT[symbolValue].deltaNbBits + tableSize <= threshold);
583
assert(normalizedDeltaFromThreshold <= bitMultiplier);
584
return (minNbBits+1)*bitMultiplier - normalizedDeltaFromThreshold;
585
}
586
}
587
588
589
/* ====== Decompression ====== */
590
591
typedef struct {
592
U16 tableLog;
593
U16 fastMode;
594
} FSE_DTableHeader; /* sizeof U32 */
595
596
typedef struct
597
{
598
unsigned short newState;
599
unsigned char symbol;
600
unsigned char nbBits;
601
} FSE_decode_t; /* size == U32 */
602
603
MEM_STATIC void FSE_initDState(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD, const FSE_DTable* dt)
604
{
605
const void* ptr = dt;
606
const FSE_DTableHeader* const DTableH = (const FSE_DTableHeader*)ptr;
607
DStatePtr->state = BIT_readBits(bitD, DTableH->tableLog);
608
BIT_reloadDStream(bitD);
609
DStatePtr->table = dt + 1;
610
}
611
612
MEM_STATIC BYTE FSE_peekSymbol(const FSE_DState_t* DStatePtr)
613
{
614
FSE_decode_t const DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
615
return DInfo.symbol;
616
}
617
618
MEM_STATIC void FSE_updateState(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD)
619
{
620
FSE_decode_t const DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
621
U32 const nbBits = DInfo.nbBits;
622
size_t const lowBits = BIT_readBits(bitD, nbBits);
623
DStatePtr->state = DInfo.newState + lowBits;
624
}
625
626
MEM_STATIC BYTE FSE_decodeSymbol(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD)
627
{
628
FSE_decode_t const DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
629
U32 const nbBits = DInfo.nbBits;
630
BYTE const symbol = DInfo.symbol;
631
size_t const lowBits = BIT_readBits(bitD, nbBits);
632
633
DStatePtr->state = DInfo.newState + lowBits;
634
return symbol;
635
}
636
637
/*! FSE_decodeSymbolFast() :
638
unsafe, only works if no symbol has a probability > 50% */
639
MEM_STATIC BYTE FSE_decodeSymbolFast(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD)
640
{
641
FSE_decode_t const DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
642
U32 const nbBits = DInfo.nbBits;
643
BYTE const symbol = DInfo.symbol;
644
size_t const lowBits = BIT_readBitsFast(bitD, nbBits);
645
646
DStatePtr->state = DInfo.newState + lowBits;
647
return symbol;
648
}
649
650
MEM_STATIC unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr)
651
{
652
return DStatePtr->state == 0;
653
}
654
655
656
657
#ifndef FSE_COMMONDEFS_ONLY
658
659
/* **************************************************************
660
* Tuning parameters
661
****************************************************************/
662
/*!MEMORY_USAGE :
663
* Memory usage formula : N->2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; etc.)
664
* Increasing memory usage improves compression ratio
665
* Reduced memory usage can improve speed, due to cache effect
666
* Recommended max value is 14, for 16KB, which nicely fits into Intel x86 L1 cache */
667
#ifndef FSE_MAX_MEMORY_USAGE
668
# define FSE_MAX_MEMORY_USAGE 14
669
#endif
670
#ifndef FSE_DEFAULT_MEMORY_USAGE
671
# define FSE_DEFAULT_MEMORY_USAGE 13
672
#endif
673
#if (FSE_DEFAULT_MEMORY_USAGE > FSE_MAX_MEMORY_USAGE)
674
# error "FSE_DEFAULT_MEMORY_USAGE must be <= FSE_MAX_MEMORY_USAGE"
675
#endif
676
677
/*!FSE_MAX_SYMBOL_VALUE :
678
* Maximum symbol value authorized.
679
* Required for proper stack allocation */
680
#ifndef FSE_MAX_SYMBOL_VALUE
681
# define FSE_MAX_SYMBOL_VALUE 255
682
#endif
683
684
/* **************************************************************
685
* template functions type & suffix
686
****************************************************************/
687
#define FSE_FUNCTION_TYPE BYTE
688
#define FSE_FUNCTION_EXTENSION
689
#define FSE_DECODE_TYPE FSE_decode_t
690
691
692
#endif /* !FSE_COMMONDEFS_ONLY */
693
694
695
/* ***************************************************************
696
* Constants
697
*****************************************************************/
698
#define FSE_MAX_TABLELOG (FSE_MAX_MEMORY_USAGE-2)
699
#define FSE_MAX_TABLESIZE (1U<<FSE_MAX_TABLELOG)
700
#define FSE_MAXTABLESIZE_MASK (FSE_MAX_TABLESIZE-1)
701
#define FSE_DEFAULT_TABLELOG (FSE_DEFAULT_MEMORY_USAGE-2)
702
#define FSE_MIN_TABLELOG 5
703
704
#define FSE_TABLELOG_ABSOLUTE_MAX 15
705
#if FSE_MAX_TABLELOG > FSE_TABLELOG_ABSOLUTE_MAX
706
# error "FSE_MAX_TABLELOG > FSE_TABLELOG_ABSOLUTE_MAX is not supported"
707
#endif
708
709
#define FSE_TABLESTEP(tableSize) (((tableSize)>>1) + ((tableSize)>>3) + 3)
710
711
712
#endif /* FSE_STATIC_LINKING_ONLY */
713
714
715
#if defined (__cplusplus)
716
}
717
#endif
718
719