extzstd 0.3.1 → 0.3.2

Sign up to get free protection for your applications and to get access to all the features.
Files changed (76) hide show
  1. checksums.yaml +4 -4
  2. data/README.md +28 -14
  3. data/contrib/zstd/CHANGELOG +114 -56
  4. data/contrib/zstd/CONTRIBUTING.md +14 -0
  5. data/contrib/zstd/Makefile +37 -31
  6. data/contrib/zstd/README.md +6 -0
  7. data/contrib/zstd/appveyor.yml +4 -1
  8. data/contrib/zstd/lib/Makefile +231 -134
  9. data/contrib/zstd/lib/README.md +28 -0
  10. data/contrib/zstd/lib/common/bitstream.h +24 -15
  11. data/contrib/zstd/lib/common/compiler.h +116 -3
  12. data/contrib/zstd/lib/common/cpu.h +0 -2
  13. data/contrib/zstd/lib/common/debug.h +11 -18
  14. data/contrib/zstd/lib/common/entropy_common.c +188 -42
  15. data/contrib/zstd/lib/common/error_private.c +1 -0
  16. data/contrib/zstd/lib/common/error_private.h +1 -1
  17. data/contrib/zstd/lib/common/fse.h +38 -11
  18. data/contrib/zstd/lib/common/fse_decompress.c +123 -16
  19. data/contrib/zstd/lib/common/huf.h +26 -5
  20. data/contrib/zstd/lib/common/mem.h +66 -93
  21. data/contrib/zstd/lib/common/pool.c +22 -16
  22. data/contrib/zstd/lib/common/pool.h +1 -1
  23. data/contrib/zstd/lib/common/threading.c +6 -5
  24. data/contrib/zstd/lib/common/xxhash.c +18 -56
  25. data/contrib/zstd/lib/common/xxhash.h +1 -1
  26. data/contrib/zstd/lib/common/zstd_common.c +9 -9
  27. data/contrib/zstd/lib/common/zstd_deps.h +111 -0
  28. data/contrib/zstd/lib/common/zstd_errors.h +1 -0
  29. data/contrib/zstd/lib/common/zstd_internal.h +89 -58
  30. data/contrib/zstd/lib/compress/fse_compress.c +30 -23
  31. data/contrib/zstd/lib/compress/hist.c +26 -28
  32. data/contrib/zstd/lib/compress/hist.h +1 -1
  33. data/contrib/zstd/lib/compress/huf_compress.c +210 -95
  34. data/contrib/zstd/lib/compress/zstd_compress.c +1339 -409
  35. data/contrib/zstd/lib/compress/zstd_compress_internal.h +119 -41
  36. data/contrib/zstd/lib/compress/zstd_compress_literals.c +4 -4
  37. data/contrib/zstd/lib/compress/zstd_compress_sequences.c +17 -3
  38. data/contrib/zstd/lib/compress/zstd_compress_superblock.c +23 -19
  39. data/contrib/zstd/lib/compress/zstd_cwksp.h +60 -24
  40. data/contrib/zstd/lib/compress/zstd_double_fast.c +22 -22
  41. data/contrib/zstd/lib/compress/zstd_fast.c +19 -19
  42. data/contrib/zstd/lib/compress/zstd_lazy.c +351 -77
  43. data/contrib/zstd/lib/compress/zstd_lazy.h +20 -0
  44. data/contrib/zstd/lib/compress/zstd_ldm.c +59 -18
  45. data/contrib/zstd/lib/compress/zstd_ldm.h +6 -0
  46. data/contrib/zstd/lib/compress/zstd_opt.c +190 -45
  47. data/contrib/zstd/lib/compress/zstdmt_compress.c +74 -406
  48. data/contrib/zstd/lib/compress/zstdmt_compress.h +26 -108
  49. data/contrib/zstd/lib/decompress/huf_decompress.c +302 -200
  50. data/contrib/zstd/lib/decompress/zstd_ddict.c +8 -8
  51. data/contrib/zstd/lib/decompress/zstd_ddict.h +1 -1
  52. data/contrib/zstd/lib/decompress/zstd_decompress.c +125 -80
  53. data/contrib/zstd/lib/decompress/zstd_decompress_block.c +145 -37
  54. data/contrib/zstd/lib/decompress/zstd_decompress_block.h +5 -2
  55. data/contrib/zstd/lib/decompress/zstd_decompress_internal.h +11 -10
  56. data/contrib/zstd/lib/dictBuilder/cover.c +29 -20
  57. data/contrib/zstd/lib/dictBuilder/cover.h +1 -1
  58. data/contrib/zstd/lib/dictBuilder/fastcover.c +20 -19
  59. data/contrib/zstd/lib/dictBuilder/zdict.c +15 -16
  60. data/contrib/zstd/lib/dictBuilder/zdict.h +1 -1
  61. data/contrib/zstd/lib/legacy/zstd_v01.c +5 -1
  62. data/contrib/zstd/lib/legacy/zstd_v02.c +5 -1
  63. data/contrib/zstd/lib/legacy/zstd_v03.c +5 -1
  64. data/contrib/zstd/lib/legacy/zstd_v04.c +6 -2
  65. data/contrib/zstd/lib/legacy/zstd_v05.c +5 -1
  66. data/contrib/zstd/lib/legacy/zstd_v06.c +5 -1
  67. data/contrib/zstd/lib/legacy/zstd_v07.c +5 -1
  68. data/contrib/zstd/lib/libzstd.pc.in +3 -3
  69. data/contrib/zstd/lib/zstd.h +348 -47
  70. data/ext/extzstd.c +6 -0
  71. data/ext/extzstd.h +6 -0
  72. data/gemstub.rb +3 -21
  73. data/lib/extzstd.rb +0 -2
  74. data/lib/extzstd/version.rb +6 -1
  75. data/test/test_basic.rb +0 -5
  76. metadata +5 -4
@@ -19,26 +19,14 @@
19
19
  /* Note : This is an internal API.
20
20
  * These APIs used to be exposed with ZSTDLIB_API,
21
21
  * because it used to be the only way to invoke MT compression.
22
- * Now, it's recommended to use ZSTD_compress2 and ZSTD_compressStream2()
23
- * instead.
24
- *
25
- * If you depend on these APIs and can't switch, then define
26
- * ZSTD_LEGACY_MULTITHREADED_API when making the dynamic library.
27
- * However, we may completely remove these functions in a future
28
- * release, so please switch soon.
22
+ * Now, you must use ZSTD_compress2 and ZSTD_compressStream2() instead.
29
23
  *
30
24
  * This API requires ZSTD_MULTITHREAD to be defined during compilation,
31
25
  * otherwise ZSTDMT_createCCtx*() will fail.
32
26
  */
33
27
 
34
- #ifdef ZSTD_LEGACY_MULTITHREADED_API
35
- # define ZSTDMT_API ZSTDLIB_API
36
- #else
37
- # define ZSTDMT_API
38
- #endif
39
-
40
28
  /* === Dependencies === */
41
- #include <stddef.h> /* size_t */
29
+ #include "../common/zstd_deps.h" /* size_t */
42
30
  #define ZSTD_STATIC_LINKING_ONLY /* ZSTD_parameters */
43
31
  #include "../zstd.h" /* ZSTD_inBuffer, ZSTD_outBuffer, ZSTDLIB_API */
44
32
 
@@ -54,78 +42,34 @@
54
42
  #define ZSTDMT_JOBSIZE_MAX (MEM_32bits() ? (512 MB) : (1024 MB))
55
43
 
56
44
 
45
+ /* ========================================================
46
+ * === Private interface, for use by ZSTD_compress.c ===
47
+ * === Not exposed in libzstd. Never invoke directly ===
48
+ * ======================================================== */
49
+
57
50
  /* === Memory management === */
58
51
  typedef struct ZSTDMT_CCtx_s ZSTDMT_CCtx;
59
52
  /* Requires ZSTD_MULTITHREAD to be defined during compilation, otherwise it will return NULL. */
60
- ZSTDMT_API ZSTDMT_CCtx* ZSTDMT_createCCtx(unsigned nbWorkers);
61
- /* Requires ZSTD_MULTITHREAD to be defined during compilation, otherwise it will return NULL. */
62
- ZSTDMT_API ZSTDMT_CCtx* ZSTDMT_createCCtx_advanced(unsigned nbWorkers,
63
- ZSTD_customMem cMem);
64
- ZSTDMT_API size_t ZSTDMT_freeCCtx(ZSTDMT_CCtx* mtctx);
65
-
66
- ZSTDMT_API size_t ZSTDMT_sizeof_CCtx(ZSTDMT_CCtx* mtctx);
67
-
68
-
69
- /* === Simple one-pass compression function === */
70
-
71
- ZSTDMT_API size_t ZSTDMT_compressCCtx(ZSTDMT_CCtx* mtctx,
72
- void* dst, size_t dstCapacity,
73
- const void* src, size_t srcSize,
74
- int compressionLevel);
75
-
53
+ ZSTDMT_CCtx* ZSTDMT_createCCtx_advanced(unsigned nbWorkers,
54
+ ZSTD_customMem cMem,
55
+ ZSTD_threadPool *pool);
56
+ size_t ZSTDMT_freeCCtx(ZSTDMT_CCtx* mtctx);
76
57
 
58
+ size_t ZSTDMT_sizeof_CCtx(ZSTDMT_CCtx* mtctx);
77
59
 
78
60
  /* === Streaming functions === */
79
61
 
80
- ZSTDMT_API size_t ZSTDMT_initCStream(ZSTDMT_CCtx* mtctx, int compressionLevel);
81
- ZSTDMT_API size_t ZSTDMT_resetCStream(ZSTDMT_CCtx* mtctx, unsigned long long pledgedSrcSize); /**< if srcSize is not known at reset time, use ZSTD_CONTENTSIZE_UNKNOWN. Note: for compatibility with older programs, 0 means the same as ZSTD_CONTENTSIZE_UNKNOWN, but it will change in the future to mean "empty" */
82
-
83
- ZSTDMT_API size_t ZSTDMT_nextInputSizeHint(const ZSTDMT_CCtx* mtctx);
84
- ZSTDMT_API size_t ZSTDMT_compressStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input);
85
-
86
- ZSTDMT_API size_t ZSTDMT_flushStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output); /**< @return : 0 == all flushed; >0 : still some data to be flushed; or an error code (ZSTD_isError()) */
87
- ZSTDMT_API size_t ZSTDMT_endStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output); /**< @return : 0 == all flushed; >0 : still some data to be flushed; or an error code (ZSTD_isError()) */
88
-
89
-
90
- /* === Advanced functions and parameters === */
91
-
92
- ZSTDMT_API size_t ZSTDMT_compress_advanced(ZSTDMT_CCtx* mtctx,
93
- void* dst, size_t dstCapacity,
94
- const void* src, size_t srcSize,
95
- const ZSTD_CDict* cdict,
96
- ZSTD_parameters params,
97
- int overlapLog);
98
-
99
- ZSTDMT_API size_t ZSTDMT_initCStream_advanced(ZSTDMT_CCtx* mtctx,
100
- const void* dict, size_t dictSize, /* dict can be released after init, a local copy is preserved within zcs */
101
- ZSTD_parameters params,
102
- unsigned long long pledgedSrcSize); /* pledgedSrcSize is optional and can be zero == unknown */
103
-
104
- ZSTDMT_API size_t ZSTDMT_initCStream_usingCDict(ZSTDMT_CCtx* mtctx,
105
- const ZSTD_CDict* cdict,
106
- ZSTD_frameParameters fparams,
107
- unsigned long long pledgedSrcSize); /* note : zero means empty */
108
-
109
- /* ZSTDMT_parameter :
110
- * List of parameters that can be set using ZSTDMT_setMTCtxParameter() */
111
- typedef enum {
112
- ZSTDMT_p_jobSize, /* Each job is compressed in parallel. By default, this value is dynamically determined depending on compression parameters. Can be set explicitly here. */
113
- ZSTDMT_p_overlapLog, /* Each job may reload a part of previous job to enhance compression ratio; 0 == no overlap, 6(default) == use 1/8th of window, >=9 == use full window. This is a "sticky" parameter : its value will be re-used on next compression job */
114
- ZSTDMT_p_rsyncable /* Enables rsyncable mode. */
115
- } ZSTDMT_parameter;
116
-
117
- /* ZSTDMT_setMTCtxParameter() :
118
- * allow setting individual parameters, one at a time, among a list of enums defined in ZSTDMT_parameter.
119
- * The function must be called typically after ZSTD_createCCtx() but __before ZSTDMT_init*() !__
120
- * Parameters not explicitly reset by ZSTDMT_init*() remain the same in consecutive compression sessions.
121
- * @return : 0, or an error code (which can be tested using ZSTD_isError()) */
122
- ZSTDMT_API size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, int value);
123
-
124
- /* ZSTDMT_getMTCtxParameter() :
125
- * Query the ZSTDMT_CCtx for a parameter value.
126
- * @return : 0, or an error code (which can be tested using ZSTD_isError()) */
127
- ZSTDMT_API size_t ZSTDMT_getMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, int* value);
62
+ size_t ZSTDMT_nextInputSizeHint(const ZSTDMT_CCtx* mtctx);
128
63
 
64
+ /*! ZSTDMT_initCStream_internal() :
65
+ * Private use only. Init streaming operation.
66
+ * expects params to be valid.
67
+ * must receive dict, or cdict, or none, but not both.
68
+ * @return : 0, or an error code */
69
+ size_t ZSTDMT_initCStream_internal(ZSTDMT_CCtx* zcs,
70
+ const void* dict, size_t dictSize, ZSTD_dictContentType_e dictContentType,
71
+ const ZSTD_CDict* cdict,
72
+ ZSTD_CCtx_params params, unsigned long long pledgedSrcSize);
129
73
 
130
74
  /*! ZSTDMT_compressStream_generic() :
131
75
  * Combines ZSTDMT_compressStream() with optional ZSTDMT_flushStream() or ZSTDMT_endStream()
@@ -134,16 +78,10 @@ ZSTDMT_API size_t ZSTDMT_getMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter
134
78
  * 0 if fully flushed
135
79
  * or an error code
136
80
  * note : needs to be init using any ZSTD_initCStream*() variant */
137
- ZSTDMT_API size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
138
- ZSTD_outBuffer* output,
139
- ZSTD_inBuffer* input,
140
- ZSTD_EndDirective endOp);
141
-
142
-
143
- /* ========================================================
144
- * === Private interface, for use by ZSTD_compress.c ===
145
- * === Not exposed in libzstd. Never invoke directly ===
146
- * ======================================================== */
81
+ size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
82
+ ZSTD_outBuffer* output,
83
+ ZSTD_inBuffer* input,
84
+ ZSTD_EndDirective endOp);
147
85
 
148
86
  /*! ZSTDMT_toFlushNow()
149
87
  * Tell how many bytes are ready to be flushed immediately.
@@ -153,15 +91,6 @@ ZSTDMT_API size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
153
91
  * therefore flushing is limited by speed of oldest job. */
154
92
  size_t ZSTDMT_toFlushNow(ZSTDMT_CCtx* mtctx);
155
93
 
156
- /*! ZSTDMT_CCtxParam_setMTCtxParameter()
157
- * like ZSTDMT_setMTCtxParameter(), but into a ZSTD_CCtx_Params */
158
- size_t ZSTDMT_CCtxParam_setMTCtxParameter(ZSTD_CCtx_params* params, ZSTDMT_parameter parameter, int value);
159
-
160
- /*! ZSTDMT_CCtxParam_setNbWorkers()
161
- * Set nbWorkers, and clamp it.
162
- * Also reset jobSize and overlapLog */
163
- size_t ZSTDMT_CCtxParam_setNbWorkers(ZSTD_CCtx_params* params, unsigned nbWorkers);
164
-
165
94
  /*! ZSTDMT_updateCParams_whileCompressing() :
166
95
  * Updates only a selected set of compression parameters, to remain compatible with current frame.
167
96
  * New parameters will be applied to next compression job. */
@@ -174,17 +103,6 @@ void ZSTDMT_updateCParams_whileCompressing(ZSTDMT_CCtx* mtctx, const ZSTD_CCtx_p
174
103
  ZSTD_frameProgression ZSTDMT_getFrameProgression(ZSTDMT_CCtx* mtctx);
175
104
 
176
105
 
177
- /*! ZSTDMT_initCStream_internal() :
178
- * Private use only. Init streaming operation.
179
- * expects params to be valid.
180
- * must receive dict, or cdict, or none, but not both.
181
- * @return : 0, or an error code */
182
- size_t ZSTDMT_initCStream_internal(ZSTDMT_CCtx* zcs,
183
- const void* dict, size_t dictSize, ZSTD_dictContentType_e dictContentType,
184
- const ZSTD_CDict* cdict,
185
- ZSTD_CCtx_params params, unsigned long long pledgedSrcSize);
186
-
187
-
188
106
  #if defined (__cplusplus)
189
107
  }
190
108
  #endif
@@ -15,7 +15,7 @@
15
15
  /* **************************************************************
16
16
  * Dependencies
17
17
  ****************************************************************/
18
- #include <string.h> /* memcpy, memset */
18
+ #include "../common/zstd_deps.h" /* ZSTD_memcpy, ZSTD_memset */
19
19
  #include "../common/compiler.h"
20
20
  #include "../common/bitstream.h" /* BIT_* */
21
21
  #include "../common/fse.h" /* to compress headers */
@@ -103,7 +103,7 @@ typedef struct { BYTE maxTableLog; BYTE tableType; BYTE tableLog; BYTE reserved;
103
103
  static DTableDesc HUF_getDTableDesc(const HUF_DTable* table)
104
104
  {
105
105
  DTableDesc dtd;
106
- memcpy(&dtd, table, sizeof(dtd));
106
+ ZSTD_memcpy(&dtd, table, sizeof(dtd));
107
107
  return dtd;
108
108
  }
109
109
 
@@ -115,29 +115,51 @@ static DTableDesc HUF_getDTableDesc(const HUF_DTable* table)
115
115
  /*-***************************/
116
116
  typedef struct { BYTE byte; BYTE nbBits; } HUF_DEltX1; /* single-symbol decoding */
117
117
 
118
+ /**
119
+ * Packs 4 HUF_DEltX1 structs into a U64. This is used to lay down 4 entries at
120
+ * a time.
121
+ */
122
+ static U64 HUF_DEltX1_set4(BYTE symbol, BYTE nbBits) {
123
+ U64 D4;
124
+ if (MEM_isLittleEndian()) {
125
+ D4 = symbol + (nbBits << 8);
126
+ } else {
127
+ D4 = (symbol << 8) + nbBits;
128
+ }
129
+ D4 *= 0x0001000100010001ULL;
130
+ return D4;
131
+ }
132
+
133
+ typedef struct {
134
+ U32 rankVal[HUF_TABLELOG_ABSOLUTEMAX + 1];
135
+ U32 rankStart[HUF_TABLELOG_ABSOLUTEMAX + 1];
136
+ U32 statsWksp[HUF_READ_STATS_WORKSPACE_SIZE_U32];
137
+ BYTE symbols[HUF_SYMBOLVALUE_MAX + 1];
138
+ BYTE huffWeight[HUF_SYMBOLVALUE_MAX + 1];
139
+ } HUF_ReadDTableX1_Workspace;
140
+
141
+
118
142
  size_t HUF_readDTableX1_wksp(HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize)
143
+ {
144
+ return HUF_readDTableX1_wksp_bmi2(DTable, src, srcSize, workSpace, wkspSize, /* bmi2 */ 0);
145
+ }
146
+
147
+ size_t HUF_readDTableX1_wksp_bmi2(HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize, int bmi2)
119
148
  {
120
149
  U32 tableLog = 0;
121
150
  U32 nbSymbols = 0;
122
151
  size_t iSize;
123
152
  void* const dtPtr = DTable + 1;
124
153
  HUF_DEltX1* const dt = (HUF_DEltX1*)dtPtr;
154
+ HUF_ReadDTableX1_Workspace* wksp = (HUF_ReadDTableX1_Workspace*)workSpace;
125
155
 
126
- U32* rankVal;
127
- BYTE* huffWeight;
128
- size_t spaceUsed32 = 0;
129
-
130
- rankVal = (U32 *)workSpace + spaceUsed32;
131
- spaceUsed32 += HUF_TABLELOG_ABSOLUTEMAX + 1;
132
- huffWeight = (BYTE *)((U32 *)workSpace + spaceUsed32);
133
- spaceUsed32 += HUF_ALIGN(HUF_SYMBOLVALUE_MAX + 1, sizeof(U32)) >> 2;
134
-
135
- if ((spaceUsed32 << 2) > wkspSize) return ERROR(tableLog_tooLarge);
156
+ DEBUG_STATIC_ASSERT(HUF_DECOMPRESS_WORKSPACE_SIZE >= sizeof(*wksp));
157
+ if (sizeof(*wksp) > wkspSize) return ERROR(tableLog_tooLarge);
136
158
 
137
159
  DEBUG_STATIC_ASSERT(sizeof(DTableDesc) == sizeof(HUF_DTable));
138
- /* memset(huffWeight, 0, sizeof(huffWeight)); */ /* is not necessary, even though some analyzer complain ... */
160
+ /* ZSTD_memset(huffWeight, 0, sizeof(huffWeight)); */ /* is not necessary, even though some analyzer complain ... */
139
161
 
140
- iSize = HUF_readStats(huffWeight, HUF_SYMBOLVALUE_MAX + 1, rankVal, &nbSymbols, &tableLog, src, srcSize);
162
+ iSize = HUF_readStats_wksp(wksp->huffWeight, HUF_SYMBOLVALUE_MAX + 1, wksp->rankVal, &nbSymbols, &tableLog, src, srcSize, wksp->statsWksp, sizeof(wksp->statsWksp), bmi2);
141
163
  if (HUF_isError(iSize)) return iSize;
142
164
 
143
165
  /* Table header */
@@ -145,52 +167,117 @@ size_t HUF_readDTableX1_wksp(HUF_DTable* DTable, const void* src, size_t srcSize
145
167
  if (tableLog > (U32)(dtd.maxTableLog+1)) return ERROR(tableLog_tooLarge); /* DTable too small, Huffman tree cannot fit in */
146
168
  dtd.tableType = 0;
147
169
  dtd.tableLog = (BYTE)tableLog;
148
- memcpy(DTable, &dtd, sizeof(dtd));
170
+ ZSTD_memcpy(DTable, &dtd, sizeof(dtd));
149
171
  }
150
172
 
151
- /* Calculate starting value for each rank */
152
- { U32 n, nextRankStart = 0;
153
- for (n=1; n<tableLog+1; n++) {
154
- U32 const current = nextRankStart;
155
- nextRankStart += (rankVal[n] << (n-1));
156
- rankVal[n] = current;
157
- } }
173
+ /* Compute symbols and rankStart given rankVal:
174
+ *
175
+ * rankVal already contains the number of values of each weight.
176
+ *
177
+ * symbols contains the symbols ordered by weight. First are the rankVal[0]
178
+ * weight 0 symbols, followed by the rankVal[1] weight 1 symbols, and so on.
179
+ * symbols[0] is filled (but unused) to avoid a branch.
180
+ *
181
+ * rankStart contains the offset where each rank belongs in the DTable.
182
+ * rankStart[0] is not filled because there are no entries in the table for
183
+ * weight 0.
184
+ */
185
+ {
186
+ int n;
187
+ int nextRankStart = 0;
188
+ int const unroll = 4;
189
+ int const nLimit = (int)nbSymbols - unroll + 1;
190
+ for (n=0; n<(int)tableLog+1; n++) {
191
+ U32 const curr = nextRankStart;
192
+ nextRankStart += wksp->rankVal[n];
193
+ wksp->rankStart[n] = curr;
194
+ }
195
+ for (n=0; n < nLimit; n += unroll) {
196
+ int u;
197
+ for (u=0; u < unroll; ++u) {
198
+ size_t const w = wksp->huffWeight[n+u];
199
+ wksp->symbols[wksp->rankStart[w]++] = (BYTE)(n+u);
200
+ }
201
+ }
202
+ for (; n < (int)nbSymbols; ++n) {
203
+ size_t const w = wksp->huffWeight[n];
204
+ wksp->symbols[wksp->rankStart[w]++] = (BYTE)n;
205
+ }
206
+ }
158
207
 
159
- /* fill DTable */
160
- { U32 n;
161
- size_t const nEnd = nbSymbols;
162
- for (n=0; n<nEnd; n++) {
163
- size_t const w = huffWeight[n];
164
- size_t const length = (1 << w) >> 1;
165
- size_t const uStart = rankVal[w];
166
- size_t const uEnd = uStart + length;
167
- size_t u;
168
- HUF_DEltX1 D;
169
- D.byte = (BYTE)n;
170
- D.nbBits = (BYTE)(tableLog + 1 - w);
171
- rankVal[w] = (U32)uEnd;
172
- if (length < 4) {
173
- /* Use length in the loop bound so the compiler knows it is short. */
174
- for (u = 0; u < length; ++u)
175
- dt[uStart + u] = D;
176
- } else {
177
- /* Unroll the loop 4 times, we know it is a power of 2. */
178
- for (u = uStart; u < uEnd; u += 4) {
179
- dt[u + 0] = D;
180
- dt[u + 1] = D;
181
- dt[u + 2] = D;
182
- dt[u + 3] = D;
183
- } } } }
208
+ /* fill DTable
209
+ * We fill all entries of each weight in order.
210
+ * That way length is a constant for each iteration of the outter loop.
211
+ * We can switch based on the length to a different inner loop which is
212
+ * optimized for that particular case.
213
+ */
214
+ {
215
+ U32 w;
216
+ int symbol=wksp->rankVal[0];
217
+ int rankStart=0;
218
+ for (w=1; w<tableLog+1; ++w) {
219
+ int const symbolCount = wksp->rankVal[w];
220
+ int const length = (1 << w) >> 1;
221
+ int uStart = rankStart;
222
+ BYTE const nbBits = (BYTE)(tableLog + 1 - w);
223
+ int s;
224
+ int u;
225
+ switch (length) {
226
+ case 1:
227
+ for (s=0; s<symbolCount; ++s) {
228
+ HUF_DEltX1 D;
229
+ D.byte = wksp->symbols[symbol + s];
230
+ D.nbBits = nbBits;
231
+ dt[uStart] = D;
232
+ uStart += 1;
233
+ }
234
+ break;
235
+ case 2:
236
+ for (s=0; s<symbolCount; ++s) {
237
+ HUF_DEltX1 D;
238
+ D.byte = wksp->symbols[symbol + s];
239
+ D.nbBits = nbBits;
240
+ dt[uStart+0] = D;
241
+ dt[uStart+1] = D;
242
+ uStart += 2;
243
+ }
244
+ break;
245
+ case 4:
246
+ for (s=0; s<symbolCount; ++s) {
247
+ U64 const D4 = HUF_DEltX1_set4(wksp->symbols[symbol + s], nbBits);
248
+ MEM_write64(dt + uStart, D4);
249
+ uStart += 4;
250
+ }
251
+ break;
252
+ case 8:
253
+ for (s=0; s<symbolCount; ++s) {
254
+ U64 const D4 = HUF_DEltX1_set4(wksp->symbols[symbol + s], nbBits);
255
+ MEM_write64(dt + uStart, D4);
256
+ MEM_write64(dt + uStart + 4, D4);
257
+ uStart += 8;
258
+ }
259
+ break;
260
+ default:
261
+ for (s=0; s<symbolCount; ++s) {
262
+ U64 const D4 = HUF_DEltX1_set4(wksp->symbols[symbol + s], nbBits);
263
+ for (u=0; u < length; u += 16) {
264
+ MEM_write64(dt + uStart + u + 0, D4);
265
+ MEM_write64(dt + uStart + u + 4, D4);
266
+ MEM_write64(dt + uStart + u + 8, D4);
267
+ MEM_write64(dt + uStart + u + 12, D4);
268
+ }
269
+ assert(u == length);
270
+ uStart += length;
271
+ }
272
+ break;
273
+ }
274
+ symbol += symbolCount;
275
+ rankStart += symbolCount * length;
276
+ }
277
+ }
184
278
  return iSize;
185
279
  }
186
280
 
187
- size_t HUF_readDTableX1(HUF_DTable* DTable, const void* src, size_t srcSize)
188
- {
189
- U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
190
- return HUF_readDTableX1_wksp(DTable, src, srcSize,
191
- workSpace, sizeof(workSpace));
192
- }
193
-
194
281
  FORCE_INLINE_TEMPLATE BYTE
195
282
  HUF_decodeSymbolX1(BIT_DStream_t* Dstream, const HUF_DEltX1* dt, const U32 dtLog)
196
283
  {
@@ -389,20 +476,6 @@ size_t HUF_decompress1X1_DCtx_wksp(HUF_DTable* DCtx, void* dst, size_t dstSize,
389
476
  }
390
477
 
391
478
 
392
- size_t HUF_decompress1X1_DCtx(HUF_DTable* DCtx, void* dst, size_t dstSize,
393
- const void* cSrc, size_t cSrcSize)
394
- {
395
- U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
396
- return HUF_decompress1X1_DCtx_wksp(DCtx, dst, dstSize, cSrc, cSrcSize,
397
- workSpace, sizeof(workSpace));
398
- }
399
-
400
- size_t HUF_decompress1X1 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
401
- {
402
- HUF_CREATE_STATIC_DTABLEX1(DTable, HUF_TABLELOG_MAX);
403
- return HUF_decompress1X1_DCtx (DTable, dst, dstSize, cSrc, cSrcSize);
404
- }
405
-
406
479
  size_t HUF_decompress4X1_usingDTable(
407
480
  void* dst, size_t dstSize,
408
481
  const void* cSrc, size_t cSrcSize,
@@ -419,8 +492,7 @@ static size_t HUF_decompress4X1_DCtx_wksp_bmi2(HUF_DTable* dctx, void* dst, size
419
492
  {
420
493
  const BYTE* ip = (const BYTE*) cSrc;
421
494
 
422
- size_t const hSize = HUF_readDTableX1_wksp (dctx, cSrc, cSrcSize,
423
- workSpace, wkspSize);
495
+ size_t const hSize = HUF_readDTableX1_wksp_bmi2(dctx, cSrc, cSrcSize, workSpace, wkspSize, bmi2);
424
496
  if (HUF_isError(hSize)) return hSize;
425
497
  if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
426
498
  ip += hSize; cSrcSize -= hSize;
@@ -436,18 +508,6 @@ size_t HUF_decompress4X1_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
436
508
  }
437
509
 
438
510
 
439
- size_t HUF_decompress4X1_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
440
- {
441
- U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
442
- return HUF_decompress4X1_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
443
- workSpace, sizeof(workSpace));
444
- }
445
- size_t HUF_decompress4X1 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
446
- {
447
- HUF_CREATE_STATIC_DTABLEX1(DTable, HUF_TABLELOG_MAX);
448
- return HUF_decompress4X1_DCtx(DTable, dst, dstSize, cSrc, cSrcSize);
449
- }
450
-
451
511
  #endif /* HUF_FORCE_DECOMPRESS_X2 */
452
512
 
453
513
 
@@ -474,7 +534,7 @@ static void HUF_fillDTableX2Level2(HUF_DEltX2* DTable, U32 sizeLog, const U32 co
474
534
  U32 rankVal[HUF_TABLELOG_MAX + 1];
475
535
 
476
536
  /* get pre-calculated rankVal */
477
- memcpy(rankVal, rankValOrigin, sizeof(rankVal));
537
+ ZSTD_memcpy(rankVal, rankValOrigin, sizeof(rankVal));
478
538
 
479
539
  /* fill skipped values */
480
540
  if (minWeight>1) {
@@ -516,7 +576,7 @@ static void HUF_fillDTableX2(HUF_DEltX2* DTable, const U32 targetLog,
516
576
  const U32 minBits = nbBitsBaseline - maxWeight;
517
577
  U32 s;
518
578
 
519
- memcpy(rankVal, rankValOrigin, sizeof(rankVal));
579
+ ZSTD_memcpy(rankVal, rankValOrigin, sizeof(rankVal));
520
580
 
521
581
  /* fill DTable */
522
582
  for (s=0; s<sortedListSize; s++) {
@@ -581,11 +641,11 @@ size_t HUF_readDTableX2_wksp(HUF_DTable* DTable,
581
641
  if ((spaceUsed32 << 2) > wkspSize) return ERROR(tableLog_tooLarge);
582
642
 
583
643
  rankStart = rankStart0 + 1;
584
- memset(rankStats, 0, sizeof(U32) * (2 * HUF_TABLELOG_MAX + 2 + 1));
644
+ ZSTD_memset(rankStats, 0, sizeof(U32) * (2 * HUF_TABLELOG_MAX + 2 + 1));
585
645
 
586
646
  DEBUG_STATIC_ASSERT(sizeof(HUF_DEltX2) == sizeof(HUF_DTable)); /* if compiler fails here, assertion is wrong */
587
647
  if (maxTableLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
588
- /* memset(weightList, 0, sizeof(weightList)); */ /* is not necessary, even though some analyzer complain ... */
648
+ /* ZSTD_memset(weightList, 0, sizeof(weightList)); */ /* is not necessary, even though some analyzer complain ... */
589
649
 
590
650
  iSize = HUF_readStats(weightList, HUF_SYMBOLVALUE_MAX + 1, rankStats, &nbSymbols, &tableLog, src, srcSize);
591
651
  if (HUF_isError(iSize)) return iSize;
@@ -599,9 +659,9 @@ size_t HUF_readDTableX2_wksp(HUF_DTable* DTable,
599
659
  /* Get start index of each weight */
600
660
  { U32 w, nextRankStart = 0;
601
661
  for (w=1; w<maxW+1; w++) {
602
- U32 current = nextRankStart;
662
+ U32 curr = nextRankStart;
603
663
  nextRankStart += rankStats[w];
604
- rankStart[w] = current;
664
+ rankStart[w] = curr;
605
665
  }
606
666
  rankStart[0] = nextRankStart; /* put all 0w symbols at the end of sorted list*/
607
667
  sizeOfSort = nextRankStart;
@@ -624,9 +684,9 @@ size_t HUF_readDTableX2_wksp(HUF_DTable* DTable,
624
684
  U32 nextRankVal = 0;
625
685
  U32 w;
626
686
  for (w=1; w<maxW+1; w++) {
627
- U32 current = nextRankVal;
687
+ U32 curr = nextRankVal;
628
688
  nextRankVal += rankStats[w] << (w+rescale);
629
- rankVal0[w] = current;
689
+ rankVal0[w] = curr;
630
690
  } }
631
691
  { U32 const minBits = tableLog+1 - maxW;
632
692
  U32 consumed;
@@ -644,23 +704,16 @@ size_t HUF_readDTableX2_wksp(HUF_DTable* DTable,
644
704
 
645
705
  dtd.tableLog = (BYTE)maxTableLog;
646
706
  dtd.tableType = 1;
647
- memcpy(DTable, &dtd, sizeof(dtd));
707
+ ZSTD_memcpy(DTable, &dtd, sizeof(dtd));
648
708
  return iSize;
649
709
  }
650
710
 
651
- size_t HUF_readDTableX2(HUF_DTable* DTable, const void* src, size_t srcSize)
652
- {
653
- U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
654
- return HUF_readDTableX2_wksp(DTable, src, srcSize,
655
- workSpace, sizeof(workSpace));
656
- }
657
-
658
711
 
659
712
  FORCE_INLINE_TEMPLATE U32
660
713
  HUF_decodeSymbolX2(void* op, BIT_DStream_t* DStream, const HUF_DEltX2* dt, const U32 dtLog)
661
714
  {
662
715
  size_t const val = BIT_lookBitsFast(DStream, dtLog); /* note : dtLog >= 1 */
663
- memcpy(op, dt+val, 2);
716
+ ZSTD_memcpy(op, dt+val, 2);
664
717
  BIT_skipBits(DStream, dt[val].nbBits);
665
718
  return dt[val].length;
666
719
  }
@@ -669,7 +722,7 @@ FORCE_INLINE_TEMPLATE U32
669
722
  HUF_decodeLastSymbolX2(void* op, BIT_DStream_t* DStream, const HUF_DEltX2* dt, const U32 dtLog)
670
723
  {
671
724
  size_t const val = BIT_lookBitsFast(DStream, dtLog); /* note : dtLog >= 1 */
672
- memcpy(op, dt+val, 1);
725
+ ZSTD_memcpy(op, dt+val, 1);
673
726
  if (dt[val].length==1) BIT_skipBits(DStream, dt[val].nbBits);
674
727
  else {
675
728
  if (DStream->bitsConsumed < (sizeof(DStream->bitContainer)*8)) {
@@ -890,20 +943,6 @@ size_t HUF_decompress1X2_DCtx_wksp(HUF_DTable* DCtx, void* dst, size_t dstSize,
890
943
  }
891
944
 
892
945
 
893
- size_t HUF_decompress1X2_DCtx(HUF_DTable* DCtx, void* dst, size_t dstSize,
894
- const void* cSrc, size_t cSrcSize)
895
- {
896
- U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
897
- return HUF_decompress1X2_DCtx_wksp(DCtx, dst, dstSize, cSrc, cSrcSize,
898
- workSpace, sizeof(workSpace));
899
- }
900
-
901
- size_t HUF_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
902
- {
903
- HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_TABLELOG_MAX);
904
- return HUF_decompress1X2_DCtx(DTable, dst, dstSize, cSrc, cSrcSize);
905
- }
906
-
907
946
  size_t HUF_decompress4X2_usingDTable(
908
947
  void* dst, size_t dstSize,
909
948
  const void* cSrc, size_t cSrcSize,
@@ -937,20 +976,6 @@ size_t HUF_decompress4X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
937
976
  }
938
977
 
939
978
 
940
- size_t HUF_decompress4X2_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize,
941
- const void* cSrc, size_t cSrcSize)
942
- {
943
- U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
944
- return HUF_decompress4X2_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
945
- workSpace, sizeof(workSpace));
946
- }
947
-
948
- size_t HUF_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
949
- {
950
- HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_TABLELOG_MAX);
951
- return HUF_decompress4X2_DCtx(DTable, dst, dstSize, cSrc, cSrcSize);
952
- }
953
-
954
979
  #endif /* HUF_FORCE_DECOMPRESS_X1 */
955
980
 
956
981
 
@@ -1051,67 +1076,6 @@ U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize)
1051
1076
  }
1052
1077
 
1053
1078
 
1054
- typedef size_t (*decompressionAlgo)(void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);
1055
-
1056
- size_t HUF_decompress (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
1057
- {
1058
- #if !defined(HUF_FORCE_DECOMPRESS_X1) && !defined(HUF_FORCE_DECOMPRESS_X2)
1059
- static const decompressionAlgo decompress[2] = { HUF_decompress4X1, HUF_decompress4X2 };
1060
- #endif
1061
-
1062
- /* validation checks */
1063
- if (dstSize == 0) return ERROR(dstSize_tooSmall);
1064
- if (cSrcSize > dstSize) return ERROR(corruption_detected); /* invalid */
1065
- if (cSrcSize == dstSize) { memcpy(dst, cSrc, dstSize); return dstSize; } /* not compressed */
1066
- if (cSrcSize == 1) { memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; } /* RLE */
1067
-
1068
- { U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
1069
- #if defined(HUF_FORCE_DECOMPRESS_X1)
1070
- (void)algoNb;
1071
- assert(algoNb == 0);
1072
- return HUF_decompress4X1(dst, dstSize, cSrc, cSrcSize);
1073
- #elif defined(HUF_FORCE_DECOMPRESS_X2)
1074
- (void)algoNb;
1075
- assert(algoNb == 1);
1076
- return HUF_decompress4X2(dst, dstSize, cSrc, cSrcSize);
1077
- #else
1078
- return decompress[algoNb](dst, dstSize, cSrc, cSrcSize);
1079
- #endif
1080
- }
1081
- }
1082
-
1083
- size_t HUF_decompress4X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
1084
- {
1085
- /* validation checks */
1086
- if (dstSize == 0) return ERROR(dstSize_tooSmall);
1087
- if (cSrcSize > dstSize) return ERROR(corruption_detected); /* invalid */
1088
- if (cSrcSize == dstSize) { memcpy(dst, cSrc, dstSize); return dstSize; } /* not compressed */
1089
- if (cSrcSize == 1) { memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; } /* RLE */
1090
-
1091
- { U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
1092
- #if defined(HUF_FORCE_DECOMPRESS_X1)
1093
- (void)algoNb;
1094
- assert(algoNb == 0);
1095
- return HUF_decompress4X1_DCtx(dctx, dst, dstSize, cSrc, cSrcSize);
1096
- #elif defined(HUF_FORCE_DECOMPRESS_X2)
1097
- (void)algoNb;
1098
- assert(algoNb == 1);
1099
- return HUF_decompress4X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize);
1100
- #else
1101
- return algoNb ? HUF_decompress4X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) :
1102
- HUF_decompress4X1_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) ;
1103
- #endif
1104
- }
1105
- }
1106
-
1107
- size_t HUF_decompress4X_hufOnly(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
1108
- {
1109
- U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
1110
- return HUF_decompress4X_hufOnly_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
1111
- workSpace, sizeof(workSpace));
1112
- }
1113
-
1114
-
1115
1079
  size_t HUF_decompress4X_hufOnly_wksp(HUF_DTable* dctx, void* dst,
1116
1080
  size_t dstSize, const void* cSrc,
1117
1081
  size_t cSrcSize, void* workSpace,
@@ -1145,8 +1109,8 @@ size_t HUF_decompress1X_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
1145
1109
  /* validation checks */
1146
1110
  if (dstSize == 0) return ERROR(dstSize_tooSmall);
1147
1111
  if (cSrcSize > dstSize) return ERROR(corruption_detected); /* invalid */
1148
- if (cSrcSize == dstSize) { memcpy(dst, cSrc, dstSize); return dstSize; } /* not compressed */
1149
- if (cSrcSize == 1) { memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; } /* RLE */
1112
+ if (cSrcSize == dstSize) { ZSTD_memcpy(dst, cSrc, dstSize); return dstSize; } /* not compressed */
1113
+ if (cSrcSize == 1) { ZSTD_memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; } /* RLE */
1150
1114
 
1151
1115
  { U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
1152
1116
  #if defined(HUF_FORCE_DECOMPRESS_X1)
@@ -1168,14 +1132,6 @@ size_t HUF_decompress1X_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
1168
1132
  }
1169
1133
  }
1170
1134
 
1171
- size_t HUF_decompress1X_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize,
1172
- const void* cSrc, size_t cSrcSize)
1173
- {
1174
- U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
1175
- return HUF_decompress1X_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
1176
- workSpace, sizeof(workSpace));
1177
- }
1178
-
1179
1135
 
1180
1136
  size_t HUF_decompress1X_usingDTable_bmi2(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable, int bmi2)
1181
1137
  {
@@ -1199,7 +1155,7 @@ size_t HUF_decompress1X1_DCtx_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t dstS
1199
1155
  {
1200
1156
  const BYTE* ip = (const BYTE*) cSrc;
1201
1157
 
1202
- size_t const hSize = HUF_readDTableX1_wksp(dctx, cSrc, cSrcSize, workSpace, wkspSize);
1158
+ size_t const hSize = HUF_readDTableX1_wksp_bmi2(dctx, cSrc, cSrcSize, workSpace, wkspSize, bmi2);
1203
1159
  if (HUF_isError(hSize)) return hSize;
1204
1160
  if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
1205
1161
  ip += hSize; cSrcSize -= hSize;
@@ -1246,3 +1202,149 @@ size_t HUF_decompress4X_hufOnly_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t ds
1246
1202
  #endif
1247
1203
  }
1248
1204
  }
1205
+
1206
+ #ifndef ZSTD_NO_UNUSED_FUNCTIONS
1207
+ #ifndef HUF_FORCE_DECOMPRESS_X2
1208
+ size_t HUF_readDTableX1(HUF_DTable* DTable, const void* src, size_t srcSize)
1209
+ {
1210
+ U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
1211
+ return HUF_readDTableX1_wksp(DTable, src, srcSize,
1212
+ workSpace, sizeof(workSpace));
1213
+ }
1214
+
1215
+ size_t HUF_decompress1X1_DCtx(HUF_DTable* DCtx, void* dst, size_t dstSize,
1216
+ const void* cSrc, size_t cSrcSize)
1217
+ {
1218
+ U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
1219
+ return HUF_decompress1X1_DCtx_wksp(DCtx, dst, dstSize, cSrc, cSrcSize,
1220
+ workSpace, sizeof(workSpace));
1221
+ }
1222
+
1223
+ size_t HUF_decompress1X1 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
1224
+ {
1225
+ HUF_CREATE_STATIC_DTABLEX1(DTable, HUF_TABLELOG_MAX);
1226
+ return HUF_decompress1X1_DCtx (DTable, dst, dstSize, cSrc, cSrcSize);
1227
+ }
1228
+ #endif
1229
+
1230
+ #ifndef HUF_FORCE_DECOMPRESS_X1
1231
+ size_t HUF_readDTableX2(HUF_DTable* DTable, const void* src, size_t srcSize)
1232
+ {
1233
+ U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
1234
+ return HUF_readDTableX2_wksp(DTable, src, srcSize,
1235
+ workSpace, sizeof(workSpace));
1236
+ }
1237
+
1238
+ size_t HUF_decompress1X2_DCtx(HUF_DTable* DCtx, void* dst, size_t dstSize,
1239
+ const void* cSrc, size_t cSrcSize)
1240
+ {
1241
+ U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
1242
+ return HUF_decompress1X2_DCtx_wksp(DCtx, dst, dstSize, cSrc, cSrcSize,
1243
+ workSpace, sizeof(workSpace));
1244
+ }
1245
+
1246
+ size_t HUF_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
1247
+ {
1248
+ HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_TABLELOG_MAX);
1249
+ return HUF_decompress1X2_DCtx(DTable, dst, dstSize, cSrc, cSrcSize);
1250
+ }
1251
+ #endif
1252
+
1253
+ #ifndef HUF_FORCE_DECOMPRESS_X2
1254
+ size_t HUF_decompress4X1_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
1255
+ {
1256
+ U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
1257
+ return HUF_decompress4X1_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
1258
+ workSpace, sizeof(workSpace));
1259
+ }
1260
+ size_t HUF_decompress4X1 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
1261
+ {
1262
+ HUF_CREATE_STATIC_DTABLEX1(DTable, HUF_TABLELOG_MAX);
1263
+ return HUF_decompress4X1_DCtx(DTable, dst, dstSize, cSrc, cSrcSize);
1264
+ }
1265
+ #endif
1266
+
1267
+ #ifndef HUF_FORCE_DECOMPRESS_X1
1268
+ size_t HUF_decompress4X2_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize,
1269
+ const void* cSrc, size_t cSrcSize)
1270
+ {
1271
+ U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
1272
+ return HUF_decompress4X2_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
1273
+ workSpace, sizeof(workSpace));
1274
+ }
1275
+
1276
+ size_t HUF_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
1277
+ {
1278
+ HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_TABLELOG_MAX);
1279
+ return HUF_decompress4X2_DCtx(DTable, dst, dstSize, cSrc, cSrcSize);
1280
+ }
1281
+ #endif
1282
+
1283
+ typedef size_t (*decompressionAlgo)(void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);
1284
+
1285
+ size_t HUF_decompress (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
1286
+ {
1287
+ #if !defined(HUF_FORCE_DECOMPRESS_X1) && !defined(HUF_FORCE_DECOMPRESS_X2)
1288
+ static const decompressionAlgo decompress[2] = { HUF_decompress4X1, HUF_decompress4X2 };
1289
+ #endif
1290
+
1291
+ /* validation checks */
1292
+ if (dstSize == 0) return ERROR(dstSize_tooSmall);
1293
+ if (cSrcSize > dstSize) return ERROR(corruption_detected); /* invalid */
1294
+ if (cSrcSize == dstSize) { ZSTD_memcpy(dst, cSrc, dstSize); return dstSize; } /* not compressed */
1295
+ if (cSrcSize == 1) { ZSTD_memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; } /* RLE */
1296
+
1297
+ { U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
1298
+ #if defined(HUF_FORCE_DECOMPRESS_X1)
1299
+ (void)algoNb;
1300
+ assert(algoNb == 0);
1301
+ return HUF_decompress4X1(dst, dstSize, cSrc, cSrcSize);
1302
+ #elif defined(HUF_FORCE_DECOMPRESS_X2)
1303
+ (void)algoNb;
1304
+ assert(algoNb == 1);
1305
+ return HUF_decompress4X2(dst, dstSize, cSrc, cSrcSize);
1306
+ #else
1307
+ return decompress[algoNb](dst, dstSize, cSrc, cSrcSize);
1308
+ #endif
1309
+ }
1310
+ }
1311
+
1312
+ size_t HUF_decompress4X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
1313
+ {
1314
+ /* validation checks */
1315
+ if (dstSize == 0) return ERROR(dstSize_tooSmall);
1316
+ if (cSrcSize > dstSize) return ERROR(corruption_detected); /* invalid */
1317
+ if (cSrcSize == dstSize) { ZSTD_memcpy(dst, cSrc, dstSize); return dstSize; } /* not compressed */
1318
+ if (cSrcSize == 1) { ZSTD_memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; } /* RLE */
1319
+
1320
+ { U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
1321
+ #if defined(HUF_FORCE_DECOMPRESS_X1)
1322
+ (void)algoNb;
1323
+ assert(algoNb == 0);
1324
+ return HUF_decompress4X1_DCtx(dctx, dst, dstSize, cSrc, cSrcSize);
1325
+ #elif defined(HUF_FORCE_DECOMPRESS_X2)
1326
+ (void)algoNb;
1327
+ assert(algoNb == 1);
1328
+ return HUF_decompress4X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize);
1329
+ #else
1330
+ return algoNb ? HUF_decompress4X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) :
1331
+ HUF_decompress4X1_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) ;
1332
+ #endif
1333
+ }
1334
+ }
1335
+
1336
+ size_t HUF_decompress4X_hufOnly(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
1337
+ {
1338
+ U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
1339
+ return HUF_decompress4X_hufOnly_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
1340
+ workSpace, sizeof(workSpace));
1341
+ }
1342
+
1343
+ size_t HUF_decompress1X_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize,
1344
+ const void* cSrc, size_t cSrcSize)
1345
+ {
1346
+ U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
1347
+ return HUF_decompress1X_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
1348
+ workSpace, sizeof(workSpace));
1349
+ }
1350
+ #endif