extzstd 0.0.2.CONCEPT → 0.0.3.CONCEPT

Sign up to get free protection for your applications and to get access to all the features.
Files changed (68) hide show
  1. checksums.yaml +4 -4
  2. data/README.md +72 -22
  3. data/Rakefile +73 -19
  4. data/contrib/zstd/README.md +68 -0
  5. data/contrib/zstd/common/bitstream.h +414 -0
  6. data/contrib/zstd/common/entropy_common.c +231 -0
  7. data/contrib/zstd/common/error_private.h +125 -0
  8. data/contrib/zstd/common/error_public.h +77 -0
  9. data/contrib/zstd/common/fse.h +628 -0
  10. data/contrib/zstd/common/fse_decompress.c +331 -0
  11. data/contrib/zstd/common/huf.h +228 -0
  12. data/contrib/zstd/common/mem.h +377 -0
  13. data/contrib/zstd/common/xxhash.c +854 -0
  14. data/contrib/zstd/common/xxhash.h +273 -0
  15. data/contrib/zstd/common/zbuff.h +197 -0
  16. data/contrib/zstd/common/zstd.h +475 -0
  17. data/contrib/zstd/common/zstd_common.c +91 -0
  18. data/contrib/zstd/common/zstd_internal.h +238 -0
  19. data/contrib/zstd/compress/fse_compress.c +807 -0
  20. data/contrib/zstd/compress/huf_compress.c +577 -0
  21. data/contrib/zstd/compress/zbuff_compress.c +327 -0
  22. data/contrib/zstd/compress/zstd_compress.c +3074 -0
  23. data/contrib/zstd/compress/zstd_opt.h +1046 -0
  24. data/contrib/zstd/decompress/huf_decompress.c +894 -0
  25. data/contrib/zstd/decompress/zbuff_decompress.c +294 -0
  26. data/contrib/zstd/decompress/zstd_decompress.c +1362 -0
  27. data/contrib/zstd/dictBuilder/divsufsort.c +1913 -0
  28. data/contrib/zstd/dictBuilder/divsufsort.h +67 -0
  29. data/contrib/zstd/dictBuilder/zdict.c +1045 -0
  30. data/contrib/zstd/dictBuilder/zdict.h +113 -0
  31. data/contrib/zstd/legacy/zstd_legacy.h +140 -0
  32. data/contrib/zstd/legacy/zstd_v01.c +2178 -0
  33. data/contrib/zstd/{zstd.h → legacy/zstd_v01.h} +46 -39
  34. data/contrib/zstd/legacy/zstd_v02.c +3748 -0
  35. data/contrib/zstd/legacy/zstd_v02.h +99 -0
  36. data/contrib/zstd/legacy/zstd_v03.c +3389 -0
  37. data/contrib/zstd/legacy/zstd_v03.h +99 -0
  38. data/contrib/zstd/legacy/zstd_v04.c +4056 -0
  39. data/contrib/zstd/legacy/zstd_v04.h +148 -0
  40. data/contrib/zstd/legacy/zstd_v05.c +4325 -0
  41. data/contrib/zstd/legacy/zstd_v05.h +171 -0
  42. data/contrib/zstd/legacy/zstd_v06.c +4581 -0
  43. data/contrib/zstd/legacy/zstd_v06.h +185 -0
  44. data/ext/extconf.rb +10 -12
  45. data/ext/extzstd.c +497 -144
  46. data/ext/extzstd.h +127 -22
  47. data/ext/extzstd_buffered.c +265 -0
  48. data/ext/extzstd_nogvls.h +174 -0
  49. data/ext/zstd_amalgam.c +18 -0
  50. data/ext/zstd_legacy_v01.c +1 -0
  51. data/ext/zstd_legacy_v02.c +1 -0
  52. data/ext/zstd_legacy_v03.c +1 -0
  53. data/ext/zstd_legacy_v04.c +1 -0
  54. data/ext/zstd_legacy_v05.c +1 -0
  55. data/ext/zstd_legacy_v06.c +1 -0
  56. data/gemstub.rb +17 -1
  57. data/lib/extzstd.rb +197 -77
  58. data/lib/extzstd/version.rb +1 -1
  59. data/test/test_basic.rb +41 -0
  60. metadata +70 -20
  61. data/contrib/zstd/Makefile +0 -115
  62. data/contrib/zstd/fse.c +0 -2466
  63. data/contrib/zstd/fse.h +0 -320
  64. data/contrib/zstd/fse_static.h +0 -282
  65. data/contrib/zstd/libzstd.pc.in +0 -14
  66. data/contrib/zstd/zstd.c +0 -1768
  67. data/contrib/zstd/zstd_static.h +0 -89
  68. data/ext/extzstd-stream.c +0 -398
@@ -1,14 +0,0 @@
1
- # ZSTD - standard compression algorithm
2
- # Copyright (C) 2014-2015, Yann Collet.
3
- # BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
4
-
5
- prefix=@PREFIX@
6
- libdir=@LIBDIR@
7
- includedir=@INCLUDEDIR@
8
-
9
- Name: zstd
10
- Description: lossless compression algorithm library
11
- URL: https://github.com/Cyan4973/zstd
12
- Version: @VERSION@
13
- Libs: -L@LIBDIR@ -lzstd
14
- Cflags: -I@INCLUDEDIR@
@@ -1,1768 +0,0 @@
1
- /*
2
- zstd - standard compression library
3
- Copyright (C) 2014-2015, Yann Collet.
4
-
5
- BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
6
-
7
- Redistribution and use in source and binary forms, with or without
8
- modification, are permitted provided that the following conditions are
9
- met:
10
- * Redistributions of source code must retain the above copyright
11
- notice, this list of conditions and the following disclaimer.
12
- * Redistributions in binary form must reproduce the above
13
- copyright notice, this list of conditions and the following disclaimer
14
- in the documentation and/or other materials provided with the
15
- distribution.
16
- THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
17
- "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
18
- LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
19
- A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
20
- OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
21
- SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
22
- LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23
- DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24
- THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25
- (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
26
- OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27
-
28
- You can contact the author at :
29
- - zstd source repository : https://github.com/Cyan4973/zstd
30
- - ztsd public forum : https://groups.google.com/forum/#!forum/lz4c
31
- */
32
-
33
- /****************************************************************
34
- * Tuning parameters
35
- *****************************************************************/
36
- /* MEMORY_USAGE :
37
- * Memory usage formula : N->2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; etc.)
38
- * Increasing memory usage improves compression ratio
39
- * Reduced memory usage can improve speed, due to cache effect */
40
- #define ZSTD_MEMORY_USAGE 17
41
-
42
-
43
- /**************************************
44
- CPU Feature Detection
45
- **************************************/
46
- /*
47
- * Automated efficient unaligned memory access detection
48
- * Based on known hardware architectures
49
- * This list will be updated thanks to feedbacks
50
- */
51
- #if defined(CPU_HAS_EFFICIENT_UNALIGNED_MEMORY_ACCESS) \
52
- || defined(__ARM_FEATURE_UNALIGNED) \
53
- || defined(__i386__) || defined(__x86_64__) \
54
- || defined(_M_IX86) || defined(_M_X64) \
55
- || defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_8__) \
56
- || (defined(_M_ARM) && (_M_ARM >= 7))
57
- # define ZSTD_UNALIGNED_ACCESS 1
58
- #else
59
- # define ZSTD_UNALIGNED_ACCESS 0
60
- #endif
61
-
62
-
63
- /********************************************************
64
- * Includes
65
- *********************************************************/
66
- #include <stdlib.h> /* calloc */
67
- #include <string.h> /* memcpy, memmove */
68
- #include <stdio.h> /* debug : printf */
69
- #include "zstd_static.h"
70
- #if defined(__clang__) || defined(__GNUC__)
71
- # include "fse.c" /* due to GCC/Clang inlining limitations, including *.c runs noticeably faster */
72
- #else
73
- # include "fse_static.h"
74
- #endif
75
-
76
-
77
- /********************************************************
78
- * Compiler specifics
79
- *********************************************************/
80
- #ifdef __AVX2__
81
- # include <immintrin.h> /* AVX2 intrinsics */
82
- #endif
83
-
84
- #ifdef _MSC_VER /* Visual Studio */
85
- # define FORCE_INLINE static __forceinline
86
- # include <intrin.h> /* For Visual 2005 */
87
- # pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */
88
- # pragma warning(disable : 4324) /* disable: C4324: padded structure */
89
- #else
90
- # define GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
91
- # ifdef __GNUC__
92
- # define FORCE_INLINE static inline __attribute__((always_inline))
93
- # else
94
- # define FORCE_INLINE static inline
95
- # endif
96
- #endif
97
-
98
-
99
- #ifndef MEM_ACCESS_MODULE
100
- #define MEM_ACCESS_MODULE
101
- /********************************************************
102
- * Basic Types
103
- *********************************************************/
104
- #if defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L /* C99 */
105
- # include <stdint.h>
106
- typedef uint8_t BYTE;
107
- typedef uint16_t U16;
108
- typedef int16_t S16;
109
- typedef uint32_t U32;
110
- typedef int32_t S32;
111
- typedef uint64_t U64;
112
- #else
113
- typedef unsigned char BYTE;
114
- typedef unsigned short U16;
115
- typedef signed short S16;
116
- typedef unsigned int U32;
117
- typedef signed int S32;
118
- typedef unsigned long long U64;
119
- #endif
120
-
121
- #endif /* MEM_ACCESS_MODULE */
122
-
123
-
124
- /********************************************************
125
- * Constants
126
- *********************************************************/
127
- static const U32 ZSTD_magicNumber = 0xFD2FB51E; /* 3rd version : seqNb header */
128
-
129
- #define HASH_LOG (ZSTD_MEMORY_USAGE - 2)
130
- #define HASH_TABLESIZE (1 << HASH_LOG)
131
- #define HASH_MASK (HASH_TABLESIZE - 1)
132
-
133
- #define KNUTH 2654435761
134
-
135
- #define BIT7 128
136
- #define BIT6 64
137
- #define BIT5 32
138
- #define BIT4 16
139
-
140
- #define KB *(1 <<10)
141
- #define MB *(1 <<20)
142
- #define GB *(1U<<30)
143
-
144
- #define BLOCKSIZE (128 KB) /* define, for static allocation */
145
- static const U32 g_maxDistance = 4 * BLOCKSIZE;
146
- static const U32 g_maxLimit = 1 GB;
147
- static const U32 g_searchStrength = 8;
148
-
149
- #define WORKPLACESIZE (BLOCKSIZE*3)
150
- #define MINMATCH 4
151
- #define MLbits 7
152
- #define LLbits 6
153
- #define Offbits 5
154
- #define MaxML ((1<<MLbits )-1)
155
- #define MaxLL ((1<<LLbits )-1)
156
- #define MaxOff ((1<<Offbits)-1)
157
- #define LitFSELog 11
158
- #define MLFSELog 10
159
- #define LLFSELog 10
160
- #define OffFSELog 9
161
- #define MAX(a,b) ((a)<(b)?(b):(a))
162
- #define MaxSeq MAX(MaxLL, MaxML)
163
-
164
- #define LITERAL_NOENTROPY 63
165
- #define COMMAND_NOENTROPY 7 /* to remove */
166
-
167
- static const size_t ZSTD_blockHeaderSize = 3;
168
- static const size_t ZSTD_frameHeaderSize = 4;
169
-
170
-
171
- /********************************************************
172
- * Memory operations
173
- *********************************************************/
174
- static unsigned ZSTD_32bits(void) { return sizeof(void*)==4; }
175
- static unsigned ZSTD_64bits(void) { return sizeof(void*)==8; }
176
-
177
- static unsigned ZSTD_isLittleEndian(void)
178
- {
179
- const union { U32 i; BYTE c[4]; } one = { 1 }; /* don't use static : performance detrimental */
180
- return one.c[0];
181
- }
182
-
183
- static U16 ZSTD_read16(const void* p) { U16 r; memcpy(&r, p, sizeof(r)); return r; }
184
-
185
- static U32 ZSTD_read32(const void* p) { U32 r; memcpy(&r, p, sizeof(r)); return r; }
186
-
187
- static U64 ZSTD_read64(const void* p) { U64 r; memcpy(&r, p, sizeof(r)); return r; }
188
-
189
- static size_t ZSTD_read_ARCH(const void* p) { size_t r; memcpy(&r, p, sizeof(r)); return r; }
190
-
191
- static void ZSTD_copy4(void* dst, const void* src) { memcpy(dst, src, 4); }
192
-
193
- static void ZSTD_copy8(void* dst, const void* src) { memcpy(dst, src, 8); }
194
-
195
- #define COPY8(d,s) { ZSTD_copy8(d,s); d+=8; s+=8; }
196
-
197
- static void ZSTD_wildcopy(void* dst, const void* src, size_t length)
198
- {
199
- const BYTE* ip = (const BYTE*)src;
200
- BYTE* op = (BYTE*)dst;
201
- BYTE* const oend = op + length;
202
- while (op < oend) COPY8(op, ip);
203
- }
204
-
205
- static U16 ZSTD_readLE16(const void* memPtr)
206
- {
207
- if (ZSTD_isLittleEndian()) return ZSTD_read16(memPtr);
208
- else
209
- {
210
- const BYTE* p = (const BYTE*)memPtr;
211
- return (U16)((U16)p[0] + ((U16)p[1]<<8));
212
- }
213
- }
214
-
215
- static void ZSTD_writeLE16(void* memPtr, U16 val)
216
- {
217
- if (ZSTD_isLittleEndian()) memcpy(memPtr, &val, sizeof(val));
218
- else
219
- {
220
- BYTE* p = (BYTE*)memPtr;
221
- p[0] = (BYTE)val;
222
- p[1] = (BYTE)(val>>8);
223
- }
224
- }
225
-
226
- static U32 ZSTD_readLE32(const void* memPtr)
227
- {
228
- if (ZSTD_isLittleEndian())
229
- return ZSTD_read32(memPtr);
230
- else
231
- {
232
- const BYTE* p = (const BYTE*)memPtr;
233
- return (U32)((U32)p[0] + ((U32)p[1]<<8) + ((U32)p[2]<<16) + ((U32)p[3]<<24));
234
- }
235
- }
236
-
237
- static void ZSTD_writeLE32(void* memPtr, U32 val32)
238
- {
239
- if (ZSTD_isLittleEndian())
240
- {
241
- memcpy(memPtr, &val32, 4);
242
- }
243
- else
244
- {
245
- BYTE* p = (BYTE*)memPtr;
246
- p[0] = (BYTE)val32;
247
- p[1] = (BYTE)(val32>>8);
248
- p[2] = (BYTE)(val32>>16);
249
- p[3] = (BYTE)(val32>>24);
250
- }
251
- }
252
-
253
- static U32 ZSTD_readBE32(const void* memPtr)
254
- {
255
- const BYTE* p = (const BYTE*)memPtr;
256
- return (U32)(((U32)p[0]<<24) + ((U32)p[1]<<16) + ((U32)p[2]<<8) + ((U32)p[3]<<0));
257
- }
258
-
259
- static void ZSTD_writeBE32(void* memPtr, U32 value)
260
- {
261
- BYTE* const p = (BYTE* const) memPtr;
262
- p[0] = (BYTE)(value>>24);
263
- p[1] = (BYTE)(value>>16);
264
- p[2] = (BYTE)(value>>8);
265
- p[3] = (BYTE)(value>>0);
266
- }
267
-
268
-
269
- /**************************************
270
- * Local structures
271
- ***************************************/
272
- typedef enum { bt_compressed, bt_raw, bt_rle, bt_end } blockType_t;
273
-
274
- typedef struct
275
- {
276
- blockType_t blockType;
277
- U32 origSize;
278
- } blockProperties_t;
279
-
280
- typedef struct {
281
- void* buffer;
282
- U32* offsetStart;
283
- U32* offset;
284
- BYTE* offCodeStart;
285
- BYTE* offCode;
286
- BYTE* litStart;
287
- BYTE* lit;
288
- BYTE* litLengthStart;
289
- BYTE* litLength;
290
- BYTE* matchLengthStart;
291
- BYTE* matchLength;
292
- BYTE* dumpsStart;
293
- BYTE* dumps;
294
- } seqStore_t;
295
-
296
- void ZSTD_resetSeqStore(seqStore_t* ssPtr)
297
- {
298
- ssPtr->offset = ssPtr->offsetStart;
299
- ssPtr->lit = ssPtr->litStart;
300
- ssPtr->litLength = ssPtr->litLengthStart;
301
- ssPtr->matchLength = ssPtr->matchLengthStart;
302
- ssPtr->dumps = ssPtr->dumpsStart;
303
- }
304
-
305
-
306
- typedef struct ZSTD_Cctx_s
307
- {
308
- const BYTE* base;
309
- U32 current;
310
- U32 nextUpdate;
311
- seqStore_t seqStore;
312
- #ifdef __AVX2__
313
- __m256i hashTable[HASH_TABLESIZE>>3];
314
- #else
315
- U32 hashTable[HASH_TABLESIZE];
316
- #endif
317
- } cctxi_t;
318
-
319
-
320
- ZSTD_Cctx* ZSTD_createCCtx(void)
321
- {
322
- ZSTD_Cctx* ctx = (ZSTD_Cctx*) malloc( sizeof(ZSTD_Cctx) );
323
- if (ctx==NULL) return NULL;
324
- ctx->seqStore.buffer = malloc(WORKPLACESIZE);
325
- if (ctx->seqStore.buffer==NULL)
326
- {
327
- free(ctx);
328
- return NULL;
329
- }
330
- ctx->seqStore.offsetStart = (U32*) (ctx->seqStore.buffer);
331
- ctx->seqStore.offCodeStart = (BYTE*) (ctx->seqStore.offsetStart + (BLOCKSIZE>>2));
332
- ctx->seqStore.litStart = ctx->seqStore.offCodeStart + (BLOCKSIZE>>2);
333
- ctx->seqStore.litLengthStart = ctx->seqStore.litStart + BLOCKSIZE;
334
- ctx->seqStore.matchLengthStart = ctx->seqStore.litLengthStart + (BLOCKSIZE>>2);
335
- ctx->seqStore.dumpsStart = ctx->seqStore.matchLengthStart + (BLOCKSIZE>>2);
336
- return ctx;
337
- }
338
-
339
- void ZSTD_resetCCtx(ZSTD_Cctx* ctx)
340
- {
341
- ctx->base = NULL;
342
- memset(ctx->hashTable, 0, HASH_TABLESIZE*4);
343
- }
344
-
345
- size_t ZSTD_freeCCtx(ZSTD_Cctx* ctx)
346
- {
347
- free(ctx->seqStore.buffer);
348
- free(ctx);
349
- return 0;
350
- }
351
-
352
-
353
- /**************************************
354
- * Error Management
355
- **************************************/
356
- /* tells if a return value is an error code */
357
- unsigned ZSTD_isError(size_t code)
358
- {
359
- return (code > (size_t)(-ZSTD_ERROR_maxCode));
360
- }
361
-
362
- #define ZSTD_GENERATE_STRING(STRING) #STRING,
363
- static const char* ZSTD_errorStrings[] = { ZSTD_LIST_ERRORS(ZSTD_GENERATE_STRING) };
364
-
365
- /* provides error code string (useful for debugging) */
366
- const char* ZSTD_getErrorName(size_t code)
367
- {
368
- static const char* codeError = "Unspecified error code";
369
- if (ZSTD_isError(code)) return ZSTD_errorStrings[-(int)(code)];
370
- return codeError;
371
- }
372
-
373
-
374
- /**************************************
375
- * Tool functions
376
- **************************************/
377
- unsigned ZSTD_versionNumber (void) { return ZSTD_VERSION_NUMBER; }
378
-
379
- static unsigned ZSTD_highbit(U32 val)
380
- {
381
- # if defined(_MSC_VER) /* Visual */
382
- unsigned long r;
383
- _BitScanReverse(&r, val);
384
- return (unsigned)r;
385
- # elif defined(__GNUC__) && (GCC_VERSION >= 304) /* GCC Intrinsic */
386
- return 31 - __builtin_clz(val);
387
- # else /* Software version */
388
- static const int DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 };
389
- U32 v = val;
390
- int r;
391
- v |= v >> 1;
392
- v |= v >> 2;
393
- v |= v >> 4;
394
- v |= v >> 8;
395
- v |= v >> 16;
396
- r = DeBruijnClz[(U32)(v * 0x07C4ACDDU) >> 27];
397
- return r;
398
- # endif
399
- }
400
-
401
- static unsigned ZSTD_NbCommonBytes (register size_t val)
402
- {
403
- if (ZSTD_isLittleEndian())
404
- {
405
- if (ZSTD_64bits())
406
- {
407
- # if defined(_MSC_VER) && defined(_WIN64) && !defined(LZ4_FORCE_SW_BITCOUNT)
408
- unsigned long r = 0;
409
- _BitScanForward64( &r, (U64)val );
410
- return (int)(r>>3);
411
- # elif defined(__GNUC__) && (GCC_VERSION >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT)
412
- return (__builtin_ctzll((U64)val) >> 3);
413
- # else
414
- static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2, 0, 3, 1, 3, 1, 4, 2, 7, 0, 2, 3, 6, 1, 5, 3, 5, 1, 3, 4, 4, 2, 5, 6, 7, 7, 0, 1, 2, 3, 3, 4, 6, 2, 6, 5, 5, 3, 4, 5, 6, 7, 1, 2, 4, 6, 4, 4, 5, 7, 2, 6, 5, 7, 6, 7, 7 };
415
- return DeBruijnBytePos[((U64)((val & -(long long)val) * 0x0218A392CDABBD3FULL)) >> 58];
416
- # endif
417
- }
418
- else /* 32 bits */
419
- {
420
- # if defined(_MSC_VER) && !defined(LZ4_FORCE_SW_BITCOUNT)
421
- unsigned long r;
422
- _BitScanForward( &r, (U32)val );
423
- return (int)(r>>3);
424
- # elif defined(__GNUC__) && (GCC_VERSION >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT)
425
- return (__builtin_ctz((U32)val) >> 3);
426
- # else
427
- static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0, 3, 2, 2, 1, 3, 2, 0, 1, 3, 3, 1, 2, 2, 2, 2, 0, 3, 1, 2, 0, 1, 0, 1, 1 };
428
- return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27];
429
- # endif
430
- }
431
- }
432
- else /* Big Endian CPU */
433
- {
434
- if (ZSTD_64bits())
435
- {
436
- # if defined(_MSC_VER) && defined(_WIN64) && !defined(LZ4_FORCE_SW_BITCOUNT)
437
- unsigned long r = 0;
438
- _BitScanReverse64( &r, val );
439
- return (unsigned)(r>>3);
440
- # elif defined(__GNUC__) && (GCC_VERSION >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT)
441
- return (__builtin_clzll(val) >> 3);
442
- # else
443
- unsigned r;
444
- const unsigned n32 = sizeof(size_t)*4; /* calculate this way due to compiler complaining in 32-bits mode */
445
- if (!(val>>n32)) { r=4; } else { r=0; val>>=n32; }
446
- if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; }
447
- r += (!val);
448
- return r;
449
- # endif
450
- }
451
- else /* 32 bits */
452
- {
453
- # if defined(_MSC_VER) && !defined(LZ4_FORCE_SW_BITCOUNT)
454
- unsigned long r = 0;
455
- _BitScanReverse( &r, (unsigned long)val );
456
- return (unsigned)(r>>3);
457
- # elif defined(__GNUC__) && (GCC_VERSION >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT)
458
- return (__builtin_clz((U32)val) >> 3);
459
- # else
460
- unsigned r;
461
- if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; }
462
- r += (!val);
463
- return r;
464
- # endif
465
- }
466
- }
467
- }
468
-
469
- static unsigned ZSTD_count(const BYTE* pIn, const BYTE* pMatch, const BYTE* pInLimit)
470
- {
471
- const BYTE* const pStart = pIn;
472
-
473
- while ((pIn<pInLimit-(sizeof(size_t)-1)))
474
- {
475
- size_t diff = ZSTD_read_ARCH(pMatch) ^ ZSTD_read_ARCH(pIn);
476
- if (!diff) { pIn+=sizeof(size_t); pMatch+=sizeof(size_t); continue; }
477
- pIn += ZSTD_NbCommonBytes(diff);
478
- return (unsigned)(pIn - pStart);
479
- }
480
-
481
- if (ZSTD_64bits()) if ((pIn<(pInLimit-3)) && (ZSTD_read32(pMatch) == ZSTD_read32(pIn))) { pIn+=4; pMatch+=4; }
482
- if ((pIn<(pInLimit-1)) && (ZSTD_read16(pMatch) == ZSTD_read16(pIn))) { pIn+=2; pMatch+=2; }
483
- if ((pIn<pInLimit) && (*pMatch == *pIn)) pIn++;
484
- return (unsigned)(pIn - pStart);
485
- }
486
-
487
-
488
- /********************************************************
489
- * Compression
490
- *********************************************************/
491
- size_t ZSTD_compressBound(size_t srcSize) /* maximum compressed size */
492
- {
493
- return FSE_compressBound(srcSize) + 12;
494
- }
495
-
496
-
497
- static size_t ZSTD_compressRle (void* dst, size_t maxDstSize, const void* src, size_t srcSize)
498
- {
499
- BYTE* const ostart = (BYTE* const)dst;
500
-
501
- /* at this stage : dstSize >= FSE_compressBound(srcSize) > (ZSTD_blockHeaderSize+1) (checked by ZSTD_compressLiterals()) */
502
- (void)maxDstSize;
503
-
504
- ostart[ZSTD_blockHeaderSize] = *(const BYTE*)src;
505
-
506
- /* Build header */
507
- ostart[0] = (BYTE)(srcSize>>16);
508
- ostart[1] = (BYTE)(srcSize>>8);
509
- ostart[2] = (BYTE) srcSize;
510
- ostart[0] += (BYTE)(bt_rle<<6);
511
-
512
- return ZSTD_blockHeaderSize+1;
513
- }
514
-
515
-
516
- static size_t ZSTD_noCompressBlock (void* dst, size_t maxDstSize, const void* src, size_t srcSize)
517
- {
518
- BYTE* const ostart = (BYTE* const)dst;
519
-
520
- if (srcSize + ZSTD_blockHeaderSize > maxDstSize) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall;
521
- memcpy(ostart + ZSTD_blockHeaderSize, src, srcSize);
522
-
523
- /* Build header */
524
- ostart[0] = (BYTE)(srcSize>>16);
525
- ostart[1] = (BYTE)(srcSize>>8);
526
- ostart[2] = (BYTE) srcSize;
527
- ostart[0] += (BYTE)(bt_raw<<6); /* is a raw (uncompressed) block */
528
-
529
- return ZSTD_blockHeaderSize+srcSize;
530
- }
531
-
532
-
533
- size_t ZSTD_minGain(size_t srcSize)
534
- {
535
- return (srcSize >> 6) + 1;
536
- }
537
-
538
-
539
- static size_t ZSTD_compressLiterals (void* dst, size_t dstSize,
540
- const void* src, size_t srcSize)
541
- {
542
- const size_t minGain = ZSTD_minGain(srcSize);
543
- BYTE* const ostart = (BYTE*)dst;
544
- size_t hsize;
545
- static const size_t LHSIZE = 5;
546
-
547
- if (dstSize < LHSIZE+1) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall; /* not enough space for compression */
548
-
549
- hsize = HUF_compress(ostart+LHSIZE, dstSize-LHSIZE, src, srcSize);
550
- if (hsize<2) return hsize; /* special cases */
551
- if (hsize >= srcSize - minGain) return 0;
552
-
553
- hsize += 2; /* work around vs fixed 3-bytes header */
554
-
555
- /* Build header */
556
- {
557
- ostart[0] = (BYTE)(bt_compressed<<6); /* is a block, is compressed */
558
- ostart[0] += (BYTE)(hsize>>16);
559
- ostart[1] = (BYTE)(hsize>>8);
560
- ostart[2] = (BYTE)(hsize>>0);
561
- ostart[0] += (BYTE)((srcSize>>16)<<3);
562
- ostart[3] = (BYTE)(srcSize>>8);
563
- ostart[4] = (BYTE)(srcSize>>0);
564
- }
565
-
566
- hsize -= 2;
567
- return hsize+LHSIZE;
568
- }
569
-
570
-
571
- static size_t ZSTD_compressSequences(BYTE* dst, size_t maxDstSize,
572
- const seqStore_t* seqStorePtr,
573
- size_t srcSize)
574
- {
575
- U32 count[MaxSeq+1];
576
- S16 norm[MaxSeq+1];
577
- size_t mostFrequent;
578
- U32 max = 255;
579
- U32 tableLog = 11;
580
- U32 CTable_LitLength [FSE_CTABLE_SIZE_U32(LLFSELog, MaxLL )];
581
- U32 CTable_OffsetBits [FSE_CTABLE_SIZE_U32(OffFSELog,MaxOff)];
582
- U32 CTable_MatchLength[FSE_CTABLE_SIZE_U32(MLFSELog, MaxML )];
583
- U32 LLtype, Offtype, MLtype; /* compressed, raw or rle */
584
- const BYTE* const op_lit_start = seqStorePtr->litStart;
585
- const BYTE* op_lit = seqStorePtr->lit;
586
- const BYTE* const llTable = seqStorePtr->litLengthStart;
587
- const BYTE* op_litLength = seqStorePtr->litLength;
588
- const BYTE* const mlTable = seqStorePtr->matchLengthStart;
589
- const U32* const offsetTable = seqStorePtr->offsetStart;
590
- BYTE* const offCodeTable = seqStorePtr->offCodeStart;
591
- BYTE* op = dst;
592
- BYTE* const oend = dst + maxDstSize;
593
- const size_t nbSeq = op_litLength - llTable;
594
- const size_t minGain = ZSTD_minGain(srcSize);
595
- const size_t maxCSize = srcSize - minGain;
596
- const size_t minSeqSize = 1 /*lastL*/ + 2 /*dHead*/ + 2 /*dumpsIn*/ + 5 /*SeqHead*/ + 3 /*SeqIn*/ + 1 /*margin*/ + ZSTD_blockHeaderSize;
597
- const size_t maxLSize = maxCSize > minSeqSize ? maxCSize - minSeqSize : 0;
598
- BYTE* seqHead;
599
-
600
-
601
- /* Compress literals */
602
- {
603
- size_t cSize;
604
- size_t litSize = op_lit - op_lit_start;
605
-
606
- if (litSize <= LITERAL_NOENTROPY) cSize = ZSTD_noCompressBlock (op, maxDstSize, op_lit_start, litSize);
607
- else
608
- {
609
- cSize = ZSTD_compressLiterals(op, maxDstSize, op_lit_start, litSize);
610
- if (cSize == 1) cSize = ZSTD_compressRle (op, maxDstSize, op_lit_start, litSize);
611
- else if (cSize == 0)
612
- {
613
- if (litSize >= maxLSize) return 0; /* block not compressible enough */
614
- cSize = ZSTD_noCompressBlock (op, maxDstSize, op_lit_start, litSize);
615
- }
616
- }
617
- if (ZSTD_isError(cSize)) return cSize;
618
- op += cSize;
619
- }
620
-
621
- /* Sequences Header */
622
- if ((oend-op) < 2+3+6) /* nbSeq + dumpsLength + 3*rleCTable*/
623
- return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall;
624
- ZSTD_writeLE16(op, (U16)nbSeq); op+=2;
625
- seqHead = op;
626
-
627
- /* dumps : contains too large lengths */
628
- {
629
- size_t dumpsLength = seqStorePtr->dumps - seqStorePtr->dumpsStart;
630
- if (dumpsLength < 512)
631
- {
632
- op[0] = (BYTE)(dumpsLength >> 8);
633
- op[1] = (BYTE)(dumpsLength);
634
- op += 2;
635
- }
636
- else
637
- {
638
- op[0] = 2;
639
- op[1] = (BYTE)(dumpsLength>>8);
640
- op[2] = (BYTE)(dumpsLength);
641
- op += 3;
642
- }
643
- if ((size_t)(oend-op) < dumpsLength+6) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall;
644
- memcpy(op, seqStorePtr->dumpsStart, dumpsLength);
645
- op += dumpsLength;
646
- }
647
-
648
- /* CTable for Literal Lengths */
649
- max = MaxLL;
650
- mostFrequent = FSE_countFast(count, &max, seqStorePtr->litLengthStart, nbSeq);
651
- if ((mostFrequent == nbSeq) && (nbSeq > 2))
652
- {
653
- *op++ = *(seqStorePtr->litLengthStart);
654
- FSE_buildCTable_rle(CTable_LitLength, (BYTE)max);
655
- LLtype = bt_rle;
656
- }
657
- else if ((nbSeq < 64) || (mostFrequent < (nbSeq >> (LLbits-1))))
658
- {
659
- FSE_buildCTable_raw(CTable_LitLength, LLbits);
660
- LLtype = bt_raw;
661
- }
662
- else
663
- {
664
- size_t NCountSize;
665
- tableLog = FSE_optimalTableLog(LLFSELog, nbSeq, max);
666
- FSE_normalizeCount(norm, tableLog, count, nbSeq, max);
667
- NCountSize = FSE_writeNCount(op, oend-op, norm, max, tableLog); /* overflow protected */
668
- if (FSE_isError(NCountSize)) return (size_t)-ZSTD_ERROR_GENERIC;
669
- op += NCountSize;
670
- FSE_buildCTable(CTable_LitLength, norm, max, tableLog);
671
- LLtype = bt_compressed;
672
- }
673
-
674
- /* CTable for Offsets codes */
675
- {
676
- /* create Offset codes */
677
- size_t i;
678
- max = MaxOff;
679
- for (i=0; i<nbSeq; i++)
680
- {
681
- offCodeTable[i] = (BYTE)ZSTD_highbit(offsetTable[i]) + 1;
682
- if (offsetTable[i]==0) offCodeTable[i]=0;
683
- }
684
- mostFrequent = FSE_countFast(count, &max, offCodeTable, nbSeq);
685
- }
686
- if ((mostFrequent == nbSeq) && (nbSeq > 2))
687
- {
688
- *op++ = *offCodeTable;
689
- FSE_buildCTable_rle(CTable_OffsetBits, (BYTE)max);
690
- Offtype = bt_rle;
691
- }
692
- else if ((nbSeq < 64) || (mostFrequent < (nbSeq >> (Offbits-1))))
693
- {
694
- FSE_buildCTable_raw(CTable_OffsetBits, Offbits);
695
- Offtype = bt_raw;
696
- }
697
- else
698
- {
699
- size_t NCountSize;
700
- tableLog = FSE_optimalTableLog(OffFSELog, nbSeq, max);
701
- FSE_normalizeCount(norm, tableLog, count, nbSeq, max);
702
- NCountSize = FSE_writeNCount(op, oend-op, norm, max, tableLog); /* overflow protected */
703
- if (FSE_isError(NCountSize)) return (size_t)-ZSTD_ERROR_GENERIC;
704
- op += NCountSize;
705
- FSE_buildCTable(CTable_OffsetBits, norm, max, tableLog);
706
- Offtype = bt_compressed;
707
- }
708
-
709
- /* CTable for MatchLengths */
710
- max = MaxML;
711
- mostFrequent = FSE_countFast(count, &max, seqStorePtr->matchLengthStart, nbSeq);
712
- if ((mostFrequent == nbSeq) && (nbSeq > 2))
713
- {
714
- *op++ = *seqStorePtr->matchLengthStart;
715
- FSE_buildCTable_rle(CTable_MatchLength, (BYTE)max);
716
- MLtype = bt_rle;
717
- }
718
- else if ((nbSeq < 64) || (mostFrequent < (nbSeq >> (MLbits-1))))
719
- {
720
- FSE_buildCTable_raw(CTable_MatchLength, MLbits);
721
- MLtype = bt_raw;
722
- }
723
- else
724
- {
725
- size_t NCountSize;
726
- tableLog = FSE_optimalTableLog(MLFSELog, nbSeq, max);
727
- FSE_normalizeCount(norm, tableLog, count, nbSeq, max);
728
- NCountSize = FSE_writeNCount(op, oend-op, norm, max, tableLog); /* overflow protected */
729
- if (FSE_isError(NCountSize)) return (size_t)-ZSTD_ERROR_GENERIC;
730
- op += NCountSize;
731
- FSE_buildCTable(CTable_MatchLength, norm, max, tableLog);
732
- MLtype = bt_compressed;
733
- }
734
-
735
- seqHead[0] += (BYTE)((LLtype<<6) + (Offtype<<4) + (MLtype<<2));
736
-
737
- /* Encoding Sequences */
738
- {
739
- size_t streamSize, errorCode;
740
- FSE_CStream_t blockStream;
741
- FSE_CState_t stateMatchLength;
742
- FSE_CState_t stateOffsetBits;
743
- FSE_CState_t stateLitLength;
744
- int i;
745
-
746
- errorCode = FSE_initCStream(&blockStream, op, oend-op);
747
- if (FSE_isError(errorCode)) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall; /* not enough space remaining */
748
- FSE_initCState(&stateMatchLength, CTable_MatchLength);
749
- FSE_initCState(&stateOffsetBits, CTable_OffsetBits);
750
- FSE_initCState(&stateLitLength, CTable_LitLength);
751
-
752
- for (i=(int)nbSeq-1; i>=0; i--)
753
- {
754
- BYTE matchLength = mlTable[i];
755
- U32 offset = offsetTable[i];
756
- BYTE offCode = offCodeTable[i]; /* 32b*/ /* 64b*/
757
- U32 nbBits = (offCode-1) * (!!offCode);
758
- BYTE litLength = llTable[i]; /* (7)*/ /* (7)*/
759
- FSE_encodeSymbol(&blockStream, &stateMatchLength, matchLength); /* 17 */ /* 17 */
760
- if (ZSTD_32bits()) FSE_flushBits(&blockStream); /* 7 */
761
- FSE_addBits(&blockStream, offset, nbBits); /* 32 */ /* 42 */
762
- if (ZSTD_32bits()) FSE_flushBits(&blockStream); /* 7 */
763
- FSE_encodeSymbol(&blockStream, &stateOffsetBits, offCode); /* 16 */ /* 51 */
764
- FSE_encodeSymbol(&blockStream, &stateLitLength, litLength); /* 26 */ /* 61 */
765
- FSE_flushBits(&blockStream); /* 7 */ /* 7 */
766
- }
767
-
768
- FSE_flushCState(&blockStream, &stateMatchLength);
769
- FSE_flushCState(&blockStream, &stateOffsetBits);
770
- FSE_flushCState(&blockStream, &stateLitLength);
771
-
772
- streamSize = FSE_closeCStream(&blockStream);
773
- if (streamSize==0) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall; /* not enough space */
774
- op += streamSize;
775
- }
776
-
777
- /* check compressibility */
778
- if ((size_t)(op-dst) >= maxCSize) return 0;
779
-
780
- return op - dst;
781
- }
782
-
783
-
784
- static void ZSTD_storeSeq(seqStore_t* seqStorePtr, size_t litLength, const BYTE* literals, size_t offset, size_t matchLength)
785
- {
786
- BYTE* op_lit = seqStorePtr->lit;
787
- BYTE* const l_end = op_lit + litLength;
788
-
789
- /* copy Literals */
790
- while (op_lit<l_end) COPY8(op_lit, literals);
791
- seqStorePtr->lit += litLength;
792
-
793
- /* literal Length */
794
- if (litLength >= MaxLL)
795
- {
796
- *(seqStorePtr->litLength++) = MaxLL;
797
- if (litLength<255 + MaxLL)
798
- *(seqStorePtr->dumps++) = (BYTE)(litLength - MaxLL);
799
- else
800
- {
801
- *(seqStorePtr->dumps++) = 255;
802
- ZSTD_writeLE32(seqStorePtr->dumps, (U32)litLength); seqStorePtr->dumps += 3;
803
- }
804
- }
805
- else *(seqStorePtr->litLength++) = (BYTE)litLength;
806
-
807
- /* match offset */
808
- *(seqStorePtr->offset++) = (U32)offset;
809
-
810
- /* match Length */
811
- if (matchLength >= MaxML)
812
- {
813
- *(seqStorePtr->matchLength++) = MaxML;
814
- if (matchLength < 255+MaxML)
815
- *(seqStorePtr->dumps++) = (BYTE)(matchLength - MaxML);
816
- else
817
- {
818
- *(seqStorePtr->dumps++) = 255;
819
- ZSTD_writeLE32(seqStorePtr->dumps, (U32)matchLength); seqStorePtr->dumps+=3;
820
- }
821
- }
822
- else *(seqStorePtr->matchLength++) = (BYTE)matchLength;
823
- }
824
-
825
-
826
- //static const U32 hashMask = (1<<HASH_LOG)-1;
827
- //static const U64 prime5bytes = 889523592379ULL;
828
- //static const U64 prime6bytes = 227718039650203ULL;
829
- static const U64 prime7bytes = 58295818150454627ULL;
830
- //static const U64 prime8bytes = 14923729446516375013ULL;
831
-
832
- //static U32 ZSTD_hashPtr(const void* p) { return (U32) _bextr_u64(*(U64*)p * prime7bytes, (56-HASH_LOG), HASH_LOG); }
833
- //static U32 ZSTD_hashPtr(const void* p) { return ( (*(U64*)p * prime7bytes) << 8 >> (64-HASH_LOG)); }
834
- //static U32 ZSTD_hashPtr(const void* p) { return ( (*(U64*)p * prime7bytes) >> (56-HASH_LOG)) & ((1<<HASH_LOG)-1); }
835
- //static U32 ZSTD_hashPtr(const void* p) { return ( ((*(U64*)p & 0xFFFFFFFFFFFFFF) * prime7bytes) >> (64-HASH_LOG)); }
836
-
837
- //static U32 ZSTD_hashPtr(const void* p) { return ( (*(U64*)p * prime8bytes) >> (64-HASH_LOG)); }
838
- static U32 ZSTD_hashPtr(const void* p) { return ( (ZSTD_read64(p) * prime7bytes) >> (56-HASH_LOG)) & HASH_MASK; }
839
- //static U32 ZSTD_hashPtr(const void* p) { return ( (*(U64*)p * prime6bytes) >> (48-HASH_LOG)) & HASH_MASK; }
840
- //static U32 ZSTD_hashPtr(const void* p) { return ( (*(U64*)p * prime5bytes) >> (40-HASH_LOG)) & HASH_MASK; }
841
- //static U32 ZSTD_hashPtr(const void* p) { return ( (*(U32*)p * KNUTH) >> (32-HASH_LOG)); }
842
-
843
- static void ZSTD_addPtr(U32* table, const BYTE* p, const BYTE* start) { table[ZSTD_hashPtr(p)] = (U32)(p-start); }
844
-
845
- static const BYTE* ZSTD_updateMatch(U32* table, const BYTE* p, const BYTE* start)
846
- {
847
- U32 h = ZSTD_hashPtr(p);
848
- const BYTE* r;
849
- r = table[h] + start;
850
- ZSTD_addPtr(table, p, start);
851
- return r;
852
- }
853
-
854
- static int ZSTD_checkMatch(const BYTE* match, const BYTE* ip)
855
- {
856
- return ZSTD_read32(match) == ZSTD_read32(ip);
857
- }
858
-
859
-
860
- static size_t ZSTD_compressBlock(void* cctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize)
861
- {
862
- cctxi_t* ctx = (cctxi_t*) cctx;
863
- U32* HashTable = (U32*)(ctx->hashTable);
864
- seqStore_t* seqStorePtr = &(ctx->seqStore);
865
- const BYTE* const base = ctx->base;
866
-
867
- const BYTE* const istart = (const BYTE*)src;
868
- const BYTE* ip = istart + 1;
869
- const BYTE* anchor = istart;
870
- const BYTE* const iend = istart + srcSize;
871
- const BYTE* const ilimit = iend - 16;
872
-
873
- size_t prevOffset=0, offset=0;
874
-
875
-
876
- /* init */
877
- ZSTD_resetSeqStore(seqStorePtr);
878
-
879
- /* Main Search Loop */
880
- while (ip < ilimit)
881
- {
882
- const BYTE* match = (const BYTE*) ZSTD_updateMatch(HashTable, ip, base);
883
-
884
- if (!ZSTD_checkMatch(match,ip)) { ip += ((ip-anchor) >> g_searchStrength) + 1; continue; }
885
-
886
- /* catch up */
887
- while ((ip>anchor) && (match>base) && (ip[-1] == match[-1])) { ip--; match--; }
888
-
889
- {
890
- size_t litLength = ip-anchor;
891
- size_t matchLength = ZSTD_count(ip+MINMATCH, match+MINMATCH, iend);
892
- size_t offsetCode;
893
- if (litLength) prevOffset = offset;
894
- offsetCode = ip-match;
895
- if (offsetCode == prevOffset) offsetCode = 0;
896
- prevOffset = offset;
897
- offset = ip-match;
898
- ZSTD_storeSeq(seqStorePtr, litLength, anchor, offsetCode, matchLength);
899
-
900
- /* Fill Table */
901
- ZSTD_addPtr(HashTable, ip+1, base);
902
- ip += matchLength + MINMATCH;
903
- if (ip<=iend-8) ZSTD_addPtr(HashTable, ip-2, base);
904
- anchor = ip;
905
- }
906
- }
907
-
908
- /* Last Literals */
909
- {
910
- size_t lastLLSize = iend - anchor;
911
- memcpy(seqStorePtr->lit, anchor, lastLLSize);
912
- seqStorePtr->lit += lastLLSize;
913
- }
914
-
915
- /* Finale compression stage */
916
- return ZSTD_compressSequences((BYTE*)dst, maxDstSize,
917
- seqStorePtr, srcSize);
918
- }
919
-
920
-
921
- size_t ZSTD_compressBegin(ZSTD_Cctx* ctx, void* dst, size_t maxDstSize)
922
- {
923
- /* Sanity check */
924
- if (maxDstSize < ZSTD_frameHeaderSize) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall;
925
-
926
- /* Init */
927
- ZSTD_resetCCtx(ctx);
928
-
929
- /* Write Header */
930
- ZSTD_writeBE32(dst, ZSTD_magicNumber);
931
-
932
- return ZSTD_frameHeaderSize;
933
- }
934
-
935
-
936
- static void ZSTD_scaleDownCtx(void* cctx, const U32 limit)
937
- {
938
- cctxi_t* ctx = (cctxi_t*) cctx;
939
- int i;
940
-
941
- #if defined(__AVX2__)
942
- /* AVX2 version */
943
- __m256i* h = ctx->hashTable;
944
- const __m256i limit8 = _mm256_set1_epi32(limit);
945
- for (i=0; i<(HASH_TABLESIZE>>3); i++)
946
- {
947
- __m256i src =_mm256_loadu_si256((const __m256i*)(h+i));
948
- const __m256i dec = _mm256_min_epu32(src, limit8);
949
- src = _mm256_sub_epi32(src, dec);
950
- _mm256_storeu_si256((__m256i*)(h+i), src);
951
- }
952
- #else
953
- /* this should be auto-vectorized by compiler */
954
- U32* h = ctx->hashTable;
955
- for (i=0; i<HASH_TABLESIZE; ++i)
956
- {
957
- U32 dec;
958
- if (h[i] > limit) dec = limit; else dec = h[i];
959
- h[i] -= dec;
960
- }
961
- #endif
962
- }
963
-
964
-
965
- static void ZSTD_limitCtx(void* cctx, const U32 limit)
966
- {
967
- cctxi_t* ctx = (cctxi_t*) cctx;
968
- int i;
969
-
970
- if (limit > g_maxLimit)
971
- {
972
- ZSTD_scaleDownCtx(cctx, limit);
973
- ctx->base += limit;
974
- ctx->current -= limit;
975
- ctx->nextUpdate -= limit;
976
- return;
977
- }
978
-
979
- #if defined(__AVX2__)
980
- /* AVX2 version */
981
- {
982
- __m256i* h = ctx->hashTable;
983
- const __m256i limit8 = _mm256_set1_epi32(limit);
984
- //printf("Address h : %0X\n", (U32)h); // address test
985
- for (i=0; i<(HASH_TABLESIZE>>3); i++)
986
- {
987
- __m256i src =_mm256_loadu_si256((const __m256i*)(h+i)); // Unfortunately, clang doesn't guarantee 32-bytes alignment
988
- src = _mm256_max_epu32(src, limit8);
989
- _mm256_storeu_si256((__m256i*)(h+i), src);
990
- }
991
- }
992
- #else
993
- /* this should be auto-vectorized by compiler */
994
- {
995
- U32* h = (U32*)(ctx->hashTable);
996
- for (i=0; i<HASH_TABLESIZE; ++i)
997
- {
998
- if (h[i] < limit) h[i] = limit;
999
- }
1000
- }
1001
- #endif
1002
- }
1003
-
1004
-
1005
- size_t ZSTD_compressContinue(ZSTD_Cctx* cctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize)
1006
- {
1007
- cctxi_t* ctx = (cctxi_t*) cctx;
1008
- const BYTE* const istart = (const BYTE* const)src;
1009
- const BYTE* ip = istart;
1010
- BYTE* const ostart = (BYTE* const)dst;
1011
- BYTE* op = ostart;
1012
- const U32 updateRate = 2 * BLOCKSIZE;
1013
-
1014
- /* Init */
1015
- if (ctx->base==NULL)
1016
- ctx->base = (const BYTE*)src, ctx->current=0, ctx->nextUpdate = g_maxDistance;
1017
- if (src != ctx->base + ctx->current) /* not contiguous */
1018
- {
1019
- ZSTD_resetCCtx(ctx);
1020
- ctx->base = (const BYTE*)src;
1021
- ctx->current = 0;
1022
- }
1023
- ctx->current += (U32)srcSize;
1024
-
1025
- while (srcSize)
1026
- {
1027
- size_t cSize;
1028
- size_t blockSize = BLOCKSIZE;
1029
- if (blockSize > srcSize) blockSize = srcSize;
1030
-
1031
- if (maxDstSize < 2*ZSTD_blockHeaderSize+1) /* one RLE block + endMark */
1032
- return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall;
1033
-
1034
- /* update hash table */
1035
- if (g_maxDistance <= BLOCKSIZE) /* static test ; yes == blocks are independent */
1036
- {
1037
- ZSTD_resetCCtx(ctx);
1038
- ctx->base = ip;
1039
- ctx->current=0;
1040
- }
1041
- else if (ip >= ctx->base + ctx->nextUpdate)
1042
- {
1043
- ctx->nextUpdate += updateRate;
1044
- ZSTD_limitCtx(ctx, ctx->nextUpdate - g_maxDistance);
1045
- }
1046
-
1047
- /* compress */
1048
- cSize = ZSTD_compressBlock(ctx, op+ZSTD_blockHeaderSize, maxDstSize-ZSTD_blockHeaderSize, ip, blockSize);
1049
- if (cSize == 0)
1050
- {
1051
- cSize = ZSTD_noCompressBlock(op, maxDstSize, ip, blockSize); /* block is not compressible */
1052
- if (ZSTD_isError(cSize)) return cSize;
1053
- }
1054
- else
1055
- {
1056
- if (ZSTD_isError(cSize)) return cSize;
1057
- op[0] = (BYTE)(cSize>>16);
1058
- op[1] = (BYTE)(cSize>>8);
1059
- op[2] = (BYTE)cSize;
1060
- op[0] += (BYTE)(bt_compressed << 6); /* is a compressed block */
1061
- cSize += 3;
1062
- }
1063
- op += cSize;
1064
- maxDstSize -= cSize;
1065
- ip += blockSize;
1066
- srcSize -= blockSize;
1067
- }
1068
-
1069
- return op-ostart;
1070
- }
1071
-
1072
-
1073
- size_t ZSTD_compressEnd(ZSTD_Cctx* ctx, void* dst, size_t maxDstSize)
1074
- {
1075
- BYTE* op = (BYTE*)dst;
1076
-
1077
- /* Sanity check */
1078
- (void)ctx;
1079
- if (maxDstSize < ZSTD_blockHeaderSize) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall;
1080
-
1081
- /* End of frame */
1082
- op[0] = (BYTE)(bt_end << 6);
1083
- op[1] = 0;
1084
- op[2] = 0;
1085
-
1086
- return 3;
1087
- }
1088
-
1089
-
1090
- static size_t ZSTD_compressCCtx(ZSTD_Cctx* ctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize)
1091
- {
1092
- BYTE* const ostart = (BYTE* const)dst;
1093
- BYTE* op = ostart;
1094
-
1095
- /* Header */
1096
- {
1097
- size_t headerSize = ZSTD_compressBegin(ctx, dst, maxDstSize);
1098
- if(ZSTD_isError(headerSize)) return headerSize;
1099
- op += headerSize;
1100
- maxDstSize -= headerSize;
1101
- }
1102
-
1103
- /* Compression */
1104
- {
1105
- size_t cSize = ZSTD_compressContinue(ctx, op, maxDstSize, src, srcSize);
1106
- if (ZSTD_isError(cSize)) return cSize;
1107
- op += cSize;
1108
- maxDstSize -= cSize;
1109
- }
1110
-
1111
- /* Close frame */
1112
- {
1113
- size_t endSize = ZSTD_compressEnd(ctx, op, maxDstSize);
1114
- if(ZSTD_isError(endSize)) return endSize;
1115
- op += endSize;
1116
- }
1117
-
1118
- return (op - ostart);
1119
- }
1120
-
1121
-
1122
- size_t ZSTD_compress(void* dst, size_t maxDstSize, const void* src, size_t srcSize)
1123
- {
1124
- ZSTD_Cctx* ctx;
1125
- size_t r;
1126
-
1127
- ctx = ZSTD_createCCtx();
1128
- if (ctx==NULL) return (size_t)-ZSTD_ERROR_GENERIC;
1129
- r = ZSTD_compressCCtx(ctx, dst, maxDstSize, src, srcSize);
1130
- ZSTD_freeCCtx(ctx);
1131
- return r;
1132
- }
1133
-
1134
-
1135
- /**************************************************************
1136
- * Decompression code
1137
- **************************************************************/
1138
-
1139
- size_t ZSTD_getcBlockSize(const void* src, size_t srcSize, blockProperties_t* bpPtr)
1140
- {
1141
- const BYTE* const in = (const BYTE* const)src;
1142
- BYTE headerFlags;
1143
- U32 cSize;
1144
-
1145
- if (srcSize < 3) return (size_t)-ZSTD_ERROR_SrcSize;
1146
-
1147
- headerFlags = *in;
1148
- cSize = in[2] + (in[1]<<8) + ((in[0] & 7)<<16);
1149
-
1150
- bpPtr->blockType = (blockType_t)(headerFlags >> 6);
1151
- bpPtr->origSize = (bpPtr->blockType == bt_rle) ? cSize : 0;
1152
-
1153
- if (bpPtr->blockType == bt_end) return 0;
1154
- if (bpPtr->blockType == bt_rle) return 1;
1155
- return cSize;
1156
- }
1157
-
1158
-
1159
- static size_t ZSTD_copyUncompressedBlock(void* dst, size_t maxDstSize, const void* src, size_t srcSize)
1160
- {
1161
- if (srcSize > maxDstSize) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall;
1162
- memcpy(dst, src, srcSize);
1163
- return srcSize;
1164
- }
1165
-
1166
-
1167
- static size_t ZSTD_decompressLiterals(void* ctx,
1168
- void* dst, size_t maxDstSize,
1169
- const void* src, size_t srcSize)
1170
- {
1171
- BYTE* op = (BYTE*)dst;
1172
- BYTE* const oend = op + maxDstSize;
1173
- const BYTE* ip = (const BYTE*)src;
1174
- size_t errorCode;
1175
- size_t litSize;
1176
-
1177
- /* check : minimum 2, for litSize, +1, for content */
1178
- if (srcSize <= 3) return (size_t)-ZSTD_ERROR_corruption;
1179
-
1180
- litSize = ip[1] + (ip[0]<<8);
1181
- litSize += ((ip[-3] >> 3) & 7) << 16; // mmmmh....
1182
- op = oend - litSize;
1183
-
1184
- (void)ctx;
1185
- if (litSize > maxDstSize) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall;
1186
- errorCode = HUF_decompress(op, litSize, ip+2, srcSize-2);
1187
- if (FSE_isError(errorCode)) return (size_t)-ZSTD_ERROR_GENERIC;
1188
- return litSize;
1189
- }
1190
-
1191
-
1192
- size_t ZSTD_decodeLiteralsBlock(void* ctx,
1193
- void* dst, size_t maxDstSize,
1194
- const BYTE** litStart, size_t* litSize,
1195
- const void* src, size_t srcSize)
1196
- {
1197
- const BYTE* const istart = (const BYTE* const)src;
1198
- const BYTE* ip = istart;
1199
- BYTE* const ostart = (BYTE* const)dst;
1200
- BYTE* const oend = ostart + maxDstSize;
1201
- blockProperties_t litbp;
1202
-
1203
- size_t litcSize = ZSTD_getcBlockSize(src, srcSize, &litbp);
1204
- if (ZSTD_isError(litcSize)) return litcSize;
1205
- if (litcSize > srcSize - ZSTD_blockHeaderSize) return (size_t)-ZSTD_ERROR_SrcSize;
1206
- ip += ZSTD_blockHeaderSize;
1207
-
1208
- switch(litbp.blockType)
1209
- {
1210
- case bt_raw:
1211
- *litStart = ip;
1212
- ip += litcSize;
1213
- *litSize = litcSize;
1214
- break;
1215
- case bt_rle:
1216
- {
1217
- size_t rleSize = litbp.origSize;
1218
- if (rleSize>maxDstSize) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall;
1219
- memset(oend - rleSize, *ip, rleSize);
1220
- *litStart = oend - rleSize;
1221
- *litSize = rleSize;
1222
- ip++;
1223
- break;
1224
- }
1225
- case bt_compressed:
1226
- {
1227
- size_t decodedLitSize = ZSTD_decompressLiterals(ctx, dst, maxDstSize, ip, litcSize);
1228
- if (ZSTD_isError(decodedLitSize)) return decodedLitSize;
1229
- *litStart = oend - decodedLitSize;
1230
- *litSize = decodedLitSize;
1231
- ip += litcSize;
1232
- break;
1233
- }
1234
- default:
1235
- return (size_t)-ZSTD_ERROR_GENERIC;
1236
- }
1237
-
1238
- return ip-istart;
1239
- }
1240
-
1241
-
1242
- size_t ZSTD_decodeSeqHeaders(int* nbSeq, const BYTE** dumpsPtr,
1243
- FSE_DTable* DTableLL, FSE_DTable* DTableML, FSE_DTable* DTableOffb,
1244
- const void* src, size_t srcSize)
1245
- {
1246
- const BYTE* const istart = (const BYTE* const)src;
1247
- const BYTE* ip = istart;
1248
- const BYTE* const iend = istart + srcSize;
1249
- U32 LLtype, Offtype, MLtype;
1250
- U32 LLlog, Offlog, MLlog;
1251
- size_t dumpsLength;
1252
-
1253
- /* check */
1254
- if (srcSize < 5) return (size_t)-ZSTD_ERROR_SrcSize;
1255
-
1256
- /* SeqHead */
1257
- *nbSeq = ZSTD_readLE16(ip); ip+=2;
1258
- LLtype = *ip >> 6;
1259
- Offtype = (*ip >> 4) & 3;
1260
- MLtype = (*ip >> 2) & 3;
1261
- if (*ip & 2)
1262
- {
1263
- dumpsLength = ip[2];
1264
- dumpsLength += ip[1] << 8;
1265
- ip += 3;
1266
- }
1267
- else
1268
- {
1269
- dumpsLength = ip[1];
1270
- dumpsLength += (ip[0] & 1) << 8;
1271
- ip += 2;
1272
- }
1273
- *dumpsPtr = ip;
1274
- ip += dumpsLength;
1275
-
1276
- /* check */
1277
- if (ip > iend-3) return (size_t)-ZSTD_ERROR_SrcSize; /* min : all 3 are "raw", hence no header, but at least xxLog bits per type */
1278
-
1279
- /* sequences */
1280
- {
1281
- S16 norm[MaxML+1]; /* assumption : MaxML >= MaxLL and MaxOff */
1282
- size_t headerSize;
1283
-
1284
- /* Build DTables */
1285
- switch(LLtype)
1286
- {
1287
- U32 max;
1288
- case bt_rle :
1289
- LLlog = 0;
1290
- FSE_buildDTable_rle(DTableLL, *ip++); break;
1291
- case bt_raw :
1292
- LLlog = LLbits;
1293
- FSE_buildDTable_raw(DTableLL, LLbits); break;
1294
- default :
1295
- max = MaxLL;
1296
- headerSize = FSE_readNCount(norm, &max, &LLlog, ip, iend-ip);
1297
- if (FSE_isError(headerSize)) return (size_t)-ZSTD_ERROR_GENERIC;
1298
- if (LLlog > LLFSELog) return (size_t)-ZSTD_ERROR_corruption;
1299
- ip += headerSize;
1300
- FSE_buildDTable(DTableLL, norm, max, LLlog);
1301
- }
1302
-
1303
- switch(Offtype)
1304
- {
1305
- U32 max;
1306
- case bt_rle :
1307
- Offlog = 0;
1308
- if (ip > iend-2) return (size_t)-ZSTD_ERROR_SrcSize; /* min : "raw", hence no header, but at least xxLog bits */
1309
- FSE_buildDTable_rle(DTableOffb, *ip++); break;
1310
- case bt_raw :
1311
- Offlog = Offbits;
1312
- FSE_buildDTable_raw(DTableOffb, Offbits); break;
1313
- default :
1314
- max = MaxOff;
1315
- headerSize = FSE_readNCount(norm, &max, &Offlog, ip, iend-ip);
1316
- if (FSE_isError(headerSize)) return (size_t)-ZSTD_ERROR_GENERIC;
1317
- if (Offlog > OffFSELog) return (size_t)-ZSTD_ERROR_corruption;
1318
- ip += headerSize;
1319
- FSE_buildDTable(DTableOffb, norm, max, Offlog);
1320
- }
1321
-
1322
- switch(MLtype)
1323
- {
1324
- U32 max;
1325
- case bt_rle :
1326
- MLlog = 0;
1327
- if (ip > iend-2) return (size_t)-ZSTD_ERROR_SrcSize; /* min : "raw", hence no header, but at least xxLog bits */
1328
- FSE_buildDTable_rle(DTableML, *ip++); break;
1329
- case bt_raw :
1330
- MLlog = MLbits;
1331
- FSE_buildDTable_raw(DTableML, MLbits); break;
1332
- default :
1333
- max = MaxML;
1334
- headerSize = FSE_readNCount(norm, &max, &MLlog, ip, iend-ip);
1335
- if (FSE_isError(headerSize)) return (size_t)-ZSTD_ERROR_GENERIC;
1336
- if (MLlog > MLFSELog) return (size_t)-ZSTD_ERROR_corruption;
1337
- ip += headerSize;
1338
- FSE_buildDTable(DTableML, norm, max, MLlog);
1339
- }
1340
- }
1341
-
1342
- return ip-istart;
1343
- }
1344
-
1345
-
1346
- typedef struct {
1347
- size_t litLength;
1348
- size_t offset;
1349
- size_t matchLength;
1350
- } seq_t;
1351
-
1352
- typedef struct {
1353
- FSE_DStream_t DStream;
1354
- FSE_DState_t stateLL;
1355
- FSE_DState_t stateOffb;
1356
- FSE_DState_t stateML;
1357
- size_t prevOffset;
1358
- const BYTE* dumps;
1359
- } seqState_t;
1360
-
1361
-
1362
- static void ZSTD_decodeSequence(seq_t* seq, seqState_t* seqState)
1363
- {
1364
- size_t litLength;
1365
- size_t prevOffset;
1366
- size_t offset;
1367
- size_t matchLength;
1368
- const BYTE* dumps = seqState->dumps;
1369
-
1370
- /* Literal length */
1371
- litLength = FSE_decodeSymbol(&(seqState->stateLL), &(seqState->DStream));
1372
- prevOffset = litLength ? seq->offset : seqState->prevOffset;
1373
- seqState->prevOffset = seq->offset;
1374
- if (litLength == MaxLL)
1375
- {
1376
- U32 add = *dumps++;
1377
- if (add < 255) litLength += add;
1378
- else
1379
- {
1380
- litLength = ZSTD_readLE32(dumps) & 0xFFFFFF;
1381
- dumps += 3;
1382
- }
1383
- }
1384
-
1385
- /* Offset */
1386
- {
1387
- U32 offsetCode, nbBits;
1388
- offsetCode = FSE_decodeSymbol(&(seqState->stateOffb), &(seqState->DStream));
1389
- if (ZSTD_32bits()) FSE_reloadDStream(&(seqState->DStream));
1390
- nbBits = offsetCode - 1;
1391
- if (offsetCode==0) nbBits = 0; /* cmove */
1392
- offset = ((size_t)1 << (nbBits & ((sizeof(offset)*8)-1))) + FSE_readBits(&(seqState->DStream), nbBits);
1393
- if (ZSTD_32bits()) FSE_reloadDStream(&(seqState->DStream));
1394
- if (offsetCode==0) offset = prevOffset;
1395
- }
1396
-
1397
- /* MatchLength */
1398
- matchLength = FSE_decodeSymbol(&(seqState->stateML), &(seqState->DStream));
1399
- if (matchLength == MaxML)
1400
- {
1401
- U32 add = *dumps++;
1402
- if (add < 255) matchLength += add;
1403
- else
1404
- {
1405
- matchLength = ZSTD_readLE32(dumps) & 0xFFFFFF; /* no pb : dumps is always followed by seq tables > 1 byte */
1406
- dumps += 3;
1407
- }
1408
- }
1409
- matchLength += MINMATCH;
1410
-
1411
- /* save result */
1412
- seq->litLength = litLength;
1413
- seq->offset = offset;
1414
- seq->matchLength = matchLength;
1415
- seqState->dumps = dumps;
1416
- }
1417
-
1418
-
1419
- static size_t ZSTD_execSequence(BYTE* op,
1420
- seq_t sequence,
1421
- const BYTE** litPtr, const BYTE* const litLimit,
1422
- BYTE* const base, BYTE* const oend)
1423
- {
1424
- static const int dec32table[] = {0, 1, 2, 1, 4, 4, 4, 4}; /* added */
1425
- static const int dec64table[] = {8, 8, 8, 7, 8, 9,10,11}; /* substracted */
1426
- const BYTE* const ostart = op;
1427
- const size_t litLength = sequence.litLength;
1428
- BYTE* const endMatch = op + litLength + sequence.matchLength; /* risk : address space overflow (32-bits) */
1429
- const BYTE* const litEnd = *litPtr + litLength;
1430
-
1431
- /* check */
1432
- if (endMatch > oend) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall; /* overwrite beyond dst buffer */
1433
- if (litEnd > litLimit) return (size_t)-ZSTD_ERROR_corruption;
1434
- if (sequence.matchLength > (size_t)(*litPtr-op)) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall; /* overwrite literal segment */
1435
-
1436
- /* copy Literals */
1437
- if (((size_t)(*litPtr - op) < 8) || ((size_t)(oend-litEnd) < 8) || (op+litLength > oend-8))
1438
- memmove(op, *litPtr, litLength); /* overwrite risk */
1439
- else
1440
- ZSTD_wildcopy(op, *litPtr, litLength);
1441
- op += litLength;
1442
- *litPtr = litEnd; /* update for next sequence */
1443
-
1444
- /* check : last match must be at a minimum distance of 8 from end of dest buffer */
1445
- if (oend-op < 8) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall;
1446
-
1447
- /* copy Match */
1448
- {
1449
- const U32 overlapRisk = (((size_t)(litEnd - endMatch)) < 12);
1450
- const BYTE* match = op - sequence.offset; /* possible underflow at op - offset ? */
1451
- size_t qutt = 12;
1452
- U64 saved[2];
1453
-
1454
- /* check */
1455
- if (match < base) return (size_t)-ZSTD_ERROR_corruption;
1456
- if (sequence.offset > (size_t)base) return (size_t)-ZSTD_ERROR_corruption;
1457
-
1458
- /* save beginning of literal sequence, in case of write overlap */
1459
- if (overlapRisk)
1460
- {
1461
- if ((endMatch + qutt) > oend) qutt = oend-endMatch;
1462
- memcpy(saved, endMatch, qutt);
1463
- }
1464
-
1465
- if (sequence.offset < 8)
1466
- {
1467
- const int dec64 = dec64table[sequence.offset];
1468
- op[0] = match[0];
1469
- op[1] = match[1];
1470
- op[2] = match[2];
1471
- op[3] = match[3];
1472
- match += dec32table[sequence.offset];
1473
- ZSTD_copy4(op+4, match);
1474
- match -= dec64;
1475
- } else { ZSTD_copy8(op, match); }
1476
- op += 8; match += 8;
1477
-
1478
- if (endMatch > oend-12)
1479
- {
1480
- if (op < oend-8)
1481
- {
1482
- ZSTD_wildcopy(op, match, (oend-8) - op);
1483
- match += (oend-8) - op;
1484
- op = oend-8;
1485
- }
1486
- while (op<endMatch) *op++ = *match++;
1487
- }
1488
- else
1489
- ZSTD_wildcopy(op, match, sequence.matchLength-8); /* works even if matchLength < 8 */
1490
-
1491
- /* restore, in case of overlap */
1492
- if (overlapRisk) memcpy(endMatch, saved, qutt);
1493
- }
1494
-
1495
- return endMatch-ostart;
1496
- }
1497
-
1498
- typedef struct ZSTD_Dctx_s
1499
- {
1500
- U32 LLTable[FSE_DTABLE_SIZE_U32(LLFSELog)];
1501
- U32 OffTable[FSE_DTABLE_SIZE_U32(OffFSELog)];
1502
- U32 MLTable[FSE_DTABLE_SIZE_U32(MLFSELog)];
1503
- void* previousDstEnd;
1504
- void* base;
1505
- size_t expected;
1506
- blockType_t bType;
1507
- U32 phase;
1508
- } dctx_t;
1509
-
1510
-
1511
- static size_t ZSTD_decompressSequences(
1512
- void* ctx,
1513
- void* dst, size_t maxDstSize,
1514
- const void* seqStart, size_t seqSize,
1515
- const BYTE* litStart, size_t litSize)
1516
- {
1517
- dctx_t* dctx = (dctx_t*)ctx;
1518
- const BYTE* ip = (const BYTE*)seqStart;
1519
- const BYTE* const iend = ip + seqSize;
1520
- BYTE* const ostart = (BYTE* const)dst;
1521
- BYTE* op = ostart;
1522
- BYTE* const oend = ostart + maxDstSize;
1523
- size_t errorCode;
1524
- const BYTE* litPtr = litStart;
1525
- const BYTE* const litEnd = litStart + litSize;
1526
- int nbSeq;
1527
- const BYTE* dumps;
1528
- U32* DTableLL = dctx->LLTable;
1529
- U32* DTableML = dctx->MLTable;
1530
- U32* DTableOffb = dctx->OffTable;
1531
- BYTE* const base = (BYTE*) (dctx->base);
1532
-
1533
- /* Build Decoding Tables */
1534
- errorCode = ZSTD_decodeSeqHeaders(&nbSeq, &dumps,
1535
- DTableLL, DTableML, DTableOffb,
1536
- ip, iend-ip);
1537
- if (ZSTD_isError(errorCode)) return errorCode;
1538
- ip += errorCode;
1539
-
1540
- /* Regen sequences */
1541
- {
1542
- seq_t sequence;
1543
- seqState_t seqState;
1544
-
1545
- memset(&sequence, 0, sizeof(sequence));
1546
- seqState.dumps = dumps;
1547
- seqState.prevOffset = 1;
1548
- errorCode = FSE_initDStream(&(seqState.DStream), ip, iend-ip);
1549
- if (FSE_isError(errorCode)) return (size_t)-ZSTD_ERROR_corruption;
1550
- FSE_initDState(&(seqState.stateLL), &(seqState.DStream), DTableLL);
1551
- FSE_initDState(&(seqState.stateOffb), &(seqState.DStream), DTableOffb);
1552
- FSE_initDState(&(seqState.stateML), &(seqState.DStream), DTableML);
1553
-
1554
- for ( ; (FSE_reloadDStream(&(seqState.DStream)) < FSE_DStream_completed) || (nbSeq>0) ; )
1555
- {
1556
- size_t oneSeqSize;
1557
- nbSeq--;
1558
- ZSTD_decodeSequence(&sequence, &seqState);
1559
- oneSeqSize = ZSTD_execSequence(op, sequence, &litPtr, litEnd, base, oend);
1560
- if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
1561
- op += oneSeqSize;
1562
- }
1563
-
1564
- /* check if reached exact end */
1565
- if (FSE_reloadDStream(&(seqState.DStream)) > FSE_DStream_completed) return (size_t)-ZSTD_ERROR_corruption; /* requested too much : data is corrupted */
1566
- if (nbSeq<0) return (size_t)-ZSTD_ERROR_corruption; /* requested too many sequences : data is corrupted */
1567
-
1568
- /* last literal segment */
1569
- {
1570
- size_t lastLLSize = litEnd - litPtr;
1571
- if (op+lastLLSize > oend) return (size_t)-ZSTD_ERROR_maxDstSize_tooSmall;
1572
- if (op != litPtr) memmove(op, litPtr, lastLLSize);
1573
- op += lastLLSize;
1574
- }
1575
- }
1576
-
1577
- return op-ostart;
1578
- }
1579
-
1580
-
1581
- static size_t ZSTD_decompressBlock(
1582
- void* ctx,
1583
- void* dst, size_t maxDstSize,
1584
- const void* src, size_t srcSize)
1585
- {
1586
- /* blockType == blockCompressed, srcSize is trusted */
1587
- const BYTE* ip = (const BYTE*)src;
1588
- const BYTE* litPtr;
1589
- size_t litSize;
1590
- size_t errorCode;
1591
-
1592
- /* Decode literals sub-block */
1593
- errorCode = ZSTD_decodeLiteralsBlock(ctx, dst, maxDstSize, &litPtr, &litSize, src, srcSize);
1594
- if (ZSTD_isError(errorCode)) return errorCode;
1595
- ip += errorCode;
1596
- srcSize -= errorCode;
1597
-
1598
- return ZSTD_decompressSequences(ctx, dst, maxDstSize, ip, srcSize, litPtr, litSize);
1599
- }
1600
-
1601
-
1602
- static size_t ZSTD_decompressDCtx(void* ctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize)
1603
- {
1604
- const BYTE* ip = (const BYTE*)src;
1605
- const BYTE* iend = ip + srcSize;
1606
- BYTE* const ostart = (BYTE* const)dst;
1607
- BYTE* op = ostart;
1608
- BYTE* const oend = ostart + maxDstSize;
1609
- size_t remainingSize = srcSize;
1610
- U32 magicNumber;
1611
- size_t errorCode=0;
1612
- blockProperties_t blockProperties;
1613
-
1614
- /* Frame Header */
1615
- if (srcSize < ZSTD_frameHeaderSize+ZSTD_blockHeaderSize) return (size_t)-ZSTD_ERROR_SrcSize;
1616
- magicNumber = ZSTD_readBE32(src);
1617
- if (magicNumber != ZSTD_magicNumber) return (size_t)-ZSTD_ERROR_MagicNumber;
1618
- ip += ZSTD_frameHeaderSize; remainingSize -= ZSTD_frameHeaderSize;
1619
-
1620
- /* Loop on each block */
1621
- while (1)
1622
- {
1623
- size_t blockSize = ZSTD_getcBlockSize(ip, iend-ip, &blockProperties);
1624
- if (ZSTD_isError(blockSize)) return blockSize;
1625
-
1626
- ip += ZSTD_blockHeaderSize;
1627
- remainingSize -= ZSTD_blockHeaderSize;
1628
- if (blockSize > remainingSize) return (size_t)-ZSTD_ERROR_SrcSize;
1629
-
1630
- switch(blockProperties.blockType)
1631
- {
1632
- case bt_compressed:
1633
- errorCode = ZSTD_decompressBlock(ctx, op, oend-op, ip, blockSize);
1634
- break;
1635
- case bt_raw :
1636
- errorCode = ZSTD_copyUncompressedBlock(op, oend-op, ip, blockSize);
1637
- break;
1638
- case bt_rle :
1639
- return (size_t)-ZSTD_ERROR_GENERIC; /* not yet supported */
1640
- break;
1641
- case bt_end :
1642
- /* end of frame */
1643
- if (remainingSize) return (size_t)-ZSTD_ERROR_SrcSize;
1644
- break;
1645
- default:
1646
- return (size_t)-ZSTD_ERROR_GENERIC;
1647
- }
1648
- if (blockSize == 0) break; /* bt_end */
1649
-
1650
- if (ZSTD_isError(errorCode)) return errorCode;
1651
- op += errorCode;
1652
- ip += blockSize;
1653
- remainingSize -= blockSize;
1654
- }
1655
-
1656
- return op-ostart;
1657
- }
1658
-
1659
- size_t ZSTD_decompress(void* dst, size_t maxDstSize, const void* src, size_t srcSize)
1660
- {
1661
- dctx_t ctx;
1662
- ctx.base = dst;
1663
- return ZSTD_decompressDCtx(&ctx, dst, maxDstSize, src, srcSize);
1664
- }
1665
-
1666
-
1667
- /*******************************
1668
- * Streaming Decompression API
1669
- *******************************/
1670
-
1671
- size_t ZSTD_resetDCtx(ZSTD_Dctx* dctx)
1672
- {
1673
- dctx->expected = ZSTD_frameHeaderSize;
1674
- dctx->phase = 0;
1675
- dctx->previousDstEnd = NULL;
1676
- dctx->base = NULL;
1677
- return 0;
1678
- }
1679
-
1680
- ZSTD_Dctx* ZSTD_createDCtx(void)
1681
- {
1682
- ZSTD_Dctx* dctx = (ZSTD_Dctx*)malloc(sizeof(ZSTD_Dctx));
1683
- if (dctx==NULL) return NULL;
1684
- ZSTD_resetDCtx(dctx);
1685
- return dctx;
1686
- }
1687
-
1688
- size_t ZSTD_freeDCtx(ZSTD_Dctx* dctx)
1689
- {
1690
- free(dctx);
1691
- return 0;
1692
- }
1693
-
1694
- size_t ZSTD_nextSrcSizeToDecompress(ZSTD_Dctx* dctx)
1695
- {
1696
- return ((dctx_t*)dctx)->expected;
1697
- }
1698
-
1699
- size_t ZSTD_decompressContinue(ZSTD_Dctx* dctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize)
1700
- {
1701
- dctx_t* ctx = (dctx_t*)dctx;
1702
-
1703
- /* Sanity check */
1704
- if (srcSize != ctx->expected) return (size_t)-ZSTD_ERROR_SrcSize;
1705
- if (dst != ctx->previousDstEnd) /* not contiguous */
1706
- ctx->base = dst;
1707
-
1708
- /* Decompress : frame header */
1709
- if (ctx->phase == 0)
1710
- {
1711
- /* Check frame magic header */
1712
- U32 magicNumber = ZSTD_readBE32(src);
1713
- if (magicNumber != ZSTD_magicNumber) return (size_t)-ZSTD_ERROR_MagicNumber;
1714
- ctx->phase = 1;
1715
- ctx->expected = ZSTD_blockHeaderSize;
1716
- return 0;
1717
- }
1718
-
1719
- /* Decompress : block header */
1720
- if (ctx->phase == 1)
1721
- {
1722
- blockProperties_t bp;
1723
- size_t blockSize = ZSTD_getcBlockSize(src, ZSTD_blockHeaderSize, &bp);
1724
- if (ZSTD_isError(blockSize)) return blockSize;
1725
- if (bp.blockType == bt_end)
1726
- {
1727
- ctx->expected = 0;
1728
- ctx->phase = 0;
1729
- }
1730
- else
1731
- {
1732
- ctx->expected = blockSize;
1733
- ctx->bType = bp.blockType;
1734
- ctx->phase = 2;
1735
- }
1736
-
1737
- return 0;
1738
- }
1739
-
1740
- /* Decompress : block content */
1741
- {
1742
- size_t rSize;
1743
- switch(ctx->bType)
1744
- {
1745
- case bt_compressed:
1746
- rSize = ZSTD_decompressBlock(ctx, dst, maxDstSize, src, srcSize);
1747
- break;
1748
- case bt_raw :
1749
- rSize = ZSTD_copyUncompressedBlock(dst, maxDstSize, src, srcSize);
1750
- break;
1751
- case bt_rle :
1752
- return (size_t)-ZSTD_ERROR_GENERIC; /* not yet handled */
1753
- break;
1754
- case bt_end : /* should never happen (filtered at phase 1) */
1755
- rSize = 0;
1756
- break;
1757
- default:
1758
- return (size_t)-ZSTD_ERROR_GENERIC;
1759
- }
1760
- ctx->phase = 1;
1761
- ctx->expected = ZSTD_blockHeaderSize;
1762
- ctx->previousDstEnd = (void*)( ((char*)dst) + rSize);
1763
- return rSize;
1764
- }
1765
-
1766
- }
1767
-
1768
-