zstdlib 0.10.0-x64-mingw32 → 0.12.0-x64-mingw32

Sign up to get free protection for your applications and to get access to all the features.
Files changed (110) hide show
  1. checksums.yaml +4 -4
  2. data/CHANGES.md +17 -0
  3. data/ext/zstdlib_c/extconf.rb +9 -4
  4. data/ext/zstdlib_c/ruby/zlib-3.2/zstdlib.c +5090 -0
  5. data/ext/zstdlib_c/ruby/zlib-3.3/zstdlib.c +5090 -0
  6. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/adler32.c +5 -27
  7. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/compress.c +5 -16
  8. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/crc32.c +94 -161
  9. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/deflate.c +362 -434
  10. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/deflate.h +43 -12
  11. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/gzclose.c +1 -3
  12. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/gzguts.h +13 -18
  13. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/gzlib.c +28 -85
  14. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/gzread.c +23 -73
  15. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/gzwrite.c +19 -65
  16. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/infback.c +17 -30
  17. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/inffast.c +1 -4
  18. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/inffast.h +1 -1
  19. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/inflate.c +36 -102
  20. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/inftrees.c +6 -11
  21. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/inftrees.h +6 -6
  22. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/trees.c +290 -355
  23. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/uncompr.c +4 -12
  24. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/zconf.h +23 -14
  25. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/zlib.h +202 -199
  26. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/zutil.c +18 -44
  27. data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/zutil.h +13 -33
  28. data/ext/zstdlib_c/zstd-1.5.5/lib/common/allocations.h +55 -0
  29. data/ext/zstdlib_c/zstd-1.5.5/lib/common/bits.h +200 -0
  30. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/bitstream.h +19 -60
  31. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/compiler.h +26 -3
  32. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/cpu.h +1 -1
  33. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/debug.c +1 -1
  34. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/debug.h +1 -1
  35. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/entropy_common.c +12 -40
  36. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/error_private.c +9 -2
  37. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/error_private.h +1 -1
  38. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/fse.h +5 -83
  39. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/fse_decompress.c +7 -99
  40. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/huf.h +65 -156
  41. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/mem.h +39 -46
  42. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/pool.c +26 -10
  43. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/pool.h +7 -1
  44. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/portability_macros.h +22 -3
  45. data/ext/zstdlib_c/zstd-1.5.5/lib/common/threading.c +176 -0
  46. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/threading.h +5 -10
  47. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/xxhash.c +2 -2
  48. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/xxhash.h +8 -8
  49. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/zstd_common.c +1 -36
  50. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/zstd_deps.h +1 -1
  51. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/zstd_internal.h +17 -118
  52. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/common/zstd_trace.h +3 -3
  53. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/clevels.h +1 -1
  54. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/fse_compress.c +7 -124
  55. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/hist.c +1 -1
  56. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/hist.h +1 -1
  57. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/huf_compress.c +234 -169
  58. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_compress.c +1243 -538
  59. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_compress_internal.h +225 -151
  60. data/ext/zstdlib_c/zstd-1.5.5/lib/compress/zstd_compress_literals.c +235 -0
  61. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_compress_literals.h +16 -8
  62. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_compress_sequences.c +3 -3
  63. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_compress_sequences.h +1 -1
  64. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_compress_superblock.c +25 -21
  65. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_compress_superblock.h +1 -1
  66. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_cwksp.h +128 -62
  67. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_double_fast.c +95 -33
  68. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_double_fast.h +3 -2
  69. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_fast.c +433 -148
  70. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_fast.h +3 -2
  71. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_lazy.c +398 -345
  72. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_lazy.h +4 -2
  73. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_ldm.c +5 -5
  74. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_ldm.h +1 -1
  75. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_ldm_geartab.h +1 -1
  76. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_opt.c +106 -80
  77. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstd_opt.h +1 -1
  78. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstdmt_compress.c +17 -9
  79. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/compress/zstdmt_compress.h +1 -1
  80. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/decompress/huf_decompress.c +434 -441
  81. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/decompress/huf_decompress_amd64.S +30 -39
  82. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/decompress/zstd_ddict.c +4 -4
  83. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/decompress/zstd_ddict.h +1 -1
  84. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/decompress/zstd_decompress.c +205 -80
  85. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/decompress/zstd_decompress_block.c +201 -81
  86. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/decompress/zstd_decompress_block.h +6 -1
  87. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/decompress/zstd_decompress_internal.h +4 -2
  88. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/zdict.h +53 -31
  89. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/zstd.h +580 -135
  90. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/lib/zstd_errors.h +27 -8
  91. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/zlibWrapper/gzclose.c +1 -1
  92. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/zlibWrapper/gzcompatibility.h +8 -8
  93. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/zlibWrapper/gzguts.h +10 -10
  94. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/zlibWrapper/gzlib.c +3 -3
  95. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/zlibWrapper/gzread.c +10 -10
  96. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/zlibWrapper/gzwrite.c +5 -5
  97. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/zlibWrapper/zstd_zlibwrapper.c +46 -44
  98. data/ext/zstdlib_c/{zstd-1.5.2 → zstd-1.5.5}/zlibWrapper/zstd_zlibwrapper.h +4 -1
  99. data/lib/2.4/zstdlib_c.so +0 -0
  100. data/lib/2.5/zstdlib_c.so +0 -0
  101. data/lib/2.6/zstdlib_c.so +0 -0
  102. data/lib/2.7/zstdlib_c.so +0 -0
  103. data/lib/3.0/zstdlib_c.so +0 -0
  104. metadata +108 -104
  105. data/ext/zstdlib_c/zstd-1.5.2/lib/common/threading.c +0 -122
  106. data/ext/zstdlib_c/zstd-1.5.2/lib/compress/zstd_compress_literals.c +0 -159
  107. /data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/crc32.h +0 -0
  108. /data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/inffixed.h +0 -0
  109. /data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/inflate.h +0 -0
  110. /data/ext/zstdlib_c/{zlib-1.2.12 → zlib-1.3.1}/trees.h +0 -0
@@ -1,6 +1,6 @@
1
1
  /* ******************************************************************
2
2
  * Huffman encoder, part of New Generation Entropy library
3
- * Copyright (c) Yann Collet, Facebook, Inc.
3
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
4
4
  *
5
5
  * You can contact the author at :
6
6
  * - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
@@ -29,9 +29,9 @@
29
29
  #include "hist.h"
30
30
  #define FSE_STATIC_LINKING_ONLY /* FSE_optimalTableLog_internal */
31
31
  #include "../common/fse.h" /* header compression */
32
- #define HUF_STATIC_LINKING_ONLY
33
32
  #include "../common/huf.h"
34
33
  #include "../common/error_private.h"
34
+ #include "../common/bits.h" /* ZSTD_highbit32 */
35
35
 
36
36
 
37
37
  /* **************************************************************
@@ -42,13 +42,67 @@
42
42
 
43
43
 
44
44
  /* **************************************************************
45
- * Utils
45
+ * Required declarations
46
46
  ****************************************************************/
47
- unsigned HUF_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue)
47
+ typedef struct nodeElt_s {
48
+ U32 count;
49
+ U16 parent;
50
+ BYTE byte;
51
+ BYTE nbBits;
52
+ } nodeElt;
53
+
54
+
55
+ /* **************************************************************
56
+ * Debug Traces
57
+ ****************************************************************/
58
+
59
+ #if DEBUGLEVEL >= 2
60
+
61
+ static size_t showU32(const U32* arr, size_t size)
48
62
  {
49
- return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 1);
63
+ size_t u;
64
+ for (u=0; u<size; u++) {
65
+ RAWLOG(6, " %u", arr[u]); (void)arr;
66
+ }
67
+ RAWLOG(6, " \n");
68
+ return size;
50
69
  }
51
70
 
71
+ static size_t HUF_getNbBits(HUF_CElt elt);
72
+
73
+ static size_t showCTableBits(const HUF_CElt* ctable, size_t size)
74
+ {
75
+ size_t u;
76
+ for (u=0; u<size; u++) {
77
+ RAWLOG(6, " %zu", HUF_getNbBits(ctable[u])); (void)ctable;
78
+ }
79
+ RAWLOG(6, " \n");
80
+ return size;
81
+
82
+ }
83
+
84
+ static size_t showHNodeSymbols(const nodeElt* hnode, size_t size)
85
+ {
86
+ size_t u;
87
+ for (u=0; u<size; u++) {
88
+ RAWLOG(6, " %u", hnode[u].byte); (void)hnode;
89
+ }
90
+ RAWLOG(6, " \n");
91
+ return size;
92
+ }
93
+
94
+ static size_t showHNodeBits(const nodeElt* hnode, size_t size)
95
+ {
96
+ size_t u;
97
+ for (u=0; u<size; u++) {
98
+ RAWLOG(6, " %u", hnode[u].nbBits); (void)hnode;
99
+ }
100
+ RAWLOG(6, " \n");
101
+ return size;
102
+ }
103
+
104
+ #endif
105
+
52
106
 
53
107
  /* *******************************************************
54
108
  * HUF : Huffman block compression
@@ -89,7 +143,10 @@ typedef struct {
89
143
  S16 norm[HUF_TABLELOG_MAX+1];
90
144
  } HUF_CompressWeightsWksp;
91
145
 
92
- static size_t HUF_compressWeights(void* dst, size_t dstSize, const void* weightTable, size_t wtSize, void* workspace, size_t workspaceSize)
146
+ static size_t
147
+ HUF_compressWeights(void* dst, size_t dstSize,
148
+ const void* weightTable, size_t wtSize,
149
+ void* workspace, size_t workspaceSize)
93
150
  {
94
151
  BYTE* const ostart = (BYTE*) dst;
95
152
  BYTE* op = ostart;
@@ -140,7 +197,7 @@ static size_t HUF_getNbBitsFast(HUF_CElt elt)
140
197
 
141
198
  static size_t HUF_getValue(HUF_CElt elt)
142
199
  {
143
- return elt & ~0xFF;
200
+ return elt & ~(size_t)0xFF;
144
201
  }
145
202
 
146
203
  static size_t HUF_getValueFast(HUF_CElt elt)
@@ -178,6 +235,8 @@ size_t HUF_writeCTable_wksp(void* dst, size_t maxDstSize,
178
235
  U32 n;
179
236
  HUF_WriteCTableWksp* wksp = (HUF_WriteCTableWksp*)HUF_alignUpWorkspace(workspace, &workspaceSize, ZSTD_ALIGNOF(U32));
180
237
 
238
+ HUF_STATIC_ASSERT(HUF_CTABLE_WORKSPACE_SIZE >= sizeof(HUF_WriteCTableWksp));
239
+
181
240
  /* check conditions */
182
241
  if (workspaceSize < sizeof(HUF_WriteCTableWksp)) return ERROR(GENERIC);
183
242
  if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge);
@@ -207,16 +266,6 @@ size_t HUF_writeCTable_wksp(void* dst, size_t maxDstSize,
207
266
  return ((maxSymbolValue+1)/2) + 1;
208
267
  }
209
268
 
210
- /*! HUF_writeCTable() :
211
- `CTable` : Huffman tree to save, using huf representation.
212
- @return : size of saved CTable */
213
- size_t HUF_writeCTable (void* dst, size_t maxDstSize,
214
- const HUF_CElt* CTable, unsigned maxSymbolValue, unsigned huffLog)
215
- {
216
- HUF_WriteCTableWksp wksp;
217
- return HUF_writeCTable_wksp(dst, maxDstSize, CTable, maxSymbolValue, huffLog, &wksp, sizeof(wksp));
218
- }
219
-
220
269
 
221
270
  size_t HUF_readCTable (HUF_CElt* CTable, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize, unsigned* hasZeroWeights)
222
271
  {
@@ -272,68 +321,64 @@ size_t HUF_readCTable (HUF_CElt* CTable, unsigned* maxSymbolValuePtr, const void
272
321
 
273
322
  U32 HUF_getNbBitsFromCTable(HUF_CElt const* CTable, U32 symbolValue)
274
323
  {
275
- const HUF_CElt* ct = CTable + 1;
324
+ const HUF_CElt* const ct = CTable + 1;
276
325
  assert(symbolValue <= HUF_SYMBOLVALUE_MAX);
277
326
  return (U32)HUF_getNbBits(ct[symbolValue]);
278
327
  }
279
328
 
280
329
 
281
- typedef struct nodeElt_s {
282
- U32 count;
283
- U16 parent;
284
- BYTE byte;
285
- BYTE nbBits;
286
- } nodeElt;
287
-
288
330
  /**
289
331
  * HUF_setMaxHeight():
290
- * Enforces maxNbBits on the Huffman tree described in huffNode.
332
+ * Try to enforce @targetNbBits on the Huffman tree described in @huffNode.
291
333
  *
292
- * It sets all nodes with nbBits > maxNbBits to be maxNbBits. Then it adjusts
293
- * the tree to so that it is a valid canonical Huffman tree.
334
+ * It attempts to convert all nodes with nbBits > @targetNbBits
335
+ * to employ @targetNbBits instead. Then it adjusts the tree
336
+ * so that it remains a valid canonical Huffman tree.
294
337
  *
295
338
  * @pre The sum of the ranks of each symbol == 2^largestBits,
296
339
  * where largestBits == huffNode[lastNonNull].nbBits.
297
340
  * @post The sum of the ranks of each symbol == 2^largestBits,
298
- * where largestBits is the return value <= maxNbBits.
341
+ * where largestBits is the return value (expected <= targetNbBits).
299
342
  *
300
- * @param huffNode The Huffman tree modified in place to enforce maxNbBits.
343
+ * @param huffNode The Huffman tree modified in place to enforce targetNbBits.
344
+ * It's presumed sorted, from most frequent to rarest symbol.
301
345
  * @param lastNonNull The symbol with the lowest count in the Huffman tree.
302
- * @param maxNbBits The maximum allowed number of bits, which the Huffman tree
346
+ * @param targetNbBits The allowed number of bits, which the Huffman tree
303
347
  * may not respect. After this function the Huffman tree will
304
- * respect maxNbBits.
305
- * @return The maximum number of bits of the Huffman tree after adjustment,
306
- * necessarily no more than maxNbBits.
348
+ * respect targetNbBits.
349
+ * @return The maximum number of bits of the Huffman tree after adjustment.
307
350
  */
308
- static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
351
+ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 targetNbBits)
309
352
  {
310
353
  const U32 largestBits = huffNode[lastNonNull].nbBits;
311
- /* early exit : no elt > maxNbBits, so the tree is already valid. */
312
- if (largestBits <= maxNbBits) return largestBits;
354
+ /* early exit : no elt > targetNbBits, so the tree is already valid. */
355
+ if (largestBits <= targetNbBits) return largestBits;
356
+
357
+ DEBUGLOG(5, "HUF_setMaxHeight (targetNbBits = %u)", targetNbBits);
313
358
 
314
359
  /* there are several too large elements (at least >= 2) */
315
360
  { int totalCost = 0;
316
- const U32 baseCost = 1 << (largestBits - maxNbBits);
361
+ const U32 baseCost = 1 << (largestBits - targetNbBits);
317
362
  int n = (int)lastNonNull;
318
363
 
319
- /* Adjust any ranks > maxNbBits to maxNbBits.
364
+ /* Adjust any ranks > targetNbBits to targetNbBits.
320
365
  * Compute totalCost, which is how far the sum of the ranks is
321
366
  * we are over 2^largestBits after adjust the offending ranks.
322
367
  */
323
- while (huffNode[n].nbBits > maxNbBits) {
368
+ while (huffNode[n].nbBits > targetNbBits) {
324
369
  totalCost += baseCost - (1 << (largestBits - huffNode[n].nbBits));
325
- huffNode[n].nbBits = (BYTE)maxNbBits;
370
+ huffNode[n].nbBits = (BYTE)targetNbBits;
326
371
  n--;
327
372
  }
328
- /* n stops at huffNode[n].nbBits <= maxNbBits */
329
- assert(huffNode[n].nbBits <= maxNbBits);
330
- /* n end at index of smallest symbol using < maxNbBits */
331
- while (huffNode[n].nbBits == maxNbBits) --n;
373
+ /* n stops at huffNode[n].nbBits <= targetNbBits */
374
+ assert(huffNode[n].nbBits <= targetNbBits);
375
+ /* n end at index of smallest symbol using < targetNbBits */
376
+ while (huffNode[n].nbBits == targetNbBits) --n;
332
377
 
333
- /* renorm totalCost from 2^largestBits to 2^maxNbBits
378
+ /* renorm totalCost from 2^largestBits to 2^targetNbBits
334
379
  * note : totalCost is necessarily a multiple of baseCost */
335
- assert((totalCost & (baseCost - 1)) == 0);
336
- totalCost >>= (largestBits - maxNbBits);
380
+ assert(((U32)totalCost & (baseCost - 1)) == 0);
381
+ totalCost >>= (largestBits - targetNbBits);
337
382
  assert(totalCost > 0);
338
383
 
339
384
  /* repay normalized cost */
@@ -342,19 +387,19 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
342
387
 
343
388
  /* Get pos of last (smallest = lowest cum. count) symbol per rank */
344
389
  ZSTD_memset(rankLast, 0xF0, sizeof(rankLast));
345
- { U32 currentNbBits = maxNbBits;
390
+ { U32 currentNbBits = targetNbBits;
346
391
  int pos;
347
392
  for (pos=n ; pos >= 0; pos--) {
348
393
  if (huffNode[pos].nbBits >= currentNbBits) continue;
349
- currentNbBits = huffNode[pos].nbBits; /* < maxNbBits */
350
- rankLast[maxNbBits-currentNbBits] = (U32)pos;
394
+ currentNbBits = huffNode[pos].nbBits; /* < targetNbBits */
395
+ rankLast[targetNbBits-currentNbBits] = (U32)pos;
351
396
  } }
352
397
 
353
398
  while (totalCost > 0) {
354
399
  /* Try to reduce the next power of 2 above totalCost because we
355
400
  * gain back half the rank.
356
401
  */
357
- U32 nBitsToDecrease = BIT_highbit32((U32)totalCost) + 1;
402
+ U32 nBitsToDecrease = ZSTD_highbit32((U32)totalCost) + 1;
358
403
  for ( ; nBitsToDecrease > 1; nBitsToDecrease--) {
359
404
  U32 const highPos = rankLast[nBitsToDecrease];
360
405
  U32 const lowPos = rankLast[nBitsToDecrease-1];
@@ -394,7 +439,7 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
394
439
  rankLast[nBitsToDecrease] = noSymbol;
395
440
  else {
396
441
  rankLast[nBitsToDecrease]--;
397
- if (huffNode[rankLast[nBitsToDecrease]].nbBits != maxNbBits-nBitsToDecrease)
442
+ if (huffNode[rankLast[nBitsToDecrease]].nbBits != targetNbBits-nBitsToDecrease)
398
443
  rankLast[nBitsToDecrease] = noSymbol; /* this rank is now empty */
399
444
  }
400
445
  } /* while (totalCost > 0) */
@@ -406,11 +451,11 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
406
451
  * TODO.
407
452
  */
408
453
  while (totalCost < 0) { /* Sometimes, cost correction overshoot */
409
- /* special case : no rank 1 symbol (using maxNbBits-1);
410
- * let's create one from largest rank 0 (using maxNbBits).
454
+ /* special case : no rank 1 symbol (using targetNbBits-1);
455
+ * let's create one from largest rank 0 (using targetNbBits).
411
456
  */
412
457
  if (rankLast[1] == noSymbol) {
413
- while (huffNode[n].nbBits == maxNbBits) n--;
458
+ while (huffNode[n].nbBits == targetNbBits) n--;
414
459
  huffNode[n+1].nbBits--;
415
460
  assert(n >= 0);
416
461
  rankLast[1] = (U32)(n+1);
@@ -424,7 +469,7 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
424
469
  } /* repay normalized cost */
425
470
  } /* there are several too large elements (at least >= 2) */
426
471
 
427
- return maxNbBits;
472
+ return targetNbBits;
428
473
  }
429
474
 
430
475
  typedef struct {
@@ -432,7 +477,7 @@ typedef struct {
432
477
  U16 curr;
433
478
  } rankPos;
434
479
 
435
- typedef nodeElt huffNodeTable[HUF_CTABLE_WORKSPACE_SIZE_U32];
480
+ typedef nodeElt huffNodeTable[2 * (HUF_SYMBOLVALUE_MAX + 1)];
436
481
 
437
482
  /* Number of buckets available for HUF_sort() */
438
483
  #define RANK_POSITION_TABLE_SIZE 192
@@ -451,8 +496,8 @@ typedef struct {
451
496
  * Let buckets 166 to 192 represent all remaining counts up to RANK_POSITION_MAX_COUNT_LOG using log2 bucketing.
452
497
  */
453
498
  #define RANK_POSITION_MAX_COUNT_LOG 32
454
- #define RANK_POSITION_LOG_BUCKETS_BEGIN (RANK_POSITION_TABLE_SIZE - 1) - RANK_POSITION_MAX_COUNT_LOG - 1 /* == 158 */
455
- #define RANK_POSITION_DISTINCT_COUNT_CUTOFF RANK_POSITION_LOG_BUCKETS_BEGIN + BIT_highbit32(RANK_POSITION_LOG_BUCKETS_BEGIN) /* == 166 */
499
+ #define RANK_POSITION_LOG_BUCKETS_BEGIN ((RANK_POSITION_TABLE_SIZE - 1) - RANK_POSITION_MAX_COUNT_LOG - 1 /* == 158 */)
500
+ #define RANK_POSITION_DISTINCT_COUNT_CUTOFF (RANK_POSITION_LOG_BUCKETS_BEGIN + ZSTD_highbit32(RANK_POSITION_LOG_BUCKETS_BEGIN) /* == 166 */)
456
501
 
457
502
  /* Return the appropriate bucket index for a given count. See definition of
458
503
  * RANK_POSITION_DISTINCT_COUNT_CUTOFF for explanation of bucketing strategy.
@@ -460,7 +505,7 @@ typedef struct {
460
505
  static U32 HUF_getIndex(U32 const count) {
461
506
  return (count < RANK_POSITION_DISTINCT_COUNT_CUTOFF)
462
507
  ? count
463
- : BIT_highbit32(count) + RANK_POSITION_LOG_BUCKETS_BEGIN;
508
+ : ZSTD_highbit32(count) + RANK_POSITION_LOG_BUCKETS_BEGIN;
464
509
  }
465
510
 
466
511
  /* Helper swap function for HUF_quickSortPartition() */
@@ -583,7 +628,7 @@ static void HUF_sort(nodeElt huffNode[], const unsigned count[], U32 const maxSy
583
628
 
584
629
  /* Sort each bucket. */
585
630
  for (n = RANK_POSITION_DISTINCT_COUNT_CUTOFF; n < RANK_POSITION_TABLE_SIZE - 1; ++n) {
586
- U32 const bucketSize = rankPosition[n].curr-rankPosition[n].base;
631
+ int const bucketSize = rankPosition[n].curr - rankPosition[n].base;
587
632
  U32 const bucketStartIdx = rankPosition[n].base;
588
633
  if (bucketSize > 1) {
589
634
  assert(bucketStartIdx < maxSymbolValue1);
@@ -594,6 +639,7 @@ static void HUF_sort(nodeElt huffNode[], const unsigned count[], U32 const maxSy
594
639
  assert(HUF_isSorted(huffNode, maxSymbolValue1));
595
640
  }
596
641
 
642
+
597
643
  /** HUF_buildCTable_wksp() :
598
644
  * Same as HUF_buildCTable(), but using externally allocated scratch buffer.
599
645
  * `workSpace` must be aligned on 4-bytes boundaries, and be at least as large as sizeof(HUF_buildCTable_wksp_tables).
@@ -614,6 +660,7 @@ static int HUF_buildTree(nodeElt* huffNode, U32 maxSymbolValue)
614
660
  int lowS, lowN;
615
661
  int nodeNb = STARTNODE;
616
662
  int n, nodeRoot;
663
+ DEBUGLOG(5, "HUF_buildTree (alphabet size = %u)", maxSymbolValue + 1);
617
664
  /* init for parents */
618
665
  nonNullRank = (int)maxSymbolValue;
619
666
  while(huffNode[nonNullRank].count == 0) nonNullRank--;
@@ -640,6 +687,8 @@ static int HUF_buildTree(nodeElt* huffNode, U32 maxSymbolValue)
640
687
  for (n=0; n<=nonNullRank; n++)
641
688
  huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1;
642
689
 
690
+ DEBUGLOG(6, "Initial distribution of bits completed (%zu sorted symbols)", showHNodeBits(huffNode, maxSymbolValue+1));
691
+
643
692
  return nonNullRank;
644
693
  }
645
694
 
@@ -677,28 +726,36 @@ static void HUF_buildCTableFromTree(HUF_CElt* CTable, nodeElt const* huffNode, i
677
726
  CTable[0] = maxNbBits;
678
727
  }
679
728
 
680
- size_t HUF_buildCTable_wksp (HUF_CElt* CTable, const unsigned* count, U32 maxSymbolValue, U32 maxNbBits, void* workSpace, size_t wkspSize)
729
+ size_t
730
+ HUF_buildCTable_wksp(HUF_CElt* CTable, const unsigned* count, U32 maxSymbolValue, U32 maxNbBits,
731
+ void* workSpace, size_t wkspSize)
681
732
  {
682
- HUF_buildCTable_wksp_tables* const wksp_tables = (HUF_buildCTable_wksp_tables*)HUF_alignUpWorkspace(workSpace, &wkspSize, ZSTD_ALIGNOF(U32));
733
+ HUF_buildCTable_wksp_tables* const wksp_tables =
734
+ (HUF_buildCTable_wksp_tables*)HUF_alignUpWorkspace(workSpace, &wkspSize, ZSTD_ALIGNOF(U32));
683
735
  nodeElt* const huffNode0 = wksp_tables->huffNodeTbl;
684
736
  nodeElt* const huffNode = huffNode0+1;
685
737
  int nonNullRank;
686
738
 
739
+ HUF_STATIC_ASSERT(HUF_CTABLE_WORKSPACE_SIZE == sizeof(HUF_buildCTable_wksp_tables));
740
+
741
+ DEBUGLOG(5, "HUF_buildCTable_wksp (alphabet size = %u)", maxSymbolValue+1);
742
+
687
743
  /* safety checks */
688
744
  if (wkspSize < sizeof(HUF_buildCTable_wksp_tables))
689
- return ERROR(workSpace_tooSmall);
745
+ return ERROR(workSpace_tooSmall);
690
746
  if (maxNbBits == 0) maxNbBits = HUF_TABLELOG_DEFAULT;
691
747
  if (maxSymbolValue > HUF_SYMBOLVALUE_MAX)
692
- return ERROR(maxSymbolValue_tooLarge);
748
+ return ERROR(maxSymbolValue_tooLarge);
693
749
  ZSTD_memset(huffNode0, 0, sizeof(huffNodeTable));
694
750
 
695
751
  /* sort, decreasing order */
696
752
  HUF_sort(huffNode, count, maxSymbolValue, wksp_tables->rankPosition);
753
+ DEBUGLOG(6, "sorted symbols completed (%zu symbols)", showHNodeSymbols(huffNode, maxSymbolValue+1));
697
754
 
698
755
  /* build tree */
699
756
  nonNullRank = HUF_buildTree(huffNode, maxSymbolValue);
700
757
 
701
- /* enforce maxTableLog */
758
+ /* determine and enforce maxTableLog */
702
759
  maxNbBits = HUF_setMaxHeight(huffNode, (U32)nonNullRank, maxNbBits);
703
760
  if (maxNbBits > HUF_TABLELOG_MAX) return ERROR(GENERIC); /* check fit into table */
704
761
 
@@ -807,7 +864,7 @@ FORCE_INLINE_TEMPLATE void HUF_addBits(HUF_CStream_t* bitC, HUF_CElt elt, int id
807
864
  #if DEBUGLEVEL >= 1
808
865
  {
809
866
  size_t const nbBits = HUF_getNbBits(elt);
810
- size_t const dirtyBits = nbBits == 0 ? 0 : BIT_highbit32((U32)nbBits) + 1;
867
+ size_t const dirtyBits = nbBits == 0 ? 0 : ZSTD_highbit32((U32)nbBits) + 1;
811
868
  (void)dirtyBits;
812
869
  /* Middle bits are 0. */
813
870
  assert(((elt >> dirtyBits) << (dirtyBits + nbBits)) == 0);
@@ -887,7 +944,7 @@ static size_t HUF_closeCStream(HUF_CStream_t* bitC)
887
944
  {
888
945
  size_t const nbBits = bitC->bitPos[0] & 0xFF;
889
946
  if (bitC->ptr >= bitC->endPtr) return 0; /* overflow detected */
890
- return (bitC->ptr - bitC->startPtr) + (nbBits > 0);
947
+ return (size_t)(bitC->ptr - bitC->startPtr) + (nbBits > 0);
891
948
  }
892
949
  }
893
950
 
@@ -1048,9 +1105,9 @@ HUF_compress1X_usingCTable_internal_default(void* dst, size_t dstSize,
1048
1105
  static size_t
1049
1106
  HUF_compress1X_usingCTable_internal(void* dst, size_t dstSize,
1050
1107
  const void* src, size_t srcSize,
1051
- const HUF_CElt* CTable, const int bmi2)
1108
+ const HUF_CElt* CTable, const int flags)
1052
1109
  {
1053
- if (bmi2) {
1110
+ if (flags & HUF_flags_bmi2) {
1054
1111
  return HUF_compress1X_usingCTable_internal_bmi2(dst, dstSize, src, srcSize, CTable);
1055
1112
  }
1056
1113
  return HUF_compress1X_usingCTable_internal_default(dst, dstSize, src, srcSize, CTable);
@@ -1061,28 +1118,23 @@ HUF_compress1X_usingCTable_internal(void* dst, size_t dstSize,
1061
1118
  static size_t
1062
1119
  HUF_compress1X_usingCTable_internal(void* dst, size_t dstSize,
1063
1120
  const void* src, size_t srcSize,
1064
- const HUF_CElt* CTable, const int bmi2)
1121
+ const HUF_CElt* CTable, const int flags)
1065
1122
  {
1066
- (void)bmi2;
1123
+ (void)flags;
1067
1124
  return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable);
1068
1125
  }
1069
1126
 
1070
1127
  #endif
1071
1128
 
1072
- size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
1129
+ size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable, int flags)
1073
1130
  {
1074
- return HUF_compress1X_usingCTable_bmi2(dst, dstSize, src, srcSize, CTable, /* bmi2 */ 0);
1075
- }
1076
-
1077
- size_t HUF_compress1X_usingCTable_bmi2(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable, int bmi2)
1078
- {
1079
- return HUF_compress1X_usingCTable_internal(dst, dstSize, src, srcSize, CTable, bmi2);
1131
+ return HUF_compress1X_usingCTable_internal(dst, dstSize, src, srcSize, CTable, flags);
1080
1132
  }
1081
1133
 
1082
1134
  static size_t
1083
1135
  HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
1084
1136
  const void* src, size_t srcSize,
1085
- const HUF_CElt* CTable, int bmi2)
1137
+ const HUF_CElt* CTable, int flags)
1086
1138
  {
1087
1139
  size_t const segmentSize = (srcSize+3)/4; /* first 3 segments */
1088
1140
  const BYTE* ip = (const BYTE*) src;
@@ -1096,7 +1148,7 @@ HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
1096
1148
  op += 6; /* jumpTable */
1097
1149
 
1098
1150
  assert(op <= oend);
1099
- { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) );
1151
+ { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, flags) );
1100
1152
  if (cSize == 0 || cSize > 65535) return 0;
1101
1153
  MEM_writeLE16(ostart, (U16)cSize);
1102
1154
  op += cSize;
@@ -1104,7 +1156,7 @@ HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
1104
1156
 
1105
1157
  ip += segmentSize;
1106
1158
  assert(op <= oend);
1107
- { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) );
1159
+ { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, flags) );
1108
1160
  if (cSize == 0 || cSize > 65535) return 0;
1109
1161
  MEM_writeLE16(ostart+2, (U16)cSize);
1110
1162
  op += cSize;
@@ -1112,7 +1164,7 @@ HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
1112
1164
 
1113
1165
  ip += segmentSize;
1114
1166
  assert(op <= oend);
1115
- { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) );
1167
+ { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, flags) );
1116
1168
  if (cSize == 0 || cSize > 65535) return 0;
1117
1169
  MEM_writeLE16(ostart+4, (U16)cSize);
1118
1170
  op += cSize;
@@ -1121,7 +1173,7 @@ HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
1121
1173
  ip += segmentSize;
1122
1174
  assert(op <= oend);
1123
1175
  assert(ip <= iend);
1124
- { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, (size_t)(iend-ip), CTable, bmi2) );
1176
+ { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, (size_t)(iend-ip), CTable, flags) );
1125
1177
  if (cSize == 0 || cSize > 65535) return 0;
1126
1178
  op += cSize;
1127
1179
  }
@@ -1129,14 +1181,9 @@ HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
1129
1181
  return (size_t)(op-ostart);
1130
1182
  }
1131
1183
 
1132
- size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
1133
- {
1134
- return HUF_compress4X_usingCTable_bmi2(dst, dstSize, src, srcSize, CTable, /* bmi2 */ 0);
1135
- }
1136
-
1137
- size_t HUF_compress4X_usingCTable_bmi2(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable, int bmi2)
1184
+ size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable, int flags)
1138
1185
  {
1139
- return HUF_compress4X_usingCTable_internal(dst, dstSize, src, srcSize, CTable, bmi2);
1186
+ return HUF_compress4X_usingCTable_internal(dst, dstSize, src, srcSize, CTable, flags);
1140
1187
  }
1141
1188
 
1142
1189
  typedef enum { HUF_singleStream, HUF_fourStreams } HUF_nbStreams_e;
@@ -1144,11 +1191,11 @@ typedef enum { HUF_singleStream, HUF_fourStreams } HUF_nbStreams_e;
1144
1191
  static size_t HUF_compressCTable_internal(
1145
1192
  BYTE* const ostart, BYTE* op, BYTE* const oend,
1146
1193
  const void* src, size_t srcSize,
1147
- HUF_nbStreams_e nbStreams, const HUF_CElt* CTable, const int bmi2)
1194
+ HUF_nbStreams_e nbStreams, const HUF_CElt* CTable, const int flags)
1148
1195
  {
1149
1196
  size_t const cSize = (nbStreams==HUF_singleStream) ?
1150
- HUF_compress1X_usingCTable_internal(op, (size_t)(oend - op), src, srcSize, CTable, bmi2) :
1151
- HUF_compress4X_usingCTable_internal(op, (size_t)(oend - op), src, srcSize, CTable, bmi2);
1197
+ HUF_compress1X_usingCTable_internal(op, (size_t)(oend - op), src, srcSize, CTable, flags) :
1198
+ HUF_compress4X_usingCTable_internal(op, (size_t)(oend - op), src, srcSize, CTable, flags);
1152
1199
  if (HUF_isError(cSize)) { return cSize; }
1153
1200
  if (cSize==0) { return 0; } /* uncompressible */
1154
1201
  op += cSize;
@@ -1171,6 +1218,79 @@ typedef struct {
1171
1218
  #define SUSPECT_INCOMPRESSIBLE_SAMPLE_SIZE 4096
1172
1219
  #define SUSPECT_INCOMPRESSIBLE_SAMPLE_RATIO 10 /* Must be >= 2 */
1173
1220
 
1221
+ unsigned HUF_cardinality(const unsigned* count, unsigned maxSymbolValue)
1222
+ {
1223
+ unsigned cardinality = 0;
1224
+ unsigned i;
1225
+
1226
+ for (i = 0; i < maxSymbolValue + 1; i++) {
1227
+ if (count[i] != 0) cardinality += 1;
1228
+ }
1229
+
1230
+ return cardinality;
1231
+ }
1232
+
1233
+ unsigned HUF_minTableLog(unsigned symbolCardinality)
1234
+ {
1235
+ U32 minBitsSymbols = ZSTD_highbit32(symbolCardinality) + 1;
1236
+ return minBitsSymbols;
1237
+ }
1238
+
1239
+ unsigned HUF_optimalTableLog(
1240
+ unsigned maxTableLog,
1241
+ size_t srcSize,
1242
+ unsigned maxSymbolValue,
1243
+ void* workSpace, size_t wkspSize,
1244
+ HUF_CElt* table,
1245
+ const unsigned* count,
1246
+ int flags)
1247
+ {
1248
+ assert(srcSize > 1); /* Not supported, RLE should be used instead */
1249
+ assert(wkspSize >= sizeof(HUF_buildCTable_wksp_tables));
1250
+
1251
+ if (!(flags & HUF_flags_optimalDepth)) {
1252
+ /* cheap evaluation, based on FSE */
1253
+ return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 1);
1254
+ }
1255
+
1256
+ { BYTE* dst = (BYTE*)workSpace + sizeof(HUF_WriteCTableWksp);
1257
+ size_t dstSize = wkspSize - sizeof(HUF_WriteCTableWksp);
1258
+ size_t maxBits, hSize, newSize;
1259
+ const unsigned symbolCardinality = HUF_cardinality(count, maxSymbolValue);
1260
+ const unsigned minTableLog = HUF_minTableLog(symbolCardinality);
1261
+ size_t optSize = ((size_t) ~0) - 1;
1262
+ unsigned optLog = maxTableLog, optLogGuess;
1263
+
1264
+ DEBUGLOG(6, "HUF_optimalTableLog: probing huf depth (srcSize=%zu)", srcSize);
1265
+
1266
+ /* Search until size increases */
1267
+ for (optLogGuess = minTableLog; optLogGuess <= maxTableLog; optLogGuess++) {
1268
+ DEBUGLOG(7, "checking for huffLog=%u", optLogGuess);
1269
+ maxBits = HUF_buildCTable_wksp(table, count, maxSymbolValue, optLogGuess, workSpace, wkspSize);
1270
+ if (ERR_isError(maxBits)) continue;
1271
+
1272
+ if (maxBits < optLogGuess && optLogGuess > minTableLog) break;
1273
+
1274
+ hSize = HUF_writeCTable_wksp(dst, dstSize, table, maxSymbolValue, (U32)maxBits, workSpace, wkspSize);
1275
+
1276
+ if (ERR_isError(hSize)) continue;
1277
+
1278
+ newSize = HUF_estimateCompressedSize(table, count, maxSymbolValue) + hSize;
1279
+
1280
+ if (newSize > optSize + 1) {
1281
+ break;
1282
+ }
1283
+
1284
+ if (newSize < optSize) {
1285
+ optSize = newSize;
1286
+ optLog = optLogGuess;
1287
+ }
1288
+ }
1289
+ assert(optLog <= HUF_TABLELOG_MAX);
1290
+ return optLog;
1291
+ }
1292
+ }
1293
+
1174
1294
  /* HUF_compress_internal() :
1175
1295
  * `workSpace_align4` must be aligned on 4-bytes boundaries,
1176
1296
  * and occupies the same space as a table of HUF_WORKSPACE_SIZE_U64 unsigned */
@@ -1180,14 +1300,14 @@ HUF_compress_internal (void* dst, size_t dstSize,
1180
1300
  unsigned maxSymbolValue, unsigned huffLog,
1181
1301
  HUF_nbStreams_e nbStreams,
1182
1302
  void* workSpace, size_t wkspSize,
1183
- HUF_CElt* oldHufTable, HUF_repeat* repeat, int preferRepeat,
1184
- const int bmi2, unsigned suspectUncompressible)
1303
+ HUF_CElt* oldHufTable, HUF_repeat* repeat, int flags)
1185
1304
  {
1186
1305
  HUF_compress_tables_t* const table = (HUF_compress_tables_t*)HUF_alignUpWorkspace(workSpace, &wkspSize, ZSTD_ALIGNOF(size_t));
1187
1306
  BYTE* const ostart = (BYTE*)dst;
1188
1307
  BYTE* const oend = ostart + dstSize;
1189
1308
  BYTE* op = ostart;
1190
1309
 
1310
+ DEBUGLOG(5, "HUF_compress_internal (srcSize=%zu)", srcSize);
1191
1311
  HUF_STATIC_ASSERT(sizeof(*table) + HUF_WORKSPACE_MAX_ALIGNMENT <= HUF_WORKSPACE_SIZE);
1192
1312
 
1193
1313
  /* checks & inits */
@@ -1201,16 +1321,17 @@ HUF_compress_internal (void* dst, size_t dstSize,
1201
1321
  if (!huffLog) huffLog = HUF_TABLELOG_DEFAULT;
1202
1322
 
1203
1323
  /* Heuristic : If old table is valid, use it for small inputs */
1204
- if (preferRepeat && repeat && *repeat == HUF_repeat_valid) {
1324
+ if ((flags & HUF_flags_preferRepeat) && repeat && *repeat == HUF_repeat_valid) {
1205
1325
  return HUF_compressCTable_internal(ostart, op, oend,
1206
1326
  src, srcSize,
1207
- nbStreams, oldHufTable, bmi2);
1327
+ nbStreams, oldHufTable, flags);
1208
1328
  }
1209
1329
 
1210
1330
  /* If uncompressible data is suspected, do a smaller sampling first */
1211
1331
  DEBUG_STATIC_ASSERT(SUSPECT_INCOMPRESSIBLE_SAMPLE_RATIO >= 2);
1212
- if (suspectUncompressible && srcSize >= (SUSPECT_INCOMPRESSIBLE_SAMPLE_SIZE * SUSPECT_INCOMPRESSIBLE_SAMPLE_RATIO)) {
1332
+ if ((flags & HUF_flags_suspectUncompressible) && srcSize >= (SUSPECT_INCOMPRESSIBLE_SAMPLE_SIZE * SUSPECT_INCOMPRESSIBLE_SAMPLE_RATIO)) {
1213
1333
  size_t largestTotal = 0;
1334
+ DEBUGLOG(5, "input suspected incompressible : sampling to check");
1214
1335
  { unsigned maxSymbolValueBegin = maxSymbolValue;
1215
1336
  CHECK_V_F(largestBegin, HIST_count_simple (table->count, &maxSymbolValueBegin, (const BYTE*)src, SUSPECT_INCOMPRESSIBLE_SAMPLE_SIZE) );
1216
1337
  largestTotal += largestBegin;
@@ -1227,6 +1348,7 @@ HUF_compress_internal (void* dst, size_t dstSize,
1227
1348
  if (largest == srcSize) { *ostart = ((const BYTE*)src)[0]; return 1; } /* single symbol, rle */
1228
1349
  if (largest <= (srcSize >> 7)+4) return 0; /* heuristic : probably not compressible enough */
1229
1350
  }
1351
+ DEBUGLOG(6, "histogram detail completed (%zu symbols)", showU32(table->count, maxSymbolValue+1));
1230
1352
 
1231
1353
  /* Check validity of previous table */
1232
1354
  if ( repeat
@@ -1235,19 +1357,20 @@ HUF_compress_internal (void* dst, size_t dstSize,
1235
1357
  *repeat = HUF_repeat_none;
1236
1358
  }
1237
1359
  /* Heuristic : use existing table for small inputs */
1238
- if (preferRepeat && repeat && *repeat != HUF_repeat_none) {
1360
+ if ((flags & HUF_flags_preferRepeat) && repeat && *repeat != HUF_repeat_none) {
1239
1361
  return HUF_compressCTable_internal(ostart, op, oend,
1240
1362
  src, srcSize,
1241
- nbStreams, oldHufTable, bmi2);
1363
+ nbStreams, oldHufTable, flags);
1242
1364
  }
1243
1365
 
1244
1366
  /* Build Huffman Tree */
1245
- huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
1367
+ huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue, &table->wksps, sizeof(table->wksps), table->CTable, table->count, flags);
1246
1368
  { size_t const maxBits = HUF_buildCTable_wksp(table->CTable, table->count,
1247
1369
  maxSymbolValue, huffLog,
1248
1370
  &table->wksps.buildCTable_wksp, sizeof(table->wksps.buildCTable_wksp));
1249
1371
  CHECK_F(maxBits);
1250
1372
  huffLog = (U32)maxBits;
1373
+ DEBUGLOG(6, "bit distribution completed (%zu symbols)", showCTableBits(table->CTable + 1, maxSymbolValue+1));
1251
1374
  }
1252
1375
  /* Zero unused symbols in CTable, so we can check it for validity */
1253
1376
  {
@@ -1266,7 +1389,7 @@ HUF_compress_internal (void* dst, size_t dstSize,
1266
1389
  if (oldSize <= hSize + newSize || hSize + 12 >= srcSize) {
1267
1390
  return HUF_compressCTable_internal(ostart, op, oend,
1268
1391
  src, srcSize,
1269
- nbStreams, oldHufTable, bmi2);
1392
+ nbStreams, oldHufTable, flags);
1270
1393
  } }
1271
1394
 
1272
1395
  /* Use the new huffman table */
@@ -1278,46 +1401,20 @@ HUF_compress_internal (void* dst, size_t dstSize,
1278
1401
  }
1279
1402
  return HUF_compressCTable_internal(ostart, op, oend,
1280
1403
  src, srcSize,
1281
- nbStreams, table->CTable, bmi2);
1282
- }
1283
-
1284
-
1285
- size_t HUF_compress1X_wksp (void* dst, size_t dstSize,
1286
- const void* src, size_t srcSize,
1287
- unsigned maxSymbolValue, unsigned huffLog,
1288
- void* workSpace, size_t wkspSize)
1289
- {
1290
- return HUF_compress_internal(dst, dstSize, src, srcSize,
1291
- maxSymbolValue, huffLog, HUF_singleStream,
1292
- workSpace, wkspSize,
1293
- NULL, NULL, 0, 0 /*bmi2*/, 0);
1404
+ nbStreams, table->CTable, flags);
1294
1405
  }
1295
1406
 
1296
1407
  size_t HUF_compress1X_repeat (void* dst, size_t dstSize,
1297
1408
  const void* src, size_t srcSize,
1298
1409
  unsigned maxSymbolValue, unsigned huffLog,
1299
1410
  void* workSpace, size_t wkspSize,
1300
- HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat,
1301
- int bmi2, unsigned suspectUncompressible)
1411
+ HUF_CElt* hufTable, HUF_repeat* repeat, int flags)
1302
1412
  {
1413
+ DEBUGLOG(5, "HUF_compress1X_repeat (srcSize = %zu)", srcSize);
1303
1414
  return HUF_compress_internal(dst, dstSize, src, srcSize,
1304
1415
  maxSymbolValue, huffLog, HUF_singleStream,
1305
1416
  workSpace, wkspSize, hufTable,
1306
- repeat, preferRepeat, bmi2, suspectUncompressible);
1307
- }
1308
-
1309
- /* HUF_compress4X_repeat():
1310
- * compress input using 4 streams.
1311
- * provide workspace to generate compression tables */
1312
- size_t HUF_compress4X_wksp (void* dst, size_t dstSize,
1313
- const void* src, size_t srcSize,
1314
- unsigned maxSymbolValue, unsigned huffLog,
1315
- void* workSpace, size_t wkspSize)
1316
- {
1317
- return HUF_compress_internal(dst, dstSize, src, srcSize,
1318
- maxSymbolValue, huffLog, HUF_fourStreams,
1319
- workSpace, wkspSize,
1320
- NULL, NULL, 0, 0 /*bmi2*/, 0);
1417
+ repeat, flags);
1321
1418
  }
1322
1419
 
1323
1420
  /* HUF_compress4X_repeat():
@@ -1328,43 +1425,11 @@ size_t HUF_compress4X_repeat (void* dst, size_t dstSize,
1328
1425
  const void* src, size_t srcSize,
1329
1426
  unsigned maxSymbolValue, unsigned huffLog,
1330
1427
  void* workSpace, size_t wkspSize,
1331
- HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2, unsigned suspectUncompressible)
1428
+ HUF_CElt* hufTable, HUF_repeat* repeat, int flags)
1332
1429
  {
1430
+ DEBUGLOG(5, "HUF_compress4X_repeat (srcSize = %zu)", srcSize);
1333
1431
  return HUF_compress_internal(dst, dstSize, src, srcSize,
1334
1432
  maxSymbolValue, huffLog, HUF_fourStreams,
1335
1433
  workSpace, wkspSize,
1336
- hufTable, repeat, preferRepeat, bmi2, suspectUncompressible);
1434
+ hufTable, repeat, flags);
1337
1435
  }
1338
-
1339
- #ifndef ZSTD_NO_UNUSED_FUNCTIONS
1340
- /** HUF_buildCTable() :
1341
- * @return : maxNbBits
1342
- * Note : count is used before tree is written, so they can safely overlap
1343
- */
1344
- size_t HUF_buildCTable (HUF_CElt* tree, const unsigned* count, unsigned maxSymbolValue, unsigned maxNbBits)
1345
- {
1346
- HUF_buildCTable_wksp_tables workspace;
1347
- return HUF_buildCTable_wksp(tree, count, maxSymbolValue, maxNbBits, &workspace, sizeof(workspace));
1348
- }
1349
-
1350
- size_t HUF_compress1X (void* dst, size_t dstSize,
1351
- const void* src, size_t srcSize,
1352
- unsigned maxSymbolValue, unsigned huffLog)
1353
- {
1354
- U64 workSpace[HUF_WORKSPACE_SIZE_U64];
1355
- return HUF_compress1X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
1356
- }
1357
-
1358
- size_t HUF_compress2 (void* dst, size_t dstSize,
1359
- const void* src, size_t srcSize,
1360
- unsigned maxSymbolValue, unsigned huffLog)
1361
- {
1362
- U64 workSpace[HUF_WORKSPACE_SIZE_U64];
1363
- return HUF_compress4X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
1364
- }
1365
-
1366
- size_t HUF_compress (void* dst, size_t maxDstSize, const void* src, size_t srcSize)
1367
- {
1368
- return HUF_compress2(dst, maxDstSize, src, srcSize, 255, HUF_TABLELOG_DEFAULT);
1369
- }
1370
- #endif