zstdlib 0.13.0-x86-linux

Sign up to get free protection for your applications and to get access to all the features.
Files changed (129) hide show
  1. checksums.yaml +7 -0
  2. data/.yardopts +6 -0
  3. data/CHANGES.md +107 -0
  4. data/Gemfile +3 -0
  5. data/README.md +107 -0
  6. data/Rakefile +59 -0
  7. data/ext/zstdlib_c/extconf.rb +59 -0
  8. data/ext/zstdlib_c/ruby/zlib-2.2/zstdlib.c +4675 -0
  9. data/ext/zstdlib_c/ruby/zlib-2.3/zstdlib.c +4702 -0
  10. data/ext/zstdlib_c/ruby/zlib-2.4/zstdlib.c +4859 -0
  11. data/ext/zstdlib_c/ruby/zlib-2.5/zstdlib.c +4864 -0
  12. data/ext/zstdlib_c/ruby/zlib-2.6/zstdlib.c +4906 -0
  13. data/ext/zstdlib_c/ruby/zlib-2.7/zstdlib.c +4895 -0
  14. data/ext/zstdlib_c/ruby/zlib-3.0/zstdlib.c +4994 -0
  15. data/ext/zstdlib_c/ruby/zlib-3.1/zstdlib.c +5076 -0
  16. data/ext/zstdlib_c/ruby/zlib-3.2/zstdlib.c +5090 -0
  17. data/ext/zstdlib_c/ruby/zlib-3.3/zstdlib.c +5090 -0
  18. data/ext/zstdlib_c/zlib-1.3.1/adler32.c +164 -0
  19. data/ext/zstdlib_c/zlib-1.3.1/compress.c +75 -0
  20. data/ext/zstdlib_c/zlib-1.3.1/crc32.c +1049 -0
  21. data/ext/zstdlib_c/zlib-1.3.1/crc32.h +9446 -0
  22. data/ext/zstdlib_c/zlib-1.3.1/deflate.c +2139 -0
  23. data/ext/zstdlib_c/zlib-1.3.1/deflate.h +377 -0
  24. data/ext/zstdlib_c/zlib-1.3.1/gzclose.c +23 -0
  25. data/ext/zstdlib_c/zlib-1.3.1/gzguts.h +214 -0
  26. data/ext/zstdlib_c/zlib-1.3.1/gzlib.c +582 -0
  27. data/ext/zstdlib_c/zlib-1.3.1/gzread.c +602 -0
  28. data/ext/zstdlib_c/zlib-1.3.1/gzwrite.c +631 -0
  29. data/ext/zstdlib_c/zlib-1.3.1/infback.c +628 -0
  30. data/ext/zstdlib_c/zlib-1.3.1/inffast.c +320 -0
  31. data/ext/zstdlib_c/zlib-1.3.1/inffast.h +11 -0
  32. data/ext/zstdlib_c/zlib-1.3.1/inffixed.h +94 -0
  33. data/ext/zstdlib_c/zlib-1.3.1/inflate.c +1526 -0
  34. data/ext/zstdlib_c/zlib-1.3.1/inflate.h +126 -0
  35. data/ext/zstdlib_c/zlib-1.3.1/inftrees.c +299 -0
  36. data/ext/zstdlib_c/zlib-1.3.1/inftrees.h +62 -0
  37. data/ext/zstdlib_c/zlib-1.3.1/trees.c +1117 -0
  38. data/ext/zstdlib_c/zlib-1.3.1/trees.h +128 -0
  39. data/ext/zstdlib_c/zlib-1.3.1/uncompr.c +85 -0
  40. data/ext/zstdlib_c/zlib-1.3.1/zconf.h +543 -0
  41. data/ext/zstdlib_c/zlib-1.3.1/zlib.h +1938 -0
  42. data/ext/zstdlib_c/zlib-1.3.1/zutil.c +299 -0
  43. data/ext/zstdlib_c/zlib-1.3.1/zutil.h +254 -0
  44. data/ext/zstdlib_c/zlib.mk +14 -0
  45. data/ext/zstdlib_c/zlibwrapper/zlibwrapper.c +10 -0
  46. data/ext/zstdlib_c/zlibwrapper.mk +14 -0
  47. data/ext/zstdlib_c/zstd-1.5.6/lib/common/allocations.h +55 -0
  48. data/ext/zstdlib_c/zstd-1.5.6/lib/common/bits.h +200 -0
  49. data/ext/zstdlib_c/zstd-1.5.6/lib/common/bitstream.h +457 -0
  50. data/ext/zstdlib_c/zstd-1.5.6/lib/common/compiler.h +450 -0
  51. data/ext/zstdlib_c/zstd-1.5.6/lib/common/cpu.h +249 -0
  52. data/ext/zstdlib_c/zstd-1.5.6/lib/common/debug.c +30 -0
  53. data/ext/zstdlib_c/zstd-1.5.6/lib/common/debug.h +116 -0
  54. data/ext/zstdlib_c/zstd-1.5.6/lib/common/entropy_common.c +340 -0
  55. data/ext/zstdlib_c/zstd-1.5.6/lib/common/error_private.c +63 -0
  56. data/ext/zstdlib_c/zstd-1.5.6/lib/common/error_private.h +168 -0
  57. data/ext/zstdlib_c/zstd-1.5.6/lib/common/fse.h +640 -0
  58. data/ext/zstdlib_c/zstd-1.5.6/lib/common/fse_decompress.c +313 -0
  59. data/ext/zstdlib_c/zstd-1.5.6/lib/common/huf.h +286 -0
  60. data/ext/zstdlib_c/zstd-1.5.6/lib/common/mem.h +426 -0
  61. data/ext/zstdlib_c/zstd-1.5.6/lib/common/pool.c +371 -0
  62. data/ext/zstdlib_c/zstd-1.5.6/lib/common/pool.h +90 -0
  63. data/ext/zstdlib_c/zstd-1.5.6/lib/common/portability_macros.h +158 -0
  64. data/ext/zstdlib_c/zstd-1.5.6/lib/common/threading.c +182 -0
  65. data/ext/zstdlib_c/zstd-1.5.6/lib/common/threading.h +150 -0
  66. data/ext/zstdlib_c/zstd-1.5.6/lib/common/xxhash.c +18 -0
  67. data/ext/zstdlib_c/zstd-1.5.6/lib/common/xxhash.h +7020 -0
  68. data/ext/zstdlib_c/zstd-1.5.6/lib/common/zstd_common.c +48 -0
  69. data/ext/zstdlib_c/zstd-1.5.6/lib/common/zstd_deps.h +111 -0
  70. data/ext/zstdlib_c/zstd-1.5.6/lib/common/zstd_internal.h +392 -0
  71. data/ext/zstdlib_c/zstd-1.5.6/lib/common/zstd_trace.h +163 -0
  72. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/clevels.h +134 -0
  73. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/fse_compress.c +625 -0
  74. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/hist.c +181 -0
  75. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/hist.h +75 -0
  76. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/huf_compress.c +1464 -0
  77. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_compress.c +7153 -0
  78. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_compress_internal.h +1534 -0
  79. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_compress_literals.c +235 -0
  80. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_compress_literals.h +39 -0
  81. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_compress_sequences.c +442 -0
  82. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_compress_sequences.h +54 -0
  83. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_compress_superblock.c +688 -0
  84. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_compress_superblock.h +32 -0
  85. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_cwksp.h +748 -0
  86. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_double_fast.c +770 -0
  87. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_double_fast.h +50 -0
  88. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_fast.c +968 -0
  89. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_fast.h +38 -0
  90. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_lazy.c +2199 -0
  91. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_lazy.h +202 -0
  92. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_ldm.c +730 -0
  93. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_ldm.h +117 -0
  94. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_ldm_geartab.h +106 -0
  95. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_opt.c +1576 -0
  96. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstd_opt.h +80 -0
  97. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstdmt_compress.c +1882 -0
  98. data/ext/zstdlib_c/zstd-1.5.6/lib/compress/zstdmt_compress.h +113 -0
  99. data/ext/zstdlib_c/zstd-1.5.6/lib/decompress/huf_decompress.c +1944 -0
  100. data/ext/zstdlib_c/zstd-1.5.6/lib/decompress/huf_decompress_amd64.S +595 -0
  101. data/ext/zstdlib_c/zstd-1.5.6/lib/decompress/zstd_ddict.c +244 -0
  102. data/ext/zstdlib_c/zstd-1.5.6/lib/decompress/zstd_ddict.h +44 -0
  103. data/ext/zstdlib_c/zstd-1.5.6/lib/decompress/zstd_decompress.c +2407 -0
  104. data/ext/zstdlib_c/zstd-1.5.6/lib/decompress/zstd_decompress_block.c +2215 -0
  105. data/ext/zstdlib_c/zstd-1.5.6/lib/decompress/zstd_decompress_block.h +73 -0
  106. data/ext/zstdlib_c/zstd-1.5.6/lib/decompress/zstd_decompress_internal.h +240 -0
  107. data/ext/zstdlib_c/zstd-1.5.6/lib/zdict.h +474 -0
  108. data/ext/zstdlib_c/zstd-1.5.6/lib/zstd.h +3089 -0
  109. data/ext/zstdlib_c/zstd-1.5.6/lib/zstd_errors.h +114 -0
  110. data/ext/zstdlib_c/zstd-1.5.6/zlibWrapper/gzclose.c +26 -0
  111. data/ext/zstdlib_c/zstd-1.5.6/zlibWrapper/gzcompatibility.h +68 -0
  112. data/ext/zstdlib_c/zstd-1.5.6/zlibWrapper/gzguts.h +229 -0
  113. data/ext/zstdlib_c/zstd-1.5.6/zlibWrapper/gzlib.c +587 -0
  114. data/ext/zstdlib_c/zstd-1.5.6/zlibWrapper/gzread.c +637 -0
  115. data/ext/zstdlib_c/zstd-1.5.6/zlibWrapper/gzwrite.c +631 -0
  116. data/ext/zstdlib_c/zstd-1.5.6/zlibWrapper/zstd_zlibwrapper.c +1200 -0
  117. data/ext/zstdlib_c/zstd-1.5.6/zlibWrapper/zstd_zlibwrapper.h +91 -0
  118. data/ext/zstdlib_c/zstd.mk +15 -0
  119. data/lib/2.4/zstdlib_c.so +0 -0
  120. data/lib/2.5/zstdlib_c.so +0 -0
  121. data/lib/2.6/zstdlib_c.so +0 -0
  122. data/lib/2.7/zstdlib_c.so +0 -0
  123. data/lib/3.0/zstdlib_c.so +0 -0
  124. data/lib/3.1/zstdlib_c.so +0 -0
  125. data/lib/3.2/zstdlib_c.so +0 -0
  126. data/lib/3.3/zstdlib_c.so +0 -0
  127. data/lib/zstdlib.rb +6 -0
  128. data/test/zstdlib_test.rb +21 -0
  129. metadata +243 -0
@@ -0,0 +1,625 @@
1
+ /* ******************************************************************
2
+ * FSE : Finite State Entropy encoder
3
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
4
+ *
5
+ * You can contact the author at :
6
+ * - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
7
+ * - Public forum : https://groups.google.com/forum/#!forum/lz4c
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
+ /* **************************************************************
16
+ * Includes
17
+ ****************************************************************/
18
+ #include "../common/compiler.h"
19
+ #include "../common/mem.h" /* U32, U16, etc. */
20
+ #include "../common/debug.h" /* assert, DEBUGLOG */
21
+ #include "hist.h" /* HIST_count_wksp */
22
+ #include "../common/bitstream.h"
23
+ #define FSE_STATIC_LINKING_ONLY
24
+ #include "../common/fse.h"
25
+ #include "../common/error_private.h"
26
+ #define ZSTD_DEPS_NEED_MALLOC
27
+ #define ZSTD_DEPS_NEED_MATH64
28
+ #include "../common/zstd_deps.h" /* ZSTD_memset */
29
+ #include "../common/bits.h" /* ZSTD_highbit32 */
30
+
31
+
32
+ /* **************************************************************
33
+ * Error Management
34
+ ****************************************************************/
35
+ #define FSE_isError ERR_isError
36
+
37
+
38
+ /* **************************************************************
39
+ * Templates
40
+ ****************************************************************/
41
+ /*
42
+ designed to be included
43
+ for type-specific functions (template emulation in C)
44
+ Objective is to write these functions only once, for improved maintenance
45
+ */
46
+
47
+ /* safety checks */
48
+ #ifndef FSE_FUNCTION_EXTENSION
49
+ # error "FSE_FUNCTION_EXTENSION must be defined"
50
+ #endif
51
+ #ifndef FSE_FUNCTION_TYPE
52
+ # error "FSE_FUNCTION_TYPE must be defined"
53
+ #endif
54
+
55
+ /* Function names */
56
+ #define FSE_CAT(X,Y) X##Y
57
+ #define FSE_FUNCTION_NAME(X,Y) FSE_CAT(X,Y)
58
+ #define FSE_TYPE_NAME(X,Y) FSE_CAT(X,Y)
59
+
60
+
61
+ /* Function templates */
62
+
63
+ /* FSE_buildCTable_wksp() :
64
+ * Same as FSE_buildCTable(), but using an externally allocated scratch buffer (`workSpace`).
65
+ * wkspSize should be sized to handle worst case situation, which is `1<<max_tableLog * sizeof(FSE_FUNCTION_TYPE)`
66
+ * workSpace must also be properly aligned with FSE_FUNCTION_TYPE requirements
67
+ */
68
+ size_t FSE_buildCTable_wksp(FSE_CTable* ct,
69
+ const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog,
70
+ void* workSpace, size_t wkspSize)
71
+ {
72
+ U32 const tableSize = 1 << tableLog;
73
+ U32 const tableMask = tableSize - 1;
74
+ void* const ptr = ct;
75
+ U16* const tableU16 = ( (U16*) ptr) + 2;
76
+ void* const FSCT = ((U32*)ptr) + 1 /* header */ + (tableLog ? tableSize>>1 : 1) ;
77
+ FSE_symbolCompressionTransform* const symbolTT = (FSE_symbolCompressionTransform*) (FSCT);
78
+ U32 const step = FSE_TABLESTEP(tableSize);
79
+ U32 const maxSV1 = maxSymbolValue+1;
80
+
81
+ U16* cumul = (U16*)workSpace; /* size = maxSV1 */
82
+ FSE_FUNCTION_TYPE* const tableSymbol = (FSE_FUNCTION_TYPE*)(cumul + (maxSV1+1)); /* size = tableSize */
83
+
84
+ U32 highThreshold = tableSize-1;
85
+
86
+ assert(((size_t)workSpace & 1) == 0); /* Must be 2 bytes-aligned */
87
+ if (FSE_BUILD_CTABLE_WORKSPACE_SIZE(maxSymbolValue, tableLog) > wkspSize) return ERROR(tableLog_tooLarge);
88
+ /* CTable header */
89
+ tableU16[-2] = (U16) tableLog;
90
+ tableU16[-1] = (U16) maxSymbolValue;
91
+ assert(tableLog < 16); /* required for threshold strategy to work */
92
+
93
+ /* For explanations on how to distribute symbol values over the table :
94
+ * https://fastcompression.blogspot.fr/2014/02/fse-distributing-symbol-values.html */
95
+
96
+ #ifdef __clang_analyzer__
97
+ ZSTD_memset(tableSymbol, 0, sizeof(*tableSymbol) * tableSize); /* useless initialization, just to keep scan-build happy */
98
+ #endif
99
+
100
+ /* symbol start positions */
101
+ { U32 u;
102
+ cumul[0] = 0;
103
+ for (u=1; u <= maxSV1; u++) {
104
+ if (normalizedCounter[u-1]==-1) { /* Low proba symbol */
105
+ cumul[u] = cumul[u-1] + 1;
106
+ tableSymbol[highThreshold--] = (FSE_FUNCTION_TYPE)(u-1);
107
+ } else {
108
+ assert(normalizedCounter[u-1] >= 0);
109
+ cumul[u] = cumul[u-1] + (U16)normalizedCounter[u-1];
110
+ assert(cumul[u] >= cumul[u-1]); /* no overflow */
111
+ } }
112
+ cumul[maxSV1] = (U16)(tableSize+1);
113
+ }
114
+
115
+ /* Spread symbols */
116
+ if (highThreshold == tableSize - 1) {
117
+ /* Case for no low prob count symbols. Lay down 8 bytes at a time
118
+ * to reduce branch misses since we are operating on a small block
119
+ */
120
+ BYTE* const spread = tableSymbol + tableSize; /* size = tableSize + 8 (may write beyond tableSize) */
121
+ { U64 const add = 0x0101010101010101ull;
122
+ size_t pos = 0;
123
+ U64 sv = 0;
124
+ U32 s;
125
+ for (s=0; s<maxSV1; ++s, sv += add) {
126
+ int i;
127
+ int const n = normalizedCounter[s];
128
+ MEM_write64(spread + pos, sv);
129
+ for (i = 8; i < n; i += 8) {
130
+ MEM_write64(spread + pos + i, sv);
131
+ }
132
+ assert(n>=0);
133
+ pos += (size_t)n;
134
+ }
135
+ }
136
+ /* Spread symbols across the table. Lack of lowprob symbols means that
137
+ * we don't need variable sized inner loop, so we can unroll the loop and
138
+ * reduce branch misses.
139
+ */
140
+ { size_t position = 0;
141
+ size_t s;
142
+ size_t const unroll = 2; /* Experimentally determined optimal unroll */
143
+ assert(tableSize % unroll == 0); /* FSE_MIN_TABLELOG is 5 */
144
+ for (s = 0; s < (size_t)tableSize; s += unroll) {
145
+ size_t u;
146
+ for (u = 0; u < unroll; ++u) {
147
+ size_t const uPosition = (position + (u * step)) & tableMask;
148
+ tableSymbol[uPosition] = spread[s + u];
149
+ }
150
+ position = (position + (unroll * step)) & tableMask;
151
+ }
152
+ assert(position == 0); /* Must have initialized all positions */
153
+ }
154
+ } else {
155
+ U32 position = 0;
156
+ U32 symbol;
157
+ for (symbol=0; symbol<maxSV1; symbol++) {
158
+ int nbOccurrences;
159
+ int const freq = normalizedCounter[symbol];
160
+ for (nbOccurrences=0; nbOccurrences<freq; nbOccurrences++) {
161
+ tableSymbol[position] = (FSE_FUNCTION_TYPE)symbol;
162
+ position = (position + step) & tableMask;
163
+ while (position > highThreshold)
164
+ position = (position + step) & tableMask; /* Low proba area */
165
+ } }
166
+ assert(position==0); /* Must have initialized all positions */
167
+ }
168
+
169
+ /* Build table */
170
+ { U32 u; for (u=0; u<tableSize; u++) {
171
+ FSE_FUNCTION_TYPE s = tableSymbol[u]; /* note : static analyzer may not understand tableSymbol is properly initialized */
172
+ tableU16[cumul[s]++] = (U16) (tableSize+u); /* TableU16 : sorted by symbol order; gives next state value */
173
+ } }
174
+
175
+ /* Build Symbol Transformation Table */
176
+ { unsigned total = 0;
177
+ unsigned s;
178
+ for (s=0; s<=maxSymbolValue; s++) {
179
+ switch (normalizedCounter[s])
180
+ {
181
+ case 0:
182
+ /* filling nonetheless, for compatibility with FSE_getMaxNbBits() */
183
+ symbolTT[s].deltaNbBits = ((tableLog+1) << 16) - (1<<tableLog);
184
+ break;
185
+
186
+ case -1:
187
+ case 1:
188
+ symbolTT[s].deltaNbBits = (tableLog << 16) - (1<<tableLog);
189
+ assert(total <= INT_MAX);
190
+ symbolTT[s].deltaFindState = (int)(total - 1);
191
+ total ++;
192
+ break;
193
+ default :
194
+ assert(normalizedCounter[s] > 1);
195
+ { U32 const maxBitsOut = tableLog - ZSTD_highbit32 ((U32)normalizedCounter[s]-1);
196
+ U32 const minStatePlus = (U32)normalizedCounter[s] << maxBitsOut;
197
+ symbolTT[s].deltaNbBits = (maxBitsOut << 16) - minStatePlus;
198
+ symbolTT[s].deltaFindState = (int)(total - (unsigned)normalizedCounter[s]);
199
+ total += (unsigned)normalizedCounter[s];
200
+ } } } }
201
+
202
+ #if 0 /* debug : symbol costs */
203
+ DEBUGLOG(5, "\n --- table statistics : ");
204
+ { U32 symbol;
205
+ for (symbol=0; symbol<=maxSymbolValue; symbol++) {
206
+ DEBUGLOG(5, "%3u: w=%3i, maxBits=%u, fracBits=%.2f",
207
+ symbol, normalizedCounter[symbol],
208
+ FSE_getMaxNbBits(symbolTT, symbol),
209
+ (double)FSE_bitCost(symbolTT, tableLog, symbol, 8) / 256);
210
+ } }
211
+ #endif
212
+
213
+ return 0;
214
+ }
215
+
216
+
217
+
218
+ #ifndef FSE_COMMONDEFS_ONLY
219
+
220
+ /*-**************************************************************
221
+ * FSE NCount encoding
222
+ ****************************************************************/
223
+ size_t FSE_NCountWriteBound(unsigned maxSymbolValue, unsigned tableLog)
224
+ {
225
+ size_t const maxHeaderSize = (((maxSymbolValue+1) * tableLog
226
+ + 4 /* bitCount initialized at 4 */
227
+ + 2 /* first two symbols may use one additional bit each */) / 8)
228
+ + 1 /* round up to whole nb bytes */
229
+ + 2 /* additional two bytes for bitstream flush */;
230
+ return maxSymbolValue ? maxHeaderSize : FSE_NCOUNTBOUND; /* maxSymbolValue==0 ? use default */
231
+ }
232
+
233
+ static size_t
234
+ FSE_writeNCount_generic (void* header, size_t headerBufferSize,
235
+ const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog,
236
+ unsigned writeIsSafe)
237
+ {
238
+ BYTE* const ostart = (BYTE*) header;
239
+ BYTE* out = ostart;
240
+ BYTE* const oend = ostart + headerBufferSize;
241
+ int nbBits;
242
+ const int tableSize = 1 << tableLog;
243
+ int remaining;
244
+ int threshold;
245
+ U32 bitStream = 0;
246
+ int bitCount = 0;
247
+ unsigned symbol = 0;
248
+ unsigned const alphabetSize = maxSymbolValue + 1;
249
+ int previousIs0 = 0;
250
+
251
+ /* Table Size */
252
+ bitStream += (tableLog-FSE_MIN_TABLELOG) << bitCount;
253
+ bitCount += 4;
254
+
255
+ /* Init */
256
+ remaining = tableSize+1; /* +1 for extra accuracy */
257
+ threshold = tableSize;
258
+ nbBits = (int)tableLog+1;
259
+
260
+ while ((symbol < alphabetSize) && (remaining>1)) { /* stops at 1 */
261
+ if (previousIs0) {
262
+ unsigned start = symbol;
263
+ while ((symbol < alphabetSize) && !normalizedCounter[symbol]) symbol++;
264
+ if (symbol == alphabetSize) break; /* incorrect distribution */
265
+ while (symbol >= start+24) {
266
+ start+=24;
267
+ bitStream += 0xFFFFU << bitCount;
268
+ if ((!writeIsSafe) && (out > oend-2))
269
+ return ERROR(dstSize_tooSmall); /* Buffer overflow */
270
+ out[0] = (BYTE) bitStream;
271
+ out[1] = (BYTE)(bitStream>>8);
272
+ out+=2;
273
+ bitStream>>=16;
274
+ }
275
+ while (symbol >= start+3) {
276
+ start+=3;
277
+ bitStream += 3U << bitCount;
278
+ bitCount += 2;
279
+ }
280
+ bitStream += (symbol-start) << bitCount;
281
+ bitCount += 2;
282
+ if (bitCount>16) {
283
+ if ((!writeIsSafe) && (out > oend - 2))
284
+ return ERROR(dstSize_tooSmall); /* Buffer overflow */
285
+ out[0] = (BYTE)bitStream;
286
+ out[1] = (BYTE)(bitStream>>8);
287
+ out += 2;
288
+ bitStream >>= 16;
289
+ bitCount -= 16;
290
+ } }
291
+ { int count = normalizedCounter[symbol++];
292
+ int const max = (2*threshold-1) - remaining;
293
+ remaining -= count < 0 ? -count : count;
294
+ count++; /* +1 for extra accuracy */
295
+ if (count>=threshold)
296
+ count += max; /* [0..max[ [max..threshold[ (...) [threshold+max 2*threshold[ */
297
+ bitStream += (U32)count << bitCount;
298
+ bitCount += nbBits;
299
+ bitCount -= (count<max);
300
+ previousIs0 = (count==1);
301
+ if (remaining<1) return ERROR(GENERIC);
302
+ while (remaining<threshold) { nbBits--; threshold>>=1; }
303
+ }
304
+ if (bitCount>16) {
305
+ if ((!writeIsSafe) && (out > oend - 2))
306
+ return ERROR(dstSize_tooSmall); /* Buffer overflow */
307
+ out[0] = (BYTE)bitStream;
308
+ out[1] = (BYTE)(bitStream>>8);
309
+ out += 2;
310
+ bitStream >>= 16;
311
+ bitCount -= 16;
312
+ } }
313
+
314
+ if (remaining != 1)
315
+ return ERROR(GENERIC); /* incorrect normalized distribution */
316
+ assert(symbol <= alphabetSize);
317
+
318
+ /* flush remaining bitStream */
319
+ if ((!writeIsSafe) && (out > oend - 2))
320
+ return ERROR(dstSize_tooSmall); /* Buffer overflow */
321
+ out[0] = (BYTE)bitStream;
322
+ out[1] = (BYTE)(bitStream>>8);
323
+ out+= (bitCount+7) /8;
324
+
325
+ assert(out >= ostart);
326
+ return (size_t)(out-ostart);
327
+ }
328
+
329
+
330
+ size_t FSE_writeNCount (void* buffer, size_t bufferSize,
331
+ const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
332
+ {
333
+ if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge); /* Unsupported */
334
+ if (tableLog < FSE_MIN_TABLELOG) return ERROR(GENERIC); /* Unsupported */
335
+
336
+ if (bufferSize < FSE_NCountWriteBound(maxSymbolValue, tableLog))
337
+ return FSE_writeNCount_generic(buffer, bufferSize, normalizedCounter, maxSymbolValue, tableLog, 0);
338
+
339
+ return FSE_writeNCount_generic(buffer, bufferSize, normalizedCounter, maxSymbolValue, tableLog, 1 /* write in buffer is safe */);
340
+ }
341
+
342
+
343
+ /*-**************************************************************
344
+ * FSE Compression Code
345
+ ****************************************************************/
346
+
347
+ /* provides the minimum logSize to safely represent a distribution */
348
+ static unsigned FSE_minTableLog(size_t srcSize, unsigned maxSymbolValue)
349
+ {
350
+ U32 minBitsSrc = ZSTD_highbit32((U32)(srcSize)) + 1;
351
+ U32 minBitsSymbols = ZSTD_highbit32(maxSymbolValue) + 2;
352
+ U32 minBits = minBitsSrc < minBitsSymbols ? minBitsSrc : minBitsSymbols;
353
+ assert(srcSize > 1); /* Not supported, RLE should be used instead */
354
+ return minBits;
355
+ }
356
+
357
+ unsigned FSE_optimalTableLog_internal(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue, unsigned minus)
358
+ {
359
+ U32 maxBitsSrc = ZSTD_highbit32((U32)(srcSize - 1)) - minus;
360
+ U32 tableLog = maxTableLog;
361
+ U32 minBits = FSE_minTableLog(srcSize, maxSymbolValue);
362
+ assert(srcSize > 1); /* Not supported, RLE should be used instead */
363
+ if (tableLog==0) tableLog = FSE_DEFAULT_TABLELOG;
364
+ if (maxBitsSrc < tableLog) tableLog = maxBitsSrc; /* Accuracy can be reduced */
365
+ if (minBits > tableLog) tableLog = minBits; /* Need a minimum to safely represent all symbol values */
366
+ if (tableLog < FSE_MIN_TABLELOG) tableLog = FSE_MIN_TABLELOG;
367
+ if (tableLog > FSE_MAX_TABLELOG) tableLog = FSE_MAX_TABLELOG;
368
+ return tableLog;
369
+ }
370
+
371
+ unsigned FSE_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue)
372
+ {
373
+ return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 2);
374
+ }
375
+
376
+ /* Secondary normalization method.
377
+ To be used when primary method fails. */
378
+
379
+ static size_t FSE_normalizeM2(short* norm, U32 tableLog, const unsigned* count, size_t total, U32 maxSymbolValue, short lowProbCount)
380
+ {
381
+ short const NOT_YET_ASSIGNED = -2;
382
+ U32 s;
383
+ U32 distributed = 0;
384
+ U32 ToDistribute;
385
+
386
+ /* Init */
387
+ U32 const lowThreshold = (U32)(total >> tableLog);
388
+ U32 lowOne = (U32)((total * 3) >> (tableLog + 1));
389
+
390
+ for (s=0; s<=maxSymbolValue; s++) {
391
+ if (count[s] == 0) {
392
+ norm[s]=0;
393
+ continue;
394
+ }
395
+ if (count[s] <= lowThreshold) {
396
+ norm[s] = lowProbCount;
397
+ distributed++;
398
+ total -= count[s];
399
+ continue;
400
+ }
401
+ if (count[s] <= lowOne) {
402
+ norm[s] = 1;
403
+ distributed++;
404
+ total -= count[s];
405
+ continue;
406
+ }
407
+
408
+ norm[s]=NOT_YET_ASSIGNED;
409
+ }
410
+ ToDistribute = (1 << tableLog) - distributed;
411
+
412
+ if (ToDistribute == 0)
413
+ return 0;
414
+
415
+ if ((total / ToDistribute) > lowOne) {
416
+ /* risk of rounding to zero */
417
+ lowOne = (U32)((total * 3) / (ToDistribute * 2));
418
+ for (s=0; s<=maxSymbolValue; s++) {
419
+ if ((norm[s] == NOT_YET_ASSIGNED) && (count[s] <= lowOne)) {
420
+ norm[s] = 1;
421
+ distributed++;
422
+ total -= count[s];
423
+ continue;
424
+ } }
425
+ ToDistribute = (1 << tableLog) - distributed;
426
+ }
427
+
428
+ if (distributed == maxSymbolValue+1) {
429
+ /* all values are pretty poor;
430
+ probably incompressible data (should have already been detected);
431
+ find max, then give all remaining points to max */
432
+ U32 maxV = 0, maxC = 0;
433
+ for (s=0; s<=maxSymbolValue; s++)
434
+ if (count[s] > maxC) { maxV=s; maxC=count[s]; }
435
+ norm[maxV] += (short)ToDistribute;
436
+ return 0;
437
+ }
438
+
439
+ if (total == 0) {
440
+ /* all of the symbols were low enough for the lowOne or lowThreshold */
441
+ for (s=0; ToDistribute > 0; s = (s+1)%(maxSymbolValue+1))
442
+ if (norm[s] > 0) { ToDistribute--; norm[s]++; }
443
+ return 0;
444
+ }
445
+
446
+ { U64 const vStepLog = 62 - tableLog;
447
+ U64 const mid = (1ULL << (vStepLog-1)) - 1;
448
+ U64 const rStep = ZSTD_div64((((U64)1<<vStepLog) * ToDistribute) + mid, (U32)total); /* scale on remaining */
449
+ U64 tmpTotal = mid;
450
+ for (s=0; s<=maxSymbolValue; s++) {
451
+ if (norm[s]==NOT_YET_ASSIGNED) {
452
+ U64 const end = tmpTotal + (count[s] * rStep);
453
+ U32 const sStart = (U32)(tmpTotal >> vStepLog);
454
+ U32 const sEnd = (U32)(end >> vStepLog);
455
+ U32 const weight = sEnd - sStart;
456
+ if (weight < 1)
457
+ return ERROR(GENERIC);
458
+ norm[s] = (short)weight;
459
+ tmpTotal = end;
460
+ } } }
461
+
462
+ return 0;
463
+ }
464
+
465
+ size_t FSE_normalizeCount (short* normalizedCounter, unsigned tableLog,
466
+ const unsigned* count, size_t total,
467
+ unsigned maxSymbolValue, unsigned useLowProbCount)
468
+ {
469
+ /* Sanity checks */
470
+ if (tableLog==0) tableLog = FSE_DEFAULT_TABLELOG;
471
+ if (tableLog < FSE_MIN_TABLELOG) return ERROR(GENERIC); /* Unsupported size */
472
+ if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge); /* Unsupported size */
473
+ if (tableLog < FSE_minTableLog(total, maxSymbolValue)) return ERROR(GENERIC); /* Too small tableLog, compression potentially impossible */
474
+
475
+ { static U32 const rtbTable[] = { 0, 473195, 504333, 520860, 550000, 700000, 750000, 830000 };
476
+ short const lowProbCount = useLowProbCount ? -1 : 1;
477
+ U64 const scale = 62 - tableLog;
478
+ U64 const step = ZSTD_div64((U64)1<<62, (U32)total); /* <== here, one division ! */
479
+ U64 const vStep = 1ULL<<(scale-20);
480
+ int stillToDistribute = 1<<tableLog;
481
+ unsigned s;
482
+ unsigned largest=0;
483
+ short largestP=0;
484
+ U32 lowThreshold = (U32)(total >> tableLog);
485
+
486
+ for (s=0; s<=maxSymbolValue; s++) {
487
+ if (count[s] == total) return 0; /* rle special case */
488
+ if (count[s] == 0) { normalizedCounter[s]=0; continue; }
489
+ if (count[s] <= lowThreshold) {
490
+ normalizedCounter[s] = lowProbCount;
491
+ stillToDistribute--;
492
+ } else {
493
+ short proba = (short)((count[s]*step) >> scale);
494
+ if (proba<8) {
495
+ U64 restToBeat = vStep * rtbTable[proba];
496
+ proba += (count[s]*step) - ((U64)proba<<scale) > restToBeat;
497
+ }
498
+ if (proba > largestP) { largestP=proba; largest=s; }
499
+ normalizedCounter[s] = proba;
500
+ stillToDistribute -= proba;
501
+ } }
502
+ if (-stillToDistribute >= (normalizedCounter[largest] >> 1)) {
503
+ /* corner case, need another normalization method */
504
+ size_t const errorCode = FSE_normalizeM2(normalizedCounter, tableLog, count, total, maxSymbolValue, lowProbCount);
505
+ if (FSE_isError(errorCode)) return errorCode;
506
+ }
507
+ else normalizedCounter[largest] += (short)stillToDistribute;
508
+ }
509
+
510
+ #if 0
511
+ { /* Print Table (debug) */
512
+ U32 s;
513
+ U32 nTotal = 0;
514
+ for (s=0; s<=maxSymbolValue; s++)
515
+ RAWLOG(2, "%3i: %4i \n", s, normalizedCounter[s]);
516
+ for (s=0; s<=maxSymbolValue; s++)
517
+ nTotal += abs(normalizedCounter[s]);
518
+ if (nTotal != (1U<<tableLog))
519
+ RAWLOG(2, "Warning !!! Total == %u != %u !!!", nTotal, 1U<<tableLog);
520
+ getchar();
521
+ }
522
+ #endif
523
+
524
+ return tableLog;
525
+ }
526
+
527
+ /* fake FSE_CTable, for rle input (always same symbol) */
528
+ size_t FSE_buildCTable_rle (FSE_CTable* ct, BYTE symbolValue)
529
+ {
530
+ void* ptr = ct;
531
+ U16* tableU16 = ( (U16*) ptr) + 2;
532
+ void* FSCTptr = (U32*)ptr + 2;
533
+ FSE_symbolCompressionTransform* symbolTT = (FSE_symbolCompressionTransform*) FSCTptr;
534
+
535
+ /* header */
536
+ tableU16[-2] = (U16) 0;
537
+ tableU16[-1] = (U16) symbolValue;
538
+
539
+ /* Build table */
540
+ tableU16[0] = 0;
541
+ tableU16[1] = 0; /* just in case */
542
+
543
+ /* Build Symbol Transformation Table */
544
+ symbolTT[symbolValue].deltaNbBits = 0;
545
+ symbolTT[symbolValue].deltaFindState = 0;
546
+
547
+ return 0;
548
+ }
549
+
550
+
551
+ static size_t FSE_compress_usingCTable_generic (void* dst, size_t dstSize,
552
+ const void* src, size_t srcSize,
553
+ const FSE_CTable* ct, const unsigned fast)
554
+ {
555
+ const BYTE* const istart = (const BYTE*) src;
556
+ const BYTE* const iend = istart + srcSize;
557
+ const BYTE* ip=iend;
558
+
559
+ BIT_CStream_t bitC;
560
+ FSE_CState_t CState1, CState2;
561
+
562
+ /* init */
563
+ if (srcSize <= 2) return 0;
564
+ { size_t const initError = BIT_initCStream(&bitC, dst, dstSize);
565
+ if (FSE_isError(initError)) return 0; /* not enough space available to write a bitstream */ }
566
+
567
+ #define FSE_FLUSHBITS(s) (fast ? BIT_flushBitsFast(s) : BIT_flushBits(s))
568
+
569
+ if (srcSize & 1) {
570
+ FSE_initCState2(&CState1, ct, *--ip);
571
+ FSE_initCState2(&CState2, ct, *--ip);
572
+ FSE_encodeSymbol(&bitC, &CState1, *--ip);
573
+ FSE_FLUSHBITS(&bitC);
574
+ } else {
575
+ FSE_initCState2(&CState2, ct, *--ip);
576
+ FSE_initCState2(&CState1, ct, *--ip);
577
+ }
578
+
579
+ /* join to mod 4 */
580
+ srcSize -= 2;
581
+ if ((sizeof(bitC.bitContainer)*8 > FSE_MAX_TABLELOG*4+7 ) && (srcSize & 2)) { /* test bit 2 */
582
+ FSE_encodeSymbol(&bitC, &CState2, *--ip);
583
+ FSE_encodeSymbol(&bitC, &CState1, *--ip);
584
+ FSE_FLUSHBITS(&bitC);
585
+ }
586
+
587
+ /* 2 or 4 encoding per loop */
588
+ while ( ip>istart ) {
589
+
590
+ FSE_encodeSymbol(&bitC, &CState2, *--ip);
591
+
592
+ if (sizeof(bitC.bitContainer)*8 < FSE_MAX_TABLELOG*2+7 ) /* this test must be static */
593
+ FSE_FLUSHBITS(&bitC);
594
+
595
+ FSE_encodeSymbol(&bitC, &CState1, *--ip);
596
+
597
+ if (sizeof(bitC.bitContainer)*8 > FSE_MAX_TABLELOG*4+7 ) { /* this test must be static */
598
+ FSE_encodeSymbol(&bitC, &CState2, *--ip);
599
+ FSE_encodeSymbol(&bitC, &CState1, *--ip);
600
+ }
601
+
602
+ FSE_FLUSHBITS(&bitC);
603
+ }
604
+
605
+ FSE_flushCState(&bitC, &CState2);
606
+ FSE_flushCState(&bitC, &CState1);
607
+ return BIT_closeCStream(&bitC);
608
+ }
609
+
610
+ size_t FSE_compress_usingCTable (void* dst, size_t dstSize,
611
+ const void* src, size_t srcSize,
612
+ const FSE_CTable* ct)
613
+ {
614
+ unsigned const fast = (dstSize >= FSE_BLOCKBOUND(srcSize));
615
+
616
+ if (fast)
617
+ return FSE_compress_usingCTable_generic(dst, dstSize, src, srcSize, ct, 1);
618
+ else
619
+ return FSE_compress_usingCTable_generic(dst, dstSize, src, srcSize, ct, 0);
620
+ }
621
+
622
+
623
+ size_t FSE_compressBound(size_t size) { return FSE_COMPRESSBOUND(size); }
624
+
625
+ #endif /* FSE_COMMONDEFS_ONLY */