extzstd 0.1 → 0.3.2

Sign up to get free protection for your applications and to get access to all the features.
Files changed (134) hide show
  1. checksums.yaml +5 -5
  2. data/HISTORY.ja.md +39 -0
  3. data/README.md +38 -56
  4. data/contrib/zstd/CHANGELOG +613 -0
  5. data/contrib/zstd/CODE_OF_CONDUCT.md +5 -0
  6. data/contrib/zstd/CONTRIBUTING.md +406 -0
  7. data/contrib/zstd/COPYING +339 -0
  8. data/contrib/zstd/Makefile +420 -0
  9. data/contrib/zstd/README.md +179 -41
  10. data/contrib/zstd/TESTING.md +44 -0
  11. data/contrib/zstd/appveyor.yml +292 -0
  12. data/contrib/zstd/lib/BUCK +234 -0
  13. data/contrib/zstd/lib/Makefile +451 -0
  14. data/contrib/zstd/lib/README.md +207 -0
  15. data/contrib/zstd/{common → lib/common}/bitstream.h +187 -138
  16. data/contrib/zstd/lib/common/compiler.h +288 -0
  17. data/contrib/zstd/lib/common/cpu.h +213 -0
  18. data/contrib/zstd/lib/common/debug.c +24 -0
  19. data/contrib/zstd/lib/common/debug.h +107 -0
  20. data/contrib/zstd/lib/common/entropy_common.c +362 -0
  21. data/contrib/zstd/{common → lib/common}/error_private.c +25 -12
  22. data/contrib/zstd/{common → lib/common}/error_private.h +14 -10
  23. data/contrib/zstd/{common → lib/common}/fse.h +173 -92
  24. data/contrib/zstd/{common → lib/common}/fse_decompress.c +149 -85
  25. data/contrib/zstd/lib/common/huf.h +361 -0
  26. data/contrib/zstd/{common → lib/common}/mem.h +115 -59
  27. data/contrib/zstd/lib/common/pool.c +350 -0
  28. data/contrib/zstd/lib/common/pool.h +84 -0
  29. data/contrib/zstd/lib/common/threading.c +122 -0
  30. data/contrib/zstd/lib/common/threading.h +155 -0
  31. data/contrib/zstd/{common → lib/common}/xxhash.c +55 -96
  32. data/contrib/zstd/{common → lib/common}/xxhash.h +23 -47
  33. data/contrib/zstd/lib/common/zstd_common.c +83 -0
  34. data/contrib/zstd/lib/common/zstd_deps.h +111 -0
  35. data/contrib/zstd/lib/common/zstd_errors.h +95 -0
  36. data/contrib/zstd/lib/common/zstd_internal.h +478 -0
  37. data/contrib/zstd/{compress → lib/compress}/fse_compress.c +214 -319
  38. data/contrib/zstd/lib/compress/hist.c +181 -0
  39. data/contrib/zstd/lib/compress/hist.h +75 -0
  40. data/contrib/zstd/lib/compress/huf_compress.c +913 -0
  41. data/contrib/zstd/lib/compress/zstd_compress.c +5208 -0
  42. data/contrib/zstd/lib/compress/zstd_compress_internal.h +1203 -0
  43. data/contrib/zstd/lib/compress/zstd_compress_literals.c +158 -0
  44. data/contrib/zstd/lib/compress/zstd_compress_literals.h +29 -0
  45. data/contrib/zstd/lib/compress/zstd_compress_sequences.c +433 -0
  46. data/contrib/zstd/lib/compress/zstd_compress_sequences.h +54 -0
  47. data/contrib/zstd/lib/compress/zstd_compress_superblock.c +849 -0
  48. data/contrib/zstd/lib/compress/zstd_compress_superblock.h +32 -0
  49. data/contrib/zstd/lib/compress/zstd_cwksp.h +561 -0
  50. data/contrib/zstd/lib/compress/zstd_double_fast.c +521 -0
  51. data/contrib/zstd/lib/compress/zstd_double_fast.h +38 -0
  52. data/contrib/zstd/lib/compress/zstd_fast.c +496 -0
  53. data/contrib/zstd/lib/compress/zstd_fast.h +37 -0
  54. data/contrib/zstd/lib/compress/zstd_lazy.c +1412 -0
  55. data/contrib/zstd/lib/compress/zstd_lazy.h +87 -0
  56. data/contrib/zstd/lib/compress/zstd_ldm.c +660 -0
  57. data/contrib/zstd/lib/compress/zstd_ldm.h +116 -0
  58. data/contrib/zstd/lib/compress/zstd_opt.c +1345 -0
  59. data/contrib/zstd/lib/compress/zstd_opt.h +56 -0
  60. data/contrib/zstd/lib/compress/zstdmt_compress.c +1811 -0
  61. data/contrib/zstd/lib/compress/zstdmt_compress.h +110 -0
  62. data/contrib/zstd/lib/decompress/huf_decompress.c +1350 -0
  63. data/contrib/zstd/lib/decompress/zstd_ddict.c +244 -0
  64. data/contrib/zstd/lib/decompress/zstd_ddict.h +44 -0
  65. data/contrib/zstd/lib/decompress/zstd_decompress.c +1930 -0
  66. data/contrib/zstd/lib/decompress/zstd_decompress_block.c +1540 -0
  67. data/contrib/zstd/lib/decompress/zstd_decompress_block.h +62 -0
  68. data/contrib/zstd/lib/decompress/zstd_decompress_internal.h +190 -0
  69. data/contrib/zstd/{common → lib/deprecated}/zbuff.h +68 -45
  70. data/contrib/zstd/lib/deprecated/zbuff_common.c +26 -0
  71. data/contrib/zstd/lib/deprecated/zbuff_compress.c +147 -0
  72. data/contrib/zstd/lib/deprecated/zbuff_decompress.c +75 -0
  73. data/contrib/zstd/lib/dictBuilder/cover.c +1245 -0
  74. data/contrib/zstd/lib/dictBuilder/cover.h +157 -0
  75. data/contrib/zstd/{dictBuilder → lib/dictBuilder}/divsufsort.c +3 -3
  76. data/contrib/zstd/{dictBuilder → lib/dictBuilder}/divsufsort.h +0 -0
  77. data/contrib/zstd/lib/dictBuilder/fastcover.c +758 -0
  78. data/contrib/zstd/{dictBuilder → lib/dictBuilder}/zdict.c +318 -194
  79. data/contrib/zstd/lib/dictBuilder/zdict.h +305 -0
  80. data/contrib/zstd/{legacy → lib/legacy}/zstd_legacy.h +171 -15
  81. data/contrib/zstd/{legacy → lib/legacy}/zstd_v01.c +191 -124
  82. data/contrib/zstd/{legacy → lib/legacy}/zstd_v01.h +19 -5
  83. data/contrib/zstd/{legacy → lib/legacy}/zstd_v02.c +125 -125
  84. data/contrib/zstd/{legacy → lib/legacy}/zstd_v02.h +19 -5
  85. data/contrib/zstd/{legacy → lib/legacy}/zstd_v03.c +125 -124
  86. data/contrib/zstd/{legacy → lib/legacy}/zstd_v03.h +20 -6
  87. data/contrib/zstd/{legacy → lib/legacy}/zstd_v04.c +151 -299
  88. data/contrib/zstd/{legacy → lib/legacy}/zstd_v04.h +19 -5
  89. data/contrib/zstd/{legacy → lib/legacy}/zstd_v05.c +237 -243
  90. data/contrib/zstd/{legacy → lib/legacy}/zstd_v05.h +19 -6
  91. data/contrib/zstd/{legacy → lib/legacy}/zstd_v06.c +130 -143
  92. data/contrib/zstd/{legacy → lib/legacy}/zstd_v06.h +18 -5
  93. data/contrib/zstd/{legacy → lib/legacy}/zstd_v07.c +158 -157
  94. data/contrib/zstd/{legacy → lib/legacy}/zstd_v07.h +19 -5
  95. data/contrib/zstd/lib/libzstd.pc.in +15 -0
  96. data/contrib/zstd/lib/zstd.h +2391 -0
  97. data/ext/depend +2 -0
  98. data/ext/extconf.rb +15 -6
  99. data/ext/extzstd.c +76 -145
  100. data/ext/extzstd.h +80 -31
  101. data/ext/extzstd_stream.c +417 -142
  102. data/ext/libzstd_conf.h +8 -0
  103. data/ext/zstd_common.c +10 -7
  104. data/ext/zstd_compress.c +14 -5
  105. data/ext/zstd_decompress.c +5 -4
  106. data/ext/zstd_dictbuilder.c +9 -4
  107. data/ext/zstd_dictbuilder_fastcover.c +3 -0
  108. data/ext/zstd_legacy_v01.c +3 -1
  109. data/ext/zstd_legacy_v02.c +3 -1
  110. data/ext/zstd_legacy_v03.c +3 -1
  111. data/ext/zstd_legacy_v04.c +3 -1
  112. data/ext/zstd_legacy_v05.c +3 -1
  113. data/ext/zstd_legacy_v06.c +3 -1
  114. data/ext/zstd_legacy_v07.c +3 -1
  115. data/gemstub.rb +10 -24
  116. data/lib/extzstd.rb +64 -179
  117. data/lib/extzstd/version.rb +6 -1
  118. data/test/test_basic.rb +9 -6
  119. metadata +113 -57
  120. data/HISTORY.ja +0 -5
  121. data/contrib/zstd/common/entropy_common.c +0 -225
  122. data/contrib/zstd/common/huf.h +0 -228
  123. data/contrib/zstd/common/zstd_common.c +0 -83
  124. data/contrib/zstd/common/zstd_errors.h +0 -60
  125. data/contrib/zstd/common/zstd_internal.h +0 -267
  126. data/contrib/zstd/compress/huf_compress.c +0 -533
  127. data/contrib/zstd/compress/zbuff_compress.c +0 -319
  128. data/contrib/zstd/compress/zstd_compress.c +0 -3264
  129. data/contrib/zstd/compress/zstd_opt.h +0 -900
  130. data/contrib/zstd/decompress/huf_decompress.c +0 -883
  131. data/contrib/zstd/decompress/zbuff_decompress.c +0 -252
  132. data/contrib/zstd/decompress/zstd_decompress.c +0 -1842
  133. data/contrib/zstd/dictBuilder/zdict.h +0 -111
  134. data/contrib/zstd/zstd.h +0 -640
@@ -0,0 +1,54 @@
1
+ /*
2
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
3
+ * All rights reserved.
4
+ *
5
+ * This source code is licensed under both the BSD-style license (found in the
6
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
7
+ * in the COPYING file in the root directory of this source tree).
8
+ * You may select, at your option, one of the above-listed licenses.
9
+ */
10
+
11
+ #ifndef ZSTD_COMPRESS_SEQUENCES_H
12
+ #define ZSTD_COMPRESS_SEQUENCES_H
13
+
14
+ #include "../common/fse.h" /* FSE_repeat, FSE_CTable */
15
+ #include "../common/zstd_internal.h" /* symbolEncodingType_e, ZSTD_strategy */
16
+
17
+ typedef enum {
18
+ ZSTD_defaultDisallowed = 0,
19
+ ZSTD_defaultAllowed = 1
20
+ } ZSTD_defaultPolicy_e;
21
+
22
+ symbolEncodingType_e
23
+ ZSTD_selectEncodingType(
24
+ FSE_repeat* repeatMode, unsigned const* count, unsigned const max,
25
+ size_t const mostFrequent, size_t nbSeq, unsigned const FSELog,
26
+ FSE_CTable const* prevCTable,
27
+ short const* defaultNorm, U32 defaultNormLog,
28
+ ZSTD_defaultPolicy_e const isDefaultAllowed,
29
+ ZSTD_strategy const strategy);
30
+
31
+ size_t
32
+ ZSTD_buildCTable(void* dst, size_t dstCapacity,
33
+ FSE_CTable* nextCTable, U32 FSELog, symbolEncodingType_e type,
34
+ unsigned* count, U32 max,
35
+ const BYTE* codeTable, size_t nbSeq,
36
+ const S16* defaultNorm, U32 defaultNormLog, U32 defaultMax,
37
+ const FSE_CTable* prevCTable, size_t prevCTableSize,
38
+ void* entropyWorkspace, size_t entropyWorkspaceSize);
39
+
40
+ size_t ZSTD_encodeSequences(
41
+ void* dst, size_t dstCapacity,
42
+ FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable,
43
+ FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable,
44
+ FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable,
45
+ seqDef const* sequences, size_t nbSeq, int longOffsets, int bmi2);
46
+
47
+ size_t ZSTD_fseBitCost(
48
+ FSE_CTable const* ctable,
49
+ unsigned const* count,
50
+ unsigned const max);
51
+
52
+ size_t ZSTD_crossEntropyCost(short const* norm, unsigned accuracyLog,
53
+ unsigned const* count, unsigned const max);
54
+ #endif /* ZSTD_COMPRESS_SEQUENCES_H */
@@ -0,0 +1,849 @@
1
+ /*
2
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
3
+ * All rights reserved.
4
+ *
5
+ * This source code is licensed under both the BSD-style license (found in the
6
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
7
+ * in the COPYING file in the root directory of this source tree).
8
+ * You may select, at your option, one of the above-listed licenses.
9
+ */
10
+
11
+ /*-*************************************
12
+ * Dependencies
13
+ ***************************************/
14
+ #include "zstd_compress_superblock.h"
15
+
16
+ #include "../common/zstd_internal.h" /* ZSTD_getSequenceLength */
17
+ #include "hist.h" /* HIST_countFast_wksp */
18
+ #include "zstd_compress_internal.h"
19
+ #include "zstd_compress_sequences.h"
20
+ #include "zstd_compress_literals.h"
21
+
22
+ /*-*************************************
23
+ * Superblock entropy buffer structs
24
+ ***************************************/
25
+ /** ZSTD_hufCTablesMetadata_t :
26
+ * Stores Literals Block Type for a super-block in hType, and
27
+ * huffman tree description in hufDesBuffer.
28
+ * hufDesSize refers to the size of huffman tree description in bytes.
29
+ * This metadata is populated in ZSTD_buildSuperBlockEntropy_literal() */
30
+ typedef struct {
31
+ symbolEncodingType_e hType;
32
+ BYTE hufDesBuffer[ZSTD_MAX_HUF_HEADER_SIZE];
33
+ size_t hufDesSize;
34
+ } ZSTD_hufCTablesMetadata_t;
35
+
36
+ /** ZSTD_fseCTablesMetadata_t :
37
+ * Stores symbol compression modes for a super-block in {ll, ol, ml}Type, and
38
+ * fse tables in fseTablesBuffer.
39
+ * fseTablesSize refers to the size of fse tables in bytes.
40
+ * This metadata is populated in ZSTD_buildSuperBlockEntropy_sequences() */
41
+ typedef struct {
42
+ symbolEncodingType_e llType;
43
+ symbolEncodingType_e ofType;
44
+ symbolEncodingType_e mlType;
45
+ BYTE fseTablesBuffer[ZSTD_MAX_FSE_HEADERS_SIZE];
46
+ size_t fseTablesSize;
47
+ size_t lastCountSize; /* This is to account for bug in 1.3.4. More detail in ZSTD_compressSubBlock_sequences() */
48
+ } ZSTD_fseCTablesMetadata_t;
49
+
50
+ typedef struct {
51
+ ZSTD_hufCTablesMetadata_t hufMetadata;
52
+ ZSTD_fseCTablesMetadata_t fseMetadata;
53
+ } ZSTD_entropyCTablesMetadata_t;
54
+
55
+
56
+ /** ZSTD_buildSuperBlockEntropy_literal() :
57
+ * Builds entropy for the super-block literals.
58
+ * Stores literals block type (raw, rle, compressed, repeat) and
59
+ * huffman description table to hufMetadata.
60
+ * @return : size of huffman description table or error code */
61
+ static size_t ZSTD_buildSuperBlockEntropy_literal(void* const src, size_t srcSize,
62
+ const ZSTD_hufCTables_t* prevHuf,
63
+ ZSTD_hufCTables_t* nextHuf,
64
+ ZSTD_hufCTablesMetadata_t* hufMetadata,
65
+ const int disableLiteralsCompression,
66
+ void* workspace, size_t wkspSize)
67
+ {
68
+ BYTE* const wkspStart = (BYTE*)workspace;
69
+ BYTE* const wkspEnd = wkspStart + wkspSize;
70
+ BYTE* const countWkspStart = wkspStart;
71
+ unsigned* const countWksp = (unsigned*)workspace;
72
+ const size_t countWkspSize = (HUF_SYMBOLVALUE_MAX + 1) * sizeof(unsigned);
73
+ BYTE* const nodeWksp = countWkspStart + countWkspSize;
74
+ const size_t nodeWkspSize = wkspEnd-nodeWksp;
75
+ unsigned maxSymbolValue = 255;
76
+ unsigned huffLog = HUF_TABLELOG_DEFAULT;
77
+ HUF_repeat repeat = prevHuf->repeatMode;
78
+
79
+ DEBUGLOG(5, "ZSTD_buildSuperBlockEntropy_literal (srcSize=%zu)", srcSize);
80
+
81
+ /* Prepare nextEntropy assuming reusing the existing table */
82
+ ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
83
+
84
+ if (disableLiteralsCompression) {
85
+ DEBUGLOG(5, "set_basic - disabled");
86
+ hufMetadata->hType = set_basic;
87
+ return 0;
88
+ }
89
+
90
+ /* small ? don't even attempt compression (speed opt) */
91
+ # define COMPRESS_LITERALS_SIZE_MIN 63
92
+ { size_t const minLitSize = (prevHuf->repeatMode == HUF_repeat_valid) ? 6 : COMPRESS_LITERALS_SIZE_MIN;
93
+ if (srcSize <= minLitSize) {
94
+ DEBUGLOG(5, "set_basic - too small");
95
+ hufMetadata->hType = set_basic;
96
+ return 0;
97
+ }
98
+ }
99
+
100
+ /* Scan input and build symbol stats */
101
+ { size_t const largest = HIST_count_wksp (countWksp, &maxSymbolValue, (const BYTE*)src, srcSize, workspace, wkspSize);
102
+ FORWARD_IF_ERROR(largest, "HIST_count_wksp failed");
103
+ if (largest == srcSize) {
104
+ DEBUGLOG(5, "set_rle");
105
+ hufMetadata->hType = set_rle;
106
+ return 0;
107
+ }
108
+ if (largest <= (srcSize >> 7)+4) {
109
+ DEBUGLOG(5, "set_basic - no gain");
110
+ hufMetadata->hType = set_basic;
111
+ return 0;
112
+ }
113
+ }
114
+
115
+ /* Validate the previous Huffman table */
116
+ if (repeat == HUF_repeat_check && !HUF_validateCTable((HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue)) {
117
+ repeat = HUF_repeat_none;
118
+ }
119
+
120
+ /* Build Huffman Tree */
121
+ ZSTD_memset(nextHuf->CTable, 0, sizeof(nextHuf->CTable));
122
+ huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
123
+ { size_t const maxBits = HUF_buildCTable_wksp((HUF_CElt*)nextHuf->CTable, countWksp,
124
+ maxSymbolValue, huffLog,
125
+ nodeWksp, nodeWkspSize);
126
+ FORWARD_IF_ERROR(maxBits, "HUF_buildCTable_wksp");
127
+ huffLog = (U32)maxBits;
128
+ { /* Build and write the CTable */
129
+ size_t const newCSize = HUF_estimateCompressedSize(
130
+ (HUF_CElt*)nextHuf->CTable, countWksp, maxSymbolValue);
131
+ size_t const hSize = HUF_writeCTable(
132
+ hufMetadata->hufDesBuffer, sizeof(hufMetadata->hufDesBuffer),
133
+ (HUF_CElt*)nextHuf->CTable, maxSymbolValue, huffLog);
134
+ /* Check against repeating the previous CTable */
135
+ if (repeat != HUF_repeat_none) {
136
+ size_t const oldCSize = HUF_estimateCompressedSize(
137
+ (HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue);
138
+ if (oldCSize < srcSize && (oldCSize <= hSize + newCSize || hSize + 12 >= srcSize)) {
139
+ DEBUGLOG(5, "set_repeat - smaller");
140
+ ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
141
+ hufMetadata->hType = set_repeat;
142
+ return 0;
143
+ }
144
+ }
145
+ if (newCSize + hSize >= srcSize) {
146
+ DEBUGLOG(5, "set_basic - no gains");
147
+ ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
148
+ hufMetadata->hType = set_basic;
149
+ return 0;
150
+ }
151
+ DEBUGLOG(5, "set_compressed (hSize=%u)", (U32)hSize);
152
+ hufMetadata->hType = set_compressed;
153
+ nextHuf->repeatMode = HUF_repeat_check;
154
+ return hSize;
155
+ }
156
+ }
157
+ }
158
+
159
+ /** ZSTD_buildSuperBlockEntropy_sequences() :
160
+ * Builds entropy for the super-block sequences.
161
+ * Stores symbol compression modes and fse table to fseMetadata.
162
+ * @return : size of fse tables or error code */
163
+ static size_t ZSTD_buildSuperBlockEntropy_sequences(seqStore_t* seqStorePtr,
164
+ const ZSTD_fseCTables_t* prevEntropy,
165
+ ZSTD_fseCTables_t* nextEntropy,
166
+ const ZSTD_CCtx_params* cctxParams,
167
+ ZSTD_fseCTablesMetadata_t* fseMetadata,
168
+ void* workspace, size_t wkspSize)
169
+ {
170
+ BYTE* const wkspStart = (BYTE*)workspace;
171
+ BYTE* const wkspEnd = wkspStart + wkspSize;
172
+ BYTE* const countWkspStart = wkspStart;
173
+ unsigned* const countWksp = (unsigned*)workspace;
174
+ const size_t countWkspSize = (MaxSeq + 1) * sizeof(unsigned);
175
+ BYTE* const cTableWksp = countWkspStart + countWkspSize;
176
+ const size_t cTableWkspSize = wkspEnd-cTableWksp;
177
+ ZSTD_strategy const strategy = cctxParams->cParams.strategy;
178
+ FSE_CTable* CTable_LitLength = nextEntropy->litlengthCTable;
179
+ FSE_CTable* CTable_OffsetBits = nextEntropy->offcodeCTable;
180
+ FSE_CTable* CTable_MatchLength = nextEntropy->matchlengthCTable;
181
+ const BYTE* const ofCodeTable = seqStorePtr->ofCode;
182
+ const BYTE* const llCodeTable = seqStorePtr->llCode;
183
+ const BYTE* const mlCodeTable = seqStorePtr->mlCode;
184
+ size_t const nbSeq = seqStorePtr->sequences - seqStorePtr->sequencesStart;
185
+ BYTE* const ostart = fseMetadata->fseTablesBuffer;
186
+ BYTE* const oend = ostart + sizeof(fseMetadata->fseTablesBuffer);
187
+ BYTE* op = ostart;
188
+
189
+ assert(cTableWkspSize >= (1 << MaxFSELog) * sizeof(FSE_FUNCTION_TYPE));
190
+ DEBUGLOG(5, "ZSTD_buildSuperBlockEntropy_sequences (nbSeq=%zu)", nbSeq);
191
+ ZSTD_memset(workspace, 0, wkspSize);
192
+
193
+ fseMetadata->lastCountSize = 0;
194
+ /* convert length/distances into codes */
195
+ ZSTD_seqToCodes(seqStorePtr);
196
+ /* build CTable for Literal Lengths */
197
+ { U32 LLtype;
198
+ unsigned max = MaxLL;
199
+ size_t const mostFrequent = HIST_countFast_wksp(countWksp, &max, llCodeTable, nbSeq, workspace, wkspSize); /* can't fail */
200
+ DEBUGLOG(5, "Building LL table");
201
+ nextEntropy->litlength_repeatMode = prevEntropy->litlength_repeatMode;
202
+ LLtype = ZSTD_selectEncodingType(&nextEntropy->litlength_repeatMode,
203
+ countWksp, max, mostFrequent, nbSeq,
204
+ LLFSELog, prevEntropy->litlengthCTable,
205
+ LL_defaultNorm, LL_defaultNormLog,
206
+ ZSTD_defaultAllowed, strategy);
207
+ assert(set_basic < set_compressed && set_rle < set_compressed);
208
+ assert(!(LLtype < set_compressed && nextEntropy->litlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
209
+ { size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_LitLength, LLFSELog, (symbolEncodingType_e)LLtype,
210
+ countWksp, max, llCodeTable, nbSeq, LL_defaultNorm, LL_defaultNormLog, MaxLL,
211
+ prevEntropy->litlengthCTable, sizeof(prevEntropy->litlengthCTable),
212
+ cTableWksp, cTableWkspSize);
213
+ FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for LitLens failed");
214
+ if (LLtype == set_compressed)
215
+ fseMetadata->lastCountSize = countSize;
216
+ op += countSize;
217
+ fseMetadata->llType = (symbolEncodingType_e) LLtype;
218
+ } }
219
+ /* build CTable for Offsets */
220
+ { U32 Offtype;
221
+ unsigned max = MaxOff;
222
+ size_t const mostFrequent = HIST_countFast_wksp(countWksp, &max, ofCodeTable, nbSeq, workspace, wkspSize); /* can't fail */
223
+ /* We can only use the basic table if max <= DefaultMaxOff, otherwise the offsets are too large */
224
+ ZSTD_defaultPolicy_e const defaultPolicy = (max <= DefaultMaxOff) ? ZSTD_defaultAllowed : ZSTD_defaultDisallowed;
225
+ DEBUGLOG(5, "Building OF table");
226
+ nextEntropy->offcode_repeatMode = prevEntropy->offcode_repeatMode;
227
+ Offtype = ZSTD_selectEncodingType(&nextEntropy->offcode_repeatMode,
228
+ countWksp, max, mostFrequent, nbSeq,
229
+ OffFSELog, prevEntropy->offcodeCTable,
230
+ OF_defaultNorm, OF_defaultNormLog,
231
+ defaultPolicy, strategy);
232
+ assert(!(Offtype < set_compressed && nextEntropy->offcode_repeatMode != FSE_repeat_none)); /* We don't copy tables */
233
+ { size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)Offtype,
234
+ countWksp, max, ofCodeTable, nbSeq, OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
235
+ prevEntropy->offcodeCTable, sizeof(prevEntropy->offcodeCTable),
236
+ cTableWksp, cTableWkspSize);
237
+ FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for Offsets failed");
238
+ if (Offtype == set_compressed)
239
+ fseMetadata->lastCountSize = countSize;
240
+ op += countSize;
241
+ fseMetadata->ofType = (symbolEncodingType_e) Offtype;
242
+ } }
243
+ /* build CTable for MatchLengths */
244
+ { U32 MLtype;
245
+ unsigned max = MaxML;
246
+ size_t const mostFrequent = HIST_countFast_wksp(countWksp, &max, mlCodeTable, nbSeq, workspace, wkspSize); /* can't fail */
247
+ DEBUGLOG(5, "Building ML table (remaining space : %i)", (int)(oend-op));
248
+ nextEntropy->matchlength_repeatMode = prevEntropy->matchlength_repeatMode;
249
+ MLtype = ZSTD_selectEncodingType(&nextEntropy->matchlength_repeatMode,
250
+ countWksp, max, mostFrequent, nbSeq,
251
+ MLFSELog, prevEntropy->matchlengthCTable,
252
+ ML_defaultNorm, ML_defaultNormLog,
253
+ ZSTD_defaultAllowed, strategy);
254
+ assert(!(MLtype < set_compressed && nextEntropy->matchlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
255
+ { size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_MatchLength, MLFSELog, (symbolEncodingType_e)MLtype,
256
+ countWksp, max, mlCodeTable, nbSeq, ML_defaultNorm, ML_defaultNormLog, MaxML,
257
+ prevEntropy->matchlengthCTable, sizeof(prevEntropy->matchlengthCTable),
258
+ cTableWksp, cTableWkspSize);
259
+ FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for MatchLengths failed");
260
+ if (MLtype == set_compressed)
261
+ fseMetadata->lastCountSize = countSize;
262
+ op += countSize;
263
+ fseMetadata->mlType = (symbolEncodingType_e) MLtype;
264
+ } }
265
+ assert((size_t) (op-ostart) <= sizeof(fseMetadata->fseTablesBuffer));
266
+ return op-ostart;
267
+ }
268
+
269
+
270
+ /** ZSTD_buildSuperBlockEntropy() :
271
+ * Builds entropy for the super-block.
272
+ * @return : 0 on success or error code */
273
+ static size_t
274
+ ZSTD_buildSuperBlockEntropy(seqStore_t* seqStorePtr,
275
+ const ZSTD_entropyCTables_t* prevEntropy,
276
+ ZSTD_entropyCTables_t* nextEntropy,
277
+ const ZSTD_CCtx_params* cctxParams,
278
+ ZSTD_entropyCTablesMetadata_t* entropyMetadata,
279
+ void* workspace, size_t wkspSize)
280
+ {
281
+ size_t const litSize = seqStorePtr->lit - seqStorePtr->litStart;
282
+ DEBUGLOG(5, "ZSTD_buildSuperBlockEntropy");
283
+ entropyMetadata->hufMetadata.hufDesSize =
284
+ ZSTD_buildSuperBlockEntropy_literal(seqStorePtr->litStart, litSize,
285
+ &prevEntropy->huf, &nextEntropy->huf,
286
+ &entropyMetadata->hufMetadata,
287
+ ZSTD_disableLiteralsCompression(cctxParams),
288
+ workspace, wkspSize);
289
+ FORWARD_IF_ERROR(entropyMetadata->hufMetadata.hufDesSize, "ZSTD_buildSuperBlockEntropy_literal failed");
290
+ entropyMetadata->fseMetadata.fseTablesSize =
291
+ ZSTD_buildSuperBlockEntropy_sequences(seqStorePtr,
292
+ &prevEntropy->fse, &nextEntropy->fse,
293
+ cctxParams,
294
+ &entropyMetadata->fseMetadata,
295
+ workspace, wkspSize);
296
+ FORWARD_IF_ERROR(entropyMetadata->fseMetadata.fseTablesSize, "ZSTD_buildSuperBlockEntropy_sequences failed");
297
+ return 0;
298
+ }
299
+
300
+ /** ZSTD_compressSubBlock_literal() :
301
+ * Compresses literals section for a sub-block.
302
+ * When we have to write the Huffman table we will sometimes choose a header
303
+ * size larger than necessary. This is because we have to pick the header size
304
+ * before we know the table size + compressed size, so we have a bound on the
305
+ * table size. If we guessed incorrectly, we fall back to uncompressed literals.
306
+ *
307
+ * We write the header when writeEntropy=1 and set entropyWrriten=1 when we succeeded
308
+ * in writing the header, otherwise it is set to 0.
309
+ *
310
+ * hufMetadata->hType has literals block type info.
311
+ * If it is set_basic, all sub-blocks literals section will be Raw_Literals_Block.
312
+ * If it is set_rle, all sub-blocks literals section will be RLE_Literals_Block.
313
+ * If it is set_compressed, first sub-block's literals section will be Compressed_Literals_Block
314
+ * If it is set_compressed, first sub-block's literals section will be Treeless_Literals_Block
315
+ * and the following sub-blocks' literals sections will be Treeless_Literals_Block.
316
+ * @return : compressed size of literals section of a sub-block
317
+ * Or 0 if it unable to compress.
318
+ * Or error code */
319
+ static size_t ZSTD_compressSubBlock_literal(const HUF_CElt* hufTable,
320
+ const ZSTD_hufCTablesMetadata_t* hufMetadata,
321
+ const BYTE* literals, size_t litSize,
322
+ void* dst, size_t dstSize,
323
+ const int bmi2, int writeEntropy, int* entropyWritten)
324
+ {
325
+ size_t const header = writeEntropy ? 200 : 0;
326
+ size_t const lhSize = 3 + (litSize >= (1 KB - header)) + (litSize >= (16 KB - header));
327
+ BYTE* const ostart = (BYTE*)dst;
328
+ BYTE* const oend = ostart + dstSize;
329
+ BYTE* op = ostart + lhSize;
330
+ U32 const singleStream = lhSize == 3;
331
+ symbolEncodingType_e hType = writeEntropy ? hufMetadata->hType : set_repeat;
332
+ size_t cLitSize = 0;
333
+
334
+ (void)bmi2; /* TODO bmi2... */
335
+
336
+ DEBUGLOG(5, "ZSTD_compressSubBlock_literal (litSize=%zu, lhSize=%zu, writeEntropy=%d)", litSize, lhSize, writeEntropy);
337
+
338
+ *entropyWritten = 0;
339
+ if (litSize == 0 || hufMetadata->hType == set_basic) {
340
+ DEBUGLOG(5, "ZSTD_compressSubBlock_literal using raw literal");
341
+ return ZSTD_noCompressLiterals(dst, dstSize, literals, litSize);
342
+ } else if (hufMetadata->hType == set_rle) {
343
+ DEBUGLOG(5, "ZSTD_compressSubBlock_literal using rle literal");
344
+ return ZSTD_compressRleLiteralsBlock(dst, dstSize, literals, litSize);
345
+ }
346
+
347
+ assert(litSize > 0);
348
+ assert(hufMetadata->hType == set_compressed || hufMetadata->hType == set_repeat);
349
+
350
+ if (writeEntropy && hufMetadata->hType == set_compressed) {
351
+ ZSTD_memcpy(op, hufMetadata->hufDesBuffer, hufMetadata->hufDesSize);
352
+ op += hufMetadata->hufDesSize;
353
+ cLitSize += hufMetadata->hufDesSize;
354
+ DEBUGLOG(5, "ZSTD_compressSubBlock_literal (hSize=%zu)", hufMetadata->hufDesSize);
355
+ }
356
+
357
+ /* TODO bmi2 */
358
+ { const size_t cSize = singleStream ? HUF_compress1X_usingCTable(op, oend-op, literals, litSize, hufTable)
359
+ : HUF_compress4X_usingCTable(op, oend-op, literals, litSize, hufTable);
360
+ op += cSize;
361
+ cLitSize += cSize;
362
+ if (cSize == 0 || ERR_isError(cSize)) {
363
+ DEBUGLOG(5, "Failed to write entropy tables %s", ZSTD_getErrorName(cSize));
364
+ return 0;
365
+ }
366
+ /* If we expand and we aren't writing a header then emit uncompressed */
367
+ if (!writeEntropy && cLitSize >= litSize) {
368
+ DEBUGLOG(5, "ZSTD_compressSubBlock_literal using raw literal because uncompressible");
369
+ return ZSTD_noCompressLiterals(dst, dstSize, literals, litSize);
370
+ }
371
+ /* If we are writing headers then allow expansion that doesn't change our header size. */
372
+ if (lhSize < (size_t)(3 + (cLitSize >= 1 KB) + (cLitSize >= 16 KB))) {
373
+ assert(cLitSize > litSize);
374
+ DEBUGLOG(5, "Literals expanded beyond allowed header size");
375
+ return ZSTD_noCompressLiterals(dst, dstSize, literals, litSize);
376
+ }
377
+ DEBUGLOG(5, "ZSTD_compressSubBlock_literal (cSize=%zu)", cSize);
378
+ }
379
+
380
+ /* Build header */
381
+ switch(lhSize)
382
+ {
383
+ case 3: /* 2 - 2 - 10 - 10 */
384
+ { U32 const lhc = hType + ((!singleStream) << 2) + ((U32)litSize<<4) + ((U32)cLitSize<<14);
385
+ MEM_writeLE24(ostart, lhc);
386
+ break;
387
+ }
388
+ case 4: /* 2 - 2 - 14 - 14 */
389
+ { U32 const lhc = hType + (2 << 2) + ((U32)litSize<<4) + ((U32)cLitSize<<18);
390
+ MEM_writeLE32(ostart, lhc);
391
+ break;
392
+ }
393
+ case 5: /* 2 - 2 - 18 - 18 */
394
+ { U32 const lhc = hType + (3 << 2) + ((U32)litSize<<4) + ((U32)cLitSize<<22);
395
+ MEM_writeLE32(ostart, lhc);
396
+ ostart[4] = (BYTE)(cLitSize >> 10);
397
+ break;
398
+ }
399
+ default: /* not possible : lhSize is {3,4,5} */
400
+ assert(0);
401
+ }
402
+ *entropyWritten = 1;
403
+ DEBUGLOG(5, "Compressed literals: %u -> %u", (U32)litSize, (U32)(op-ostart));
404
+ return op-ostart;
405
+ }
406
+
407
+ static size_t ZSTD_seqDecompressedSize(seqStore_t const* seqStore, const seqDef* sequences, size_t nbSeq, size_t litSize, int lastSequence) {
408
+ const seqDef* const sstart = sequences;
409
+ const seqDef* const send = sequences + nbSeq;
410
+ const seqDef* sp = sstart;
411
+ size_t matchLengthSum = 0;
412
+ size_t litLengthSum = 0;
413
+ while (send-sp > 0) {
414
+ ZSTD_sequenceLength const seqLen = ZSTD_getSequenceLength(seqStore, sp);
415
+ litLengthSum += seqLen.litLength;
416
+ matchLengthSum += seqLen.matchLength;
417
+ sp++;
418
+ }
419
+ assert(litLengthSum <= litSize);
420
+ if (!lastSequence) {
421
+ assert(litLengthSum == litSize);
422
+ }
423
+ return matchLengthSum + litSize;
424
+ }
425
+
426
+ /** ZSTD_compressSubBlock_sequences() :
427
+ * Compresses sequences section for a sub-block.
428
+ * fseMetadata->llType, fseMetadata->ofType, and fseMetadata->mlType have
429
+ * symbol compression modes for the super-block.
430
+ * The first successfully compressed block will have these in its header.
431
+ * We set entropyWritten=1 when we succeed in compressing the sequences.
432
+ * The following sub-blocks will always have repeat mode.
433
+ * @return : compressed size of sequences section of a sub-block
434
+ * Or 0 if it is unable to compress
435
+ * Or error code. */
436
+ static size_t ZSTD_compressSubBlock_sequences(const ZSTD_fseCTables_t* fseTables,
437
+ const ZSTD_fseCTablesMetadata_t* fseMetadata,
438
+ const seqDef* sequences, size_t nbSeq,
439
+ const BYTE* llCode, const BYTE* mlCode, const BYTE* ofCode,
440
+ const ZSTD_CCtx_params* cctxParams,
441
+ void* dst, size_t dstCapacity,
442
+ const int bmi2, int writeEntropy, int* entropyWritten)
443
+ {
444
+ const int longOffsets = cctxParams->cParams.windowLog > STREAM_ACCUMULATOR_MIN;
445
+ BYTE* const ostart = (BYTE*)dst;
446
+ BYTE* const oend = ostart + dstCapacity;
447
+ BYTE* op = ostart;
448
+ BYTE* seqHead;
449
+
450
+ DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (nbSeq=%zu, writeEntropy=%d, longOffsets=%d)", nbSeq, writeEntropy, longOffsets);
451
+
452
+ *entropyWritten = 0;
453
+ /* Sequences Header */
454
+ RETURN_ERROR_IF((oend-op) < 3 /*max nbSeq Size*/ + 1 /*seqHead*/,
455
+ dstSize_tooSmall, "");
456
+ if (nbSeq < 0x7F)
457
+ *op++ = (BYTE)nbSeq;
458
+ else if (nbSeq < LONGNBSEQ)
459
+ op[0] = (BYTE)((nbSeq>>8) + 0x80), op[1] = (BYTE)nbSeq, op+=2;
460
+ else
461
+ op[0]=0xFF, MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ)), op+=3;
462
+ if (nbSeq==0) {
463
+ return op - ostart;
464
+ }
465
+
466
+ /* seqHead : flags for FSE encoding type */
467
+ seqHead = op++;
468
+
469
+ DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (seqHeadSize=%u)", (unsigned)(op-ostart));
470
+
471
+ if (writeEntropy) {
472
+ const U32 LLtype = fseMetadata->llType;
473
+ const U32 Offtype = fseMetadata->ofType;
474
+ const U32 MLtype = fseMetadata->mlType;
475
+ DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (fseTablesSize=%zu)", fseMetadata->fseTablesSize);
476
+ *seqHead = (BYTE)((LLtype<<6) + (Offtype<<4) + (MLtype<<2));
477
+ ZSTD_memcpy(op, fseMetadata->fseTablesBuffer, fseMetadata->fseTablesSize);
478
+ op += fseMetadata->fseTablesSize;
479
+ } else {
480
+ const U32 repeat = set_repeat;
481
+ *seqHead = (BYTE)((repeat<<6) + (repeat<<4) + (repeat<<2));
482
+ }
483
+
484
+ { size_t const bitstreamSize = ZSTD_encodeSequences(
485
+ op, oend - op,
486
+ fseTables->matchlengthCTable, mlCode,
487
+ fseTables->offcodeCTable, ofCode,
488
+ fseTables->litlengthCTable, llCode,
489
+ sequences, nbSeq,
490
+ longOffsets, bmi2);
491
+ FORWARD_IF_ERROR(bitstreamSize, "ZSTD_encodeSequences failed");
492
+ op += bitstreamSize;
493
+ /* zstd versions <= 1.3.4 mistakenly report corruption when
494
+ * FSE_readNCount() receives a buffer < 4 bytes.
495
+ * Fixed by https://github.com/facebook/zstd/pull/1146.
496
+ * This can happen when the last set_compressed table present is 2
497
+ * bytes and the bitstream is only one byte.
498
+ * In this exceedingly rare case, we will simply emit an uncompressed
499
+ * block, since it isn't worth optimizing.
500
+ */
501
+ #ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
502
+ if (writeEntropy && fseMetadata->lastCountSize && fseMetadata->lastCountSize + bitstreamSize < 4) {
503
+ /* NCountSize >= 2 && bitstreamSize > 0 ==> lastCountSize == 3 */
504
+ assert(fseMetadata->lastCountSize + bitstreamSize == 3);
505
+ DEBUGLOG(5, "Avoiding bug in zstd decoder in versions <= 1.3.4 by "
506
+ "emitting an uncompressed block.");
507
+ return 0;
508
+ }
509
+ #endif
510
+ DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (bitstreamSize=%zu)", bitstreamSize);
511
+ }
512
+
513
+ /* zstd versions <= 1.4.0 mistakenly report error when
514
+ * sequences section body size is less than 3 bytes.
515
+ * Fixed by https://github.com/facebook/zstd/pull/1664.
516
+ * This can happen when the previous sequences section block is compressed
517
+ * with rle mode and the current block's sequences section is compressed
518
+ * with repeat mode where sequences section body size can be 1 byte.
519
+ */
520
+ #ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
521
+ if (op-seqHead < 4) {
522
+ DEBUGLOG(5, "Avoiding bug in zstd decoder in versions <= 1.4.0 by emitting "
523
+ "an uncompressed block when sequences are < 4 bytes");
524
+ return 0;
525
+ }
526
+ #endif
527
+
528
+ *entropyWritten = 1;
529
+ return op - ostart;
530
+ }
531
+
532
+ /** ZSTD_compressSubBlock() :
533
+ * Compresses a single sub-block.
534
+ * @return : compressed size of the sub-block
535
+ * Or 0 if it failed to compress. */
536
+ static size_t ZSTD_compressSubBlock(const ZSTD_entropyCTables_t* entropy,
537
+ const ZSTD_entropyCTablesMetadata_t* entropyMetadata,
538
+ const seqDef* sequences, size_t nbSeq,
539
+ const BYTE* literals, size_t litSize,
540
+ const BYTE* llCode, const BYTE* mlCode, const BYTE* ofCode,
541
+ const ZSTD_CCtx_params* cctxParams,
542
+ void* dst, size_t dstCapacity,
543
+ const int bmi2,
544
+ int writeLitEntropy, int writeSeqEntropy,
545
+ int* litEntropyWritten, int* seqEntropyWritten,
546
+ U32 lastBlock)
547
+ {
548
+ BYTE* const ostart = (BYTE*)dst;
549
+ BYTE* const oend = ostart + dstCapacity;
550
+ BYTE* op = ostart + ZSTD_blockHeaderSize;
551
+ DEBUGLOG(5, "ZSTD_compressSubBlock (litSize=%zu, nbSeq=%zu, writeLitEntropy=%d, writeSeqEntropy=%d, lastBlock=%d)",
552
+ litSize, nbSeq, writeLitEntropy, writeSeqEntropy, lastBlock);
553
+ { size_t cLitSize = ZSTD_compressSubBlock_literal((const HUF_CElt*)entropy->huf.CTable,
554
+ &entropyMetadata->hufMetadata, literals, litSize,
555
+ op, oend-op, bmi2, writeLitEntropy, litEntropyWritten);
556
+ FORWARD_IF_ERROR(cLitSize, "ZSTD_compressSubBlock_literal failed");
557
+ if (cLitSize == 0) return 0;
558
+ op += cLitSize;
559
+ }
560
+ { size_t cSeqSize = ZSTD_compressSubBlock_sequences(&entropy->fse,
561
+ &entropyMetadata->fseMetadata,
562
+ sequences, nbSeq,
563
+ llCode, mlCode, ofCode,
564
+ cctxParams,
565
+ op, oend-op,
566
+ bmi2, writeSeqEntropy, seqEntropyWritten);
567
+ FORWARD_IF_ERROR(cSeqSize, "ZSTD_compressSubBlock_sequences failed");
568
+ if (cSeqSize == 0) return 0;
569
+ op += cSeqSize;
570
+ }
571
+ /* Write block header */
572
+ { size_t cSize = (op-ostart)-ZSTD_blockHeaderSize;
573
+ U32 const cBlockHeader24 = lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3);
574
+ MEM_writeLE24(ostart, cBlockHeader24);
575
+ }
576
+ return op-ostart;
577
+ }
578
+
579
+ static size_t ZSTD_estimateSubBlockSize_literal(const BYTE* literals, size_t litSize,
580
+ const ZSTD_hufCTables_t* huf,
581
+ const ZSTD_hufCTablesMetadata_t* hufMetadata,
582
+ void* workspace, size_t wkspSize,
583
+ int writeEntropy)
584
+ {
585
+ unsigned* const countWksp = (unsigned*)workspace;
586
+ unsigned maxSymbolValue = 255;
587
+ size_t literalSectionHeaderSize = 3; /* Use hard coded size of 3 bytes */
588
+
589
+ if (hufMetadata->hType == set_basic) return litSize;
590
+ else if (hufMetadata->hType == set_rle) return 1;
591
+ else if (hufMetadata->hType == set_compressed || hufMetadata->hType == set_repeat) {
592
+ size_t const largest = HIST_count_wksp (countWksp, &maxSymbolValue, (const BYTE*)literals, litSize, workspace, wkspSize);
593
+ if (ZSTD_isError(largest)) return litSize;
594
+ { size_t cLitSizeEstimate = HUF_estimateCompressedSize((const HUF_CElt*)huf->CTable, countWksp, maxSymbolValue);
595
+ if (writeEntropy) cLitSizeEstimate += hufMetadata->hufDesSize;
596
+ return cLitSizeEstimate + literalSectionHeaderSize;
597
+ } }
598
+ assert(0); /* impossible */
599
+ return 0;
600
+ }
601
+
602
+ static size_t ZSTD_estimateSubBlockSize_symbolType(symbolEncodingType_e type,
603
+ const BYTE* codeTable, unsigned maxCode,
604
+ size_t nbSeq, const FSE_CTable* fseCTable,
605
+ const U32* additionalBits,
606
+ short const* defaultNorm, U32 defaultNormLog, U32 defaultMax,
607
+ void* workspace, size_t wkspSize)
608
+ {
609
+ unsigned* const countWksp = (unsigned*)workspace;
610
+ const BYTE* ctp = codeTable;
611
+ const BYTE* const ctStart = ctp;
612
+ const BYTE* const ctEnd = ctStart + nbSeq;
613
+ size_t cSymbolTypeSizeEstimateInBits = 0;
614
+ unsigned max = maxCode;
615
+
616
+ HIST_countFast_wksp(countWksp, &max, codeTable, nbSeq, workspace, wkspSize); /* can't fail */
617
+ if (type == set_basic) {
618
+ /* We selected this encoding type, so it must be valid. */
619
+ assert(max <= defaultMax);
620
+ cSymbolTypeSizeEstimateInBits = max <= defaultMax
621
+ ? ZSTD_crossEntropyCost(defaultNorm, defaultNormLog, countWksp, max)
622
+ : ERROR(GENERIC);
623
+ } else if (type == set_rle) {
624
+ cSymbolTypeSizeEstimateInBits = 0;
625
+ } else if (type == set_compressed || type == set_repeat) {
626
+ cSymbolTypeSizeEstimateInBits = ZSTD_fseBitCost(fseCTable, countWksp, max);
627
+ }
628
+ if (ZSTD_isError(cSymbolTypeSizeEstimateInBits)) return nbSeq * 10;
629
+ while (ctp < ctEnd) {
630
+ if (additionalBits) cSymbolTypeSizeEstimateInBits += additionalBits[*ctp];
631
+ else cSymbolTypeSizeEstimateInBits += *ctp; /* for offset, offset code is also the number of additional bits */
632
+ ctp++;
633
+ }
634
+ return cSymbolTypeSizeEstimateInBits / 8;
635
+ }
636
+
637
+ static size_t ZSTD_estimateSubBlockSize_sequences(const BYTE* ofCodeTable,
638
+ const BYTE* llCodeTable,
639
+ const BYTE* mlCodeTable,
640
+ size_t nbSeq,
641
+ const ZSTD_fseCTables_t* fseTables,
642
+ const ZSTD_fseCTablesMetadata_t* fseMetadata,
643
+ void* workspace, size_t wkspSize,
644
+ int writeEntropy)
645
+ {
646
+ size_t sequencesSectionHeaderSize = 3; /* Use hard coded size of 3 bytes */
647
+ size_t cSeqSizeEstimate = 0;
648
+ cSeqSizeEstimate += ZSTD_estimateSubBlockSize_symbolType(fseMetadata->ofType, ofCodeTable, MaxOff,
649
+ nbSeq, fseTables->offcodeCTable, NULL,
650
+ OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
651
+ workspace, wkspSize);
652
+ cSeqSizeEstimate += ZSTD_estimateSubBlockSize_symbolType(fseMetadata->llType, llCodeTable, MaxLL,
653
+ nbSeq, fseTables->litlengthCTable, LL_bits,
654
+ LL_defaultNorm, LL_defaultNormLog, MaxLL,
655
+ workspace, wkspSize);
656
+ cSeqSizeEstimate += ZSTD_estimateSubBlockSize_symbolType(fseMetadata->mlType, mlCodeTable, MaxML,
657
+ nbSeq, fseTables->matchlengthCTable, ML_bits,
658
+ ML_defaultNorm, ML_defaultNormLog, MaxML,
659
+ workspace, wkspSize);
660
+ if (writeEntropy) cSeqSizeEstimate += fseMetadata->fseTablesSize;
661
+ return cSeqSizeEstimate + sequencesSectionHeaderSize;
662
+ }
663
+
664
+ static size_t ZSTD_estimateSubBlockSize(const BYTE* literals, size_t litSize,
665
+ const BYTE* ofCodeTable,
666
+ const BYTE* llCodeTable,
667
+ const BYTE* mlCodeTable,
668
+ size_t nbSeq,
669
+ const ZSTD_entropyCTables_t* entropy,
670
+ const ZSTD_entropyCTablesMetadata_t* entropyMetadata,
671
+ void* workspace, size_t wkspSize,
672
+ int writeLitEntropy, int writeSeqEntropy) {
673
+ size_t cSizeEstimate = 0;
674
+ cSizeEstimate += ZSTD_estimateSubBlockSize_literal(literals, litSize,
675
+ &entropy->huf, &entropyMetadata->hufMetadata,
676
+ workspace, wkspSize, writeLitEntropy);
677
+ cSizeEstimate += ZSTD_estimateSubBlockSize_sequences(ofCodeTable, llCodeTable, mlCodeTable,
678
+ nbSeq, &entropy->fse, &entropyMetadata->fseMetadata,
679
+ workspace, wkspSize, writeSeqEntropy);
680
+ return cSizeEstimate + ZSTD_blockHeaderSize;
681
+ }
682
+
683
+ static int ZSTD_needSequenceEntropyTables(ZSTD_fseCTablesMetadata_t const* fseMetadata)
684
+ {
685
+ if (fseMetadata->llType == set_compressed || fseMetadata->llType == set_rle)
686
+ return 1;
687
+ if (fseMetadata->mlType == set_compressed || fseMetadata->mlType == set_rle)
688
+ return 1;
689
+ if (fseMetadata->ofType == set_compressed || fseMetadata->ofType == set_rle)
690
+ return 1;
691
+ return 0;
692
+ }
693
+
694
+ /** ZSTD_compressSubBlock_multi() :
695
+ * Breaks super-block into multiple sub-blocks and compresses them.
696
+ * Entropy will be written to the first block.
697
+ * The following blocks will use repeat mode to compress.
698
+ * All sub-blocks are compressed blocks (no raw or rle blocks).
699
+ * @return : compressed size of the super block (which is multiple ZSTD blocks)
700
+ * Or 0 if it failed to compress. */
701
+ static size_t ZSTD_compressSubBlock_multi(const seqStore_t* seqStorePtr,
702
+ const ZSTD_compressedBlockState_t* prevCBlock,
703
+ ZSTD_compressedBlockState_t* nextCBlock,
704
+ const ZSTD_entropyCTablesMetadata_t* entropyMetadata,
705
+ const ZSTD_CCtx_params* cctxParams,
706
+ void* dst, size_t dstCapacity,
707
+ const void* src, size_t srcSize,
708
+ const int bmi2, U32 lastBlock,
709
+ void* workspace, size_t wkspSize)
710
+ {
711
+ const seqDef* const sstart = seqStorePtr->sequencesStart;
712
+ const seqDef* const send = seqStorePtr->sequences;
713
+ const seqDef* sp = sstart;
714
+ const BYTE* const lstart = seqStorePtr->litStart;
715
+ const BYTE* const lend = seqStorePtr->lit;
716
+ const BYTE* lp = lstart;
717
+ BYTE const* ip = (BYTE const*)src;
718
+ BYTE const* const iend = ip + srcSize;
719
+ BYTE* const ostart = (BYTE*)dst;
720
+ BYTE* const oend = ostart + dstCapacity;
721
+ BYTE* op = ostart;
722
+ const BYTE* llCodePtr = seqStorePtr->llCode;
723
+ const BYTE* mlCodePtr = seqStorePtr->mlCode;
724
+ const BYTE* ofCodePtr = seqStorePtr->ofCode;
725
+ size_t targetCBlockSize = cctxParams->targetCBlockSize;
726
+ size_t litSize, seqCount;
727
+ int writeLitEntropy = entropyMetadata->hufMetadata.hType == set_compressed;
728
+ int writeSeqEntropy = 1;
729
+ int lastSequence = 0;
730
+
731
+ DEBUGLOG(5, "ZSTD_compressSubBlock_multi (litSize=%u, nbSeq=%u)",
732
+ (unsigned)(lend-lp), (unsigned)(send-sstart));
733
+
734
+ litSize = 0;
735
+ seqCount = 0;
736
+ do {
737
+ size_t cBlockSizeEstimate = 0;
738
+ if (sstart == send) {
739
+ lastSequence = 1;
740
+ } else {
741
+ const seqDef* const sequence = sp + seqCount;
742
+ lastSequence = sequence == send - 1;
743
+ litSize += ZSTD_getSequenceLength(seqStorePtr, sequence).litLength;
744
+ seqCount++;
745
+ }
746
+ if (lastSequence) {
747
+ assert(lp <= lend);
748
+ assert(litSize <= (size_t)(lend - lp));
749
+ litSize = (size_t)(lend - lp);
750
+ }
751
+ /* I think there is an optimization opportunity here.
752
+ * Calling ZSTD_estimateSubBlockSize for every sequence can be wasteful
753
+ * since it recalculates estimate from scratch.
754
+ * For example, it would recount literal distribution and symbol codes everytime.
755
+ */
756
+ cBlockSizeEstimate = ZSTD_estimateSubBlockSize(lp, litSize, ofCodePtr, llCodePtr, mlCodePtr, seqCount,
757
+ &nextCBlock->entropy, entropyMetadata,
758
+ workspace, wkspSize, writeLitEntropy, writeSeqEntropy);
759
+ if (cBlockSizeEstimate > targetCBlockSize || lastSequence) {
760
+ int litEntropyWritten = 0;
761
+ int seqEntropyWritten = 0;
762
+ const size_t decompressedSize = ZSTD_seqDecompressedSize(seqStorePtr, sp, seqCount, litSize, lastSequence);
763
+ const size_t cSize = ZSTD_compressSubBlock(&nextCBlock->entropy, entropyMetadata,
764
+ sp, seqCount,
765
+ lp, litSize,
766
+ llCodePtr, mlCodePtr, ofCodePtr,
767
+ cctxParams,
768
+ op, oend-op,
769
+ bmi2, writeLitEntropy, writeSeqEntropy,
770
+ &litEntropyWritten, &seqEntropyWritten,
771
+ lastBlock && lastSequence);
772
+ FORWARD_IF_ERROR(cSize, "ZSTD_compressSubBlock failed");
773
+ if (cSize > 0 && cSize < decompressedSize) {
774
+ DEBUGLOG(5, "Committed the sub-block");
775
+ assert(ip + decompressedSize <= iend);
776
+ ip += decompressedSize;
777
+ sp += seqCount;
778
+ lp += litSize;
779
+ op += cSize;
780
+ llCodePtr += seqCount;
781
+ mlCodePtr += seqCount;
782
+ ofCodePtr += seqCount;
783
+ litSize = 0;
784
+ seqCount = 0;
785
+ /* Entropy only needs to be written once */
786
+ if (litEntropyWritten) {
787
+ writeLitEntropy = 0;
788
+ }
789
+ if (seqEntropyWritten) {
790
+ writeSeqEntropy = 0;
791
+ }
792
+ }
793
+ }
794
+ } while (!lastSequence);
795
+ if (writeLitEntropy) {
796
+ DEBUGLOG(5, "ZSTD_compressSubBlock_multi has literal entropy tables unwritten");
797
+ ZSTD_memcpy(&nextCBlock->entropy.huf, &prevCBlock->entropy.huf, sizeof(prevCBlock->entropy.huf));
798
+ }
799
+ if (writeSeqEntropy && ZSTD_needSequenceEntropyTables(&entropyMetadata->fseMetadata)) {
800
+ /* If we haven't written our entropy tables, then we've violated our contract and
801
+ * must emit an uncompressed block.
802
+ */
803
+ DEBUGLOG(5, "ZSTD_compressSubBlock_multi has sequence entropy tables unwritten");
804
+ return 0;
805
+ }
806
+ if (ip < iend) {
807
+ size_t const cSize = ZSTD_noCompressBlock(op, oend - op, ip, iend - ip, lastBlock);
808
+ DEBUGLOG(5, "ZSTD_compressSubBlock_multi last sub-block uncompressed, %zu bytes", (size_t)(iend - ip));
809
+ FORWARD_IF_ERROR(cSize, "ZSTD_noCompressBlock failed");
810
+ assert(cSize != 0);
811
+ op += cSize;
812
+ /* We have to regenerate the repcodes because we've skipped some sequences */
813
+ if (sp < send) {
814
+ seqDef const* seq;
815
+ repcodes_t rep;
816
+ ZSTD_memcpy(&rep, prevCBlock->rep, sizeof(rep));
817
+ for (seq = sstart; seq < sp; ++seq) {
818
+ rep = ZSTD_updateRep(rep.rep, seq->offset - 1, ZSTD_getSequenceLength(seqStorePtr, seq).litLength == 0);
819
+ }
820
+ ZSTD_memcpy(nextCBlock->rep, &rep, sizeof(rep));
821
+ }
822
+ }
823
+ DEBUGLOG(5, "ZSTD_compressSubBlock_multi compressed");
824
+ return op-ostart;
825
+ }
826
+
827
+ size_t ZSTD_compressSuperBlock(ZSTD_CCtx* zc,
828
+ void* dst, size_t dstCapacity,
829
+ void const* src, size_t srcSize,
830
+ unsigned lastBlock) {
831
+ ZSTD_entropyCTablesMetadata_t entropyMetadata;
832
+
833
+ FORWARD_IF_ERROR(ZSTD_buildSuperBlockEntropy(&zc->seqStore,
834
+ &zc->blockState.prevCBlock->entropy,
835
+ &zc->blockState.nextCBlock->entropy,
836
+ &zc->appliedParams,
837
+ &entropyMetadata,
838
+ zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */), "");
839
+
840
+ return ZSTD_compressSubBlock_multi(&zc->seqStore,
841
+ zc->blockState.prevCBlock,
842
+ zc->blockState.nextCBlock,
843
+ &entropyMetadata,
844
+ &zc->appliedParams,
845
+ dst, dstCapacity,
846
+ src, srcSize,
847
+ zc->bmi2, lastBlock,
848
+ zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */);
849
+ }