2 Common functions of New Generation Entropy library
3 Copyright (C) 2016, Yann Collet.
5 BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
7 Redistribution and use in source and binary forms, with or without
8 modification, are permitted provided that the following conditions are
11 * Redistributions of source code must retain the above copyright
12 notice, this list of conditions and the following disclaimer.
13 * Redistributions in binary form must reproduce the above
14 copyright notice, this list of conditions and the following disclaimer
15 in the documentation and/or other materials provided with the
18 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
21 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
22 OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
23 SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
24 LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
25 DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
26 THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
28 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30 You can contact the author at :
31 - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
32 - Public forum : https://groups.google.com/forum/#!forum/lz4c
33 *************************************************************************** */
35 /* *************************************
37 ***************************************/
39 #include "error_private.h" /* ERR_*, ERROR */
40 #define FSE_STATIC_LINKING_ONLY /* FSE_MIN_TABLELOG */
42 #define HUF_STATIC_LINKING_ONLY /* HUF_TABLELOG_ABSOLUTEMAX */
47 unsigned FSE_versionNumber(void) { return FSE_VERSION_NUMBER; }
50 /*=== Error Management ===*/
51 unsigned FSE_isError(size_t code) { return ERR_isError(code); }
52 const char* FSE_getErrorName(size_t code) { return ERR_getErrorName(code); }
54 unsigned HUF_isError(size_t code) { return ERR_isError(code); }
55 const char* HUF_getErrorName(size_t code) { return ERR_getErrorName(code); }
58 /*-**************************************************************
59 * FSE NCount encoding-decoding
60 ****************************************************************/
61 size_t FSE_readNCount (short* normalizedCounter, unsigned* maxSVPtr, unsigned* tableLogPtr,
62 const void* headerBuffer, size_t hbSize)
64 const BYTE* const istart = (const BYTE*) headerBuffer;
65 const BYTE* const iend = istart + hbSize;
66 const BYTE* ip = istart;
76 /* This function only works when hbSize >= 4 */
78 memset(buffer, 0, sizeof(buffer));
79 memcpy(buffer, headerBuffer, hbSize);
80 { size_t const countSize = FSE_readNCount(normalizedCounter, maxSVPtr, tableLogPtr,
81 buffer, sizeof(buffer));
82 if (FSE_isError(countSize)) return countSize;
83 if (countSize > hbSize) return ERROR(corruption_detected);
89 memset(normalizedCounter, 0, (*maxSVPtr+1) * sizeof(normalizedCounter[0])); /* all symbols not present in NCount have a frequency of 0 */
90 bitStream = MEM_readLE32(ip);
91 nbBits = (bitStream & 0xF) + FSE_MIN_TABLELOG; /* extract tableLog */
92 if (nbBits > FSE_TABLELOG_ABSOLUTE_MAX) return ERROR(tableLog_tooLarge);
95 *tableLogPtr = nbBits;
96 remaining = (1<<nbBits)+1;
97 threshold = 1<<nbBits;
100 while ((remaining>1) & (charnum<=*maxSVPtr)) {
102 unsigned n0 = charnum;
103 while ((bitStream & 0xFFFF) == 0xFFFF) {
107 bitStream = MEM_readLE32(ip) >> bitCount;
112 while ((bitStream & 3) == 3) {
119 if (n0 > *maxSVPtr) return ERROR(maxSymbolValue_tooSmall);
120 while (charnum < n0) normalizedCounter[charnum++] = 0;
121 if ((ip <= iend-7) || (ip + (bitCount>>3) <= iend-4)) {
122 assert((bitCount >> 3) <= 3); /* For first condition to work */
125 bitStream = MEM_readLE32(ip) >> bitCount;
129 { int const max = (2*threshold-1) - remaining;
132 if ((bitStream & (threshold-1)) < (U32)max) {
133 count = bitStream & (threshold-1);
134 bitCount += nbBits-1;
136 count = bitStream & (2*threshold-1);
137 if (count >= threshold) count -= max;
141 count--; /* extra accuracy */
142 remaining -= count < 0 ? -count : count; /* -1 means +1 */
143 normalizedCounter[charnum++] = (short)count;
145 while (remaining < threshold) {
150 if ((ip <= iend-7) || (ip + (bitCount>>3) <= iend-4)) {
154 bitCount -= (int)(8 * (iend - 4 - ip));
157 bitStream = MEM_readLE32(ip) >> (bitCount & 31);
158 } } /* while ((remaining>1) & (charnum<=*maxSVPtr)) */
159 if (remaining != 1) return ERROR(corruption_detected);
160 if (bitCount > 32) return ERROR(corruption_detected);
161 *maxSVPtr = charnum-1;
163 ip += (bitCount+7)>>3;
168 /*! HUF_readStats() :
169 Read compact Huffman tree, saved by HUF_writeCTable().
170 `huffWeight` is destination buffer.
171 `rankStats` is assumed to be a table of at least HUF_TABLELOG_MAX U32.
172 @return : size read from `src` , or an error Code .
173 Note : Needed by HUF_readCTable() and HUF_readDTableX?() .
175 size_t HUF_readStats(BYTE* huffWeight, size_t hwSize, U32* rankStats,
176 U32* nbSymbolsPtr, U32* tableLogPtr,
177 const void* src, size_t srcSize)
180 const BYTE* ip = (const BYTE*) src;
184 if (!srcSize) return ERROR(srcSize_wrong);
186 /* memset(huffWeight, 0, hwSize); *//* is not necessary, even though some analyzer complain ... */
188 if (iSize >= 128) { /* special header */
190 iSize = ((oSize+1)/2);
191 if (iSize+1 > srcSize) return ERROR(srcSize_wrong);
192 if (oSize >= hwSize) return ERROR(corruption_detected);
195 for (n=0; n<oSize; n+=2) {
196 huffWeight[n] = ip[n/2] >> 4;
197 huffWeight[n+1] = ip[n/2] & 15;
199 else { /* header compressed with FSE (normal case) */
200 FSE_DTable fseWorkspace[FSE_DTABLE_SIZE_U32(6)]; /* 6 is max possible tableLog for HUF header (maybe even 5, to be tested) */
201 if (iSize+1 > srcSize) return ERROR(srcSize_wrong);
202 oSize = FSE_decompress_wksp(huffWeight, hwSize-1, ip+1, iSize, fseWorkspace, 6); /* max (hwSize-1) values decoded, as last one is implied */
203 if (FSE_isError(oSize)) return oSize;
206 /* collect weight stats */
207 memset(rankStats, 0, (HUF_TABLELOG_MAX + 1) * sizeof(U32));
209 { U32 n; for (n=0; n<oSize; n++) {
210 if (huffWeight[n] >= HUF_TABLELOG_MAX) return ERROR(corruption_detected);
211 rankStats[huffWeight[n]]++;
212 weightTotal += (1 << huffWeight[n]) >> 1;
214 if (weightTotal == 0) return ERROR(corruption_detected);
216 /* get last non-null symbol weight (implied, total must be 2^n) */
217 { U32 const tableLog = BIT_highbit32(weightTotal) + 1;
218 if (tableLog > HUF_TABLELOG_MAX) return ERROR(corruption_detected);
219 *tableLogPtr = tableLog;
220 /* determine last weight */
221 { U32 const total = 1 << tableLog;
222 U32 const rest = total - weightTotal;
223 U32 const verif = 1 << BIT_highbit32(rest);
224 U32 const lastWeight = BIT_highbit32(rest) + 1;
225 if (verif != rest) return ERROR(corruption_detected); /* last value must be a clean power of 2 */
226 huffWeight[oSize] = (BYTE)lastWeight;
227 rankStats[lastWeight]++;
230 /* check tree construction validity */
231 if ((rankStats[1] < 2) || (rankStats[1] & 1)) return ERROR(corruption_detected); /* by construction : at least 2 elts of rank 1, must be even */
234 *nbSymbolsPtr = (U32)(oSize+1);