diff options
author | Ruslan Kovalev <ruslan.a.kovalev@gmail.com> | 2022-02-10 16:46:44 +0300 |
---|---|---|
committer | Daniil Cherednik <dcherednik@yandex-team.ru> | 2022-02-10 16:46:44 +0300 |
commit | 59e19371de37995fcb36beb16cd6ec030af960bc (patch) | |
tree | fa68e36093ebff8b805462e9e6d331fe9d348214 /contrib/libs/zstd/lib/compress/huf_compress.c | |
parent | 89db6fe2fe2c32d2a832ddfeb04e8d078e301084 (diff) | |
download | ydb-59e19371de37995fcb36beb16cd6ec030af960bc.tar.gz |
Restoring authorship annotation for Ruslan Kovalev <ruslan.a.kovalev@gmail.com>. Commit 1 of 2.
Diffstat (limited to 'contrib/libs/zstd/lib/compress/huf_compress.c')
-rw-r--r-- | contrib/libs/zstd/lib/compress/huf_compress.c | 540 |
1 files changed, 270 insertions, 270 deletions
diff --git a/contrib/libs/zstd/lib/compress/huf_compress.c b/contrib/libs/zstd/lib/compress/huf_compress.c index 2b3d6adc2a..137671d31a 100644 --- a/contrib/libs/zstd/lib/compress/huf_compress.c +++ b/contrib/libs/zstd/lib/compress/huf_compress.c @@ -1,4 +1,4 @@ -/* ****************************************************************** +/* ****************************************************************** * Huffman encoder, part of New Generation Entropy library * Copyright (c) Yann Collet, Facebook, Inc. * @@ -10,49 +10,49 @@ * LICENSE file in the root directory of this source tree) and the GPLv2 (found * in the COPYING file in the root directory of this source tree). * You may select, at your option, one of the above-listed licenses. -****************************************************************** */ - -/* ************************************************************** -* Compiler specifics -****************************************************************/ -#ifdef _MSC_VER /* Visual Studio */ -# pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */ -#endif - - -/* ************************************************************** -* Includes -****************************************************************/ +****************************************************************** */ + +/* ************************************************************** +* Compiler specifics +****************************************************************/ +#ifdef _MSC_VER /* Visual Studio */ +# pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */ +#endif + + +/* ************************************************************** +* Includes +****************************************************************/ #include "../common/zstd_deps.h" /* ZSTD_memcpy, ZSTD_memset */ #include "../common/compiler.h" #include "../common/bitstream.h" #include "hist.h" -#define FSE_STATIC_LINKING_ONLY /* FSE_optimalTableLog_internal */ +#define FSE_STATIC_LINKING_ONLY /* FSE_optimalTableLog_internal */ #include "../common/fse.h" /* header compression */ -#define HUF_STATIC_LINKING_ONLY +#define HUF_STATIC_LINKING_ONLY #include "../common/huf.h" #include "../common/error_private.h" - - -/* ************************************************************** -* Error Management -****************************************************************/ + + +/* ************************************************************** +* Error Management +****************************************************************/ #define HUF_isError ERR_isError #define HUF_STATIC_ASSERT(c) DEBUG_STATIC_ASSERT(c) /* use only *after* variable declarations */ - - -/* ************************************************************** -* Utils -****************************************************************/ -unsigned HUF_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue) -{ - return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 1); -} - - -/* ******************************************************* -* HUF : Huffman block compression -*********************************************************/ + + +/* ************************************************************** +* Utils +****************************************************************/ +unsigned HUF_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue) +{ + return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 1); +} + + +/* ******************************************************* +* HUF : Huffman block compression +*********************************************************/ #define HUF_WORKSPACE_MAX_ALIGNMENT 8 static void* HUF_alignUpWorkspace(void* workspace, size_t* workspaceSizePtr, size_t align) @@ -166,7 +166,7 @@ static void HUF_setValue(HUF_CElt* elt, size_t value) typedef struct { HUF_CompressWeightsWksp wksp; BYTE bitsToWeight[HUF_TABLELOG_MAX + 1]; /* precomputed conversion table */ - BYTE huffWeight[HUF_SYMBOLVALUE_MAX]; + BYTE huffWeight[HUF_SYMBOLVALUE_MAX]; } HUF_WriteCTableWksp; size_t HUF_writeCTable_wksp(void* dst, size_t maxDstSize, @@ -174,39 +174,39 @@ size_t HUF_writeCTable_wksp(void* dst, size_t maxDstSize, void* workspace, size_t workspaceSize) { HUF_CElt const* const ct = CTable + 1; - BYTE* op = (BYTE*)dst; - U32 n; + BYTE* op = (BYTE*)dst; + U32 n; HUF_WriteCTableWksp* wksp = (HUF_WriteCTableWksp*)HUF_alignUpWorkspace(workspace, &workspaceSize, ZSTD_ALIGNOF(U32)); - + /* check conditions */ if (workspaceSize < sizeof(HUF_WriteCTableWksp)) return ERROR(GENERIC); if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge); - - /* convert to weight */ + + /* convert to weight */ wksp->bitsToWeight[0] = 0; - for (n=1; n<huffLog+1; n++) + for (n=1; n<huffLog+1; n++) wksp->bitsToWeight[n] = (BYTE)(huffLog + 1 - n); - for (n=0; n<maxSymbolValue; n++) + for (n=0; n<maxSymbolValue; n++) wksp->huffWeight[n] = wksp->bitsToWeight[HUF_getNbBits(ct[n])]; - + /* attempt weights compression by FSE */ if (maxDstSize < 1) return ERROR(dstSize_tooSmall); { CHECK_V_F(hSize, HUF_compressWeights(op+1, maxDstSize-1, wksp->huffWeight, maxSymbolValue, &wksp->wksp, sizeof(wksp->wksp)) ); if ((hSize>1) & (hSize < maxSymbolValue/2)) { /* FSE compressed */ op[0] = (BYTE)hSize; return hSize+1; - } } - + } } + /* write raw values as 4-bits (max : 15) */ if (maxSymbolValue > (256-128)) return ERROR(GENERIC); /* should not happen : likely means source cannot be compressed */ - if (((maxSymbolValue+1)/2) + 1 > maxDstSize) return ERROR(dstSize_tooSmall); /* not enough space within dst buffer */ - op[0] = (BYTE)(128 /*special case*/ + (maxSymbolValue-1)); + if (((maxSymbolValue+1)/2) + 1 > maxDstSize) return ERROR(dstSize_tooSmall); /* not enough space within dst buffer */ + op[0] = (BYTE)(128 /*special case*/ + (maxSymbolValue-1)); wksp->huffWeight[maxSymbolValue] = 0; /* to be sure it doesn't cause msan issue in final combination */ - for (n=0; n<maxSymbolValue; n+=2) + for (n=0; n<maxSymbolValue; n+=2) op[(n/2)+1] = (BYTE)((wksp->huffWeight[n] << 4) + wksp->huffWeight[n+1]); - return ((maxSymbolValue+1)/2) + 1; -} - + return ((maxSymbolValue+1)/2) + 1; +} + /*! HUF_writeCTable() : `CTable` : Huffman tree to save, using huf representation. @return : size of saved CTable */ @@ -216,67 +216,67 @@ size_t HUF_writeCTable (void* dst, size_t maxDstSize, HUF_WriteCTableWksp wksp; return HUF_writeCTable_wksp(dst, maxDstSize, CTable, maxSymbolValue, huffLog, &wksp, sizeof(wksp)); } - + size_t HUF_readCTable (HUF_CElt* CTable, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize, unsigned* hasZeroWeights) -{ +{ BYTE huffWeight[HUF_SYMBOLVALUE_MAX + 1]; /* init not required, even though some static analyzer may complain */ - U32 rankVal[HUF_TABLELOG_ABSOLUTEMAX + 1]; /* large enough for values from 0 to 16 */ - U32 tableLog = 0; - U32 nbSymbols = 0; + U32 rankVal[HUF_TABLELOG_ABSOLUTEMAX + 1]; /* large enough for values from 0 to 16 */ + U32 tableLog = 0; + U32 nbSymbols = 0; HUF_CElt* const ct = CTable + 1; - - /* get symbol weights */ + + /* get symbol weights */ CHECK_V_F(readSize, HUF_readStats(huffWeight, HUF_SYMBOLVALUE_MAX+1, rankVal, &nbSymbols, &tableLog, src, srcSize)); *hasZeroWeights = (rankVal[0] > 0); - - /* check result */ - if (tableLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge); + + /* check result */ + if (tableLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge); if (nbSymbols > *maxSymbolValuePtr+1) return ERROR(maxSymbolValue_tooSmall); - + CTable[0] = tableLog; - /* Prepare base value per rank */ - { U32 n, nextRankStart = 0; - for (n=1; n<=tableLog; n++) { + /* Prepare base value per rank */ + { U32 n, nextRankStart = 0; + for (n=1; n<=tableLog; n++) { U32 curr = nextRankStart; - nextRankStart += (rankVal[n] << (n-1)); + nextRankStart += (rankVal[n] << (n-1)); rankVal[n] = curr; - } } - - /* fill nbBits */ - { U32 n; for (n=0; n<nbSymbols; n++) { - const U32 w = huffWeight[n]; + } } + + /* fill nbBits */ + { U32 n; for (n=0; n<nbSymbols; n++) { + const U32 w = huffWeight[n]; HUF_setNbBits(ct + n, (BYTE)(tableLog + 1 - w) & -(w != 0)); - } } - - /* fill val */ + } } + + /* fill val */ { U16 nbPerRank[HUF_TABLELOG_MAX+2] = {0}; /* support w=0=>n=tableLog+1 */ U16 valPerRank[HUF_TABLELOG_MAX+2] = {0}; { U32 n; for (n=0; n<nbSymbols; n++) nbPerRank[HUF_getNbBits(ct[n])]++; } - /* determine stating value per rank */ + /* determine stating value per rank */ valPerRank[tableLog+1] = 0; /* for w==0 */ - { U16 min = 0; + { U16 min = 0; U32 n; for (n=tableLog; n>0; n--) { /* start at n=tablelog <-> w=1 */ valPerRank[n] = min; /* get starting value within each rank */ - min += nbPerRank[n]; - min >>= 1; - } } - /* assign value within rank, symbol order */ + min += nbPerRank[n]; + min >>= 1; + } } + /* assign value within rank, symbol order */ { U32 n; for (n=0; n<nbSymbols; n++) HUF_setValue(ct + n, valPerRank[HUF_getNbBits(ct[n])]++); } - } - + } + *maxSymbolValuePtr = nbSymbols - 1; - return readSize; -} - + return readSize; +} + U32 HUF_getNbBitsFromCTable(HUF_CElt const* CTable, U32 symbolValue) { const HUF_CElt* ct = CTable + 1; assert(symbolValue <= HUF_SYMBOLVALUE_MAX); return (U32)HUF_getNbBits(ct[symbolValue]); } - + typedef struct nodeElt_s { U32 count; @@ -305,83 +305,83 @@ typedef struct nodeElt_s { * @return The maximum number of bits of the Huffman tree after adjustment, * necessarily no more than maxNbBits. */ -static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits) -{ - const U32 largestBits = huffNode[lastNonNull].nbBits; +static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits) +{ + const U32 largestBits = huffNode[lastNonNull].nbBits; /* early exit : no elt > maxNbBits, so the tree is already valid. */ if (largestBits <= maxNbBits) return largestBits; - - /* there are several too large elements (at least >= 2) */ - { int totalCost = 0; - const U32 baseCost = 1 << (largestBits - maxNbBits); + + /* there are several too large elements (at least >= 2) */ + { int totalCost = 0; + const U32 baseCost = 1 << (largestBits - maxNbBits); int n = (int)lastNonNull; - + /* Adjust any ranks > maxNbBits to maxNbBits. * Compute totalCost, which is how far the sum of the ranks is * we are over 2^largestBits after adjust the offending ranks. */ - while (huffNode[n].nbBits > maxNbBits) { - totalCost += baseCost - (1 << (largestBits - huffNode[n].nbBits)); - huffNode[n].nbBits = (BYTE)maxNbBits; + while (huffNode[n].nbBits > maxNbBits) { + totalCost += baseCost - (1 << (largestBits - huffNode[n].nbBits)); + huffNode[n].nbBits = (BYTE)maxNbBits; n--; } /* n stops at huffNode[n].nbBits <= maxNbBits */ assert(huffNode[n].nbBits <= maxNbBits); /* n end at index of smallest symbol using < maxNbBits */ while (huffNode[n].nbBits == maxNbBits) --n; - + /* renorm totalCost from 2^largestBits to 2^maxNbBits * note : totalCost is necessarily a multiple of baseCost */ assert((totalCost & (baseCost - 1)) == 0); totalCost >>= (largestBits - maxNbBits); assert(totalCost > 0); - - /* repay normalized cost */ - { U32 const noSymbol = 0xF0F0F0F0; - U32 rankLast[HUF_TABLELOG_MAX+2]; - + + /* repay normalized cost */ + { U32 const noSymbol = 0xF0F0F0F0; + U32 rankLast[HUF_TABLELOG_MAX+2]; + /* Get pos of last (smallest = lowest cum. count) symbol per rank */ ZSTD_memset(rankLast, 0xF0, sizeof(rankLast)); - { U32 currentNbBits = maxNbBits; + { U32 currentNbBits = maxNbBits; int pos; - for (pos=n ; pos >= 0; pos--) { - if (huffNode[pos].nbBits >= currentNbBits) continue; - currentNbBits = huffNode[pos].nbBits; /* < maxNbBits */ + for (pos=n ; pos >= 0; pos--) { + if (huffNode[pos].nbBits >= currentNbBits) continue; + currentNbBits = huffNode[pos].nbBits; /* < maxNbBits */ rankLast[maxNbBits-currentNbBits] = (U32)pos; - } } - - while (totalCost > 0) { + } } + + while (totalCost > 0) { /* Try to reduce the next power of 2 above totalCost because we * gain back half the rank. */ U32 nBitsToDecrease = BIT_highbit32((U32)totalCost) + 1; - for ( ; nBitsToDecrease > 1; nBitsToDecrease--) { + for ( ; nBitsToDecrease > 1; nBitsToDecrease--) { U32 const highPos = rankLast[nBitsToDecrease]; U32 const lowPos = rankLast[nBitsToDecrease-1]; - if (highPos == noSymbol) continue; + if (highPos == noSymbol) continue; /* Decrease highPos if no symbols of lowPos or if it is * not cheaper to remove 2 lowPos than highPos. */ - if (lowPos == noSymbol) break; - { U32 const highTotal = huffNode[highPos].count; - U32 const lowTotal = 2 * huffNode[lowPos].count; - if (highTotal <= lowTotal) break; - } } - /* only triggered when no more rank 1 symbol left => find closest one (note : there is necessarily at least one !) */ + if (lowPos == noSymbol) break; + { U32 const highTotal = huffNode[highPos].count; + U32 const lowTotal = 2 * huffNode[lowPos].count; + if (highTotal <= lowTotal) break; + } } + /* only triggered when no more rank 1 symbol left => find closest one (note : there is necessarily at least one !) */ assert(rankLast[nBitsToDecrease] != noSymbol || nBitsToDecrease == 1); /* HUF_MAX_TABLELOG test just to please gcc 5+; but it should not be necessary */ while ((nBitsToDecrease<=HUF_TABLELOG_MAX) && (rankLast[nBitsToDecrease] == noSymbol)) nBitsToDecrease++; assert(rankLast[nBitsToDecrease] != noSymbol); /* Increase the number of bits to gain back half the rank cost. */ - totalCost -= 1 << (nBitsToDecrease-1); + totalCost -= 1 << (nBitsToDecrease-1); huffNode[rankLast[nBitsToDecrease]].nbBits++; /* Fix up the new rank. * If the new rank was empty, this symbol is now its smallest. * Otherwise, this symbol will be the largest in the new rank so no adjustment. */ - if (rankLast[nBitsToDecrease-1] == noSymbol) + if (rankLast[nBitsToDecrease-1] == noSymbol) rankLast[nBitsToDecrease-1] = rankLast[nBitsToDecrease]; /* Fix up the old rank. * If the symbol was at position 0, meaning it was the highest weight symbol in the tree, @@ -390,48 +390,48 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits) * the smallest node in the rank. If the previous position belongs to a different rank, * then the rank is now empty. */ - if (rankLast[nBitsToDecrease] == 0) /* special case, reached largest symbol */ - rankLast[nBitsToDecrease] = noSymbol; - else { - rankLast[nBitsToDecrease]--; - if (huffNode[rankLast[nBitsToDecrease]].nbBits != maxNbBits-nBitsToDecrease) - rankLast[nBitsToDecrease] = noSymbol; /* this rank is now empty */ + if (rankLast[nBitsToDecrease] == 0) /* special case, reached largest symbol */ + rankLast[nBitsToDecrease] = noSymbol; + else { + rankLast[nBitsToDecrease]--; + if (huffNode[rankLast[nBitsToDecrease]].nbBits != maxNbBits-nBitsToDecrease) + rankLast[nBitsToDecrease] = noSymbol; /* this rank is now empty */ } } /* while (totalCost > 0) */ - + /* If we've removed too much weight, then we have to add it back. * To avoid overshooting again, we only adjust the smallest rank. * We take the largest nodes from the lowest rank 0 and move them * to rank 1. There's guaranteed to be enough rank 0 symbols because * TODO. */ - while (totalCost < 0) { /* Sometimes, cost correction overshoot */ + while (totalCost < 0) { /* Sometimes, cost correction overshoot */ /* special case : no rank 1 symbol (using maxNbBits-1); * let's create one from largest rank 0 (using maxNbBits). */ if (rankLast[1] == noSymbol) { - while (huffNode[n].nbBits == maxNbBits) n--; - huffNode[n+1].nbBits--; + while (huffNode[n].nbBits == maxNbBits) n--; + huffNode[n+1].nbBits--; assert(n >= 0); rankLast[1] = (U32)(n+1); - totalCost++; - continue; - } - huffNode[ rankLast[1] + 1 ].nbBits--; - rankLast[1]++; - totalCost ++; + totalCost++; + continue; + } + huffNode[ rankLast[1] + 1 ].nbBits--; + rankLast[1]++; + totalCost ++; } } /* repay normalized cost */ } /* there are several too large elements (at least >= 2) */ - - return maxNbBits; -} - -typedef struct { + + return maxNbBits; +} + +typedef struct { U16 base; U16 curr; -} rankPos; - +} rankPos; + typedef nodeElt huffNodeTable[HUF_CTABLE_WORKSPACE_SIZE_U32]; /* Number of buckets available for HUF_sort() */ @@ -550,7 +550,7 @@ static void HUF_simpleQuickSort(nodeElt arr[], int low, int high) { static void HUF_sort(nodeElt huffNode[], const unsigned count[], U32 const maxSymbolValue, rankPos rankPosition[]) { U32 n; U32 const maxSymbolValue1 = maxSymbolValue+1; - + /* Compute base and set curr to base. * For symbol s let lowerRank = HUF_getIndex(count[n]) and rank = lowerRank + 1. * See HUF_getIndex to see bucketing strategy. @@ -562,7 +562,7 @@ static void HUF_sort(nodeElt huffNode[], const unsigned count[], U32 const maxSy U32 lowerRank = HUF_getIndex(count[n]); assert(lowerRank < RANK_POSITION_TABLE_SIZE - 1); rankPosition[lowerRank].base++; - } + } assert(rankPosition[RANK_POSITION_TABLE_SIZE - 1].base == 0); /* Set up the rankPosition table */ @@ -573,13 +573,13 @@ static void HUF_sort(nodeElt huffNode[], const unsigned count[], U32 const maxSy /* Insert each symbol into their appropriate bucket, setting up rankPosition table. */ for (n = 0; n < maxSymbolValue1; ++n) { - U32 const c = count[n]; + U32 const c = count[n]; U32 const r = HUF_getIndex(c) + 1; U32 const pos = rankPosition[r].curr++; assert(pos < maxSymbolValue1); - huffNode[pos].count = c; - huffNode[pos].byte = (BYTE)n; - } + huffNode[pos].count = c; + huffNode[pos].byte = (BYTE)n; + } /* Sort each bucket. */ for (n = RANK_POSITION_DISTINCT_COUNT_CUTOFF; n < RANK_POSITION_TABLE_SIZE - 1; ++n) { @@ -592,13 +592,13 @@ static void HUF_sort(nodeElt huffNode[], const unsigned count[], U32 const maxSy } assert(HUF_isSorted(huffNode, maxSymbolValue1)); -} - +} + /** HUF_buildCTable_wksp() : * Same as HUF_buildCTable(), but using externally allocated scratch buffer. * `workSpace` must be aligned on 4-bytes boundaries, and be at least as large as sizeof(HUF_buildCTable_wksp_tables). */ -#define STARTNODE (HUF_SYMBOLVALUE_MAX+1) +#define STARTNODE (HUF_SYMBOLVALUE_MAX+1) /* HUF_buildTree(): * Takes the huffNode array sorted by HUF_sort() and builds an unlimited-depth Huffman tree. @@ -608,38 +608,38 @@ static void HUF_sort(nodeElt huffNode[], const unsigned count[], U32 const maxSy * @return The smallest node in the Huffman tree (by count). */ static int HUF_buildTree(nodeElt* huffNode, U32 maxSymbolValue) -{ +{ nodeElt* const huffNode0 = huffNode - 1; int nonNullRank; - int lowS, lowN; + int lowS, lowN; int nodeNb = STARTNODE; int n, nodeRoot; - /* init for parents */ + /* init for parents */ nonNullRank = (int)maxSymbolValue; - while(huffNode[nonNullRank].count == 0) nonNullRank--; - lowS = nonNullRank; nodeRoot = nodeNb + lowS - 1; lowN = nodeNb; - huffNode[nodeNb].count = huffNode[lowS].count + huffNode[lowS-1].count; + while(huffNode[nonNullRank].count == 0) nonNullRank--; + lowS = nonNullRank; nodeRoot = nodeNb + lowS - 1; lowN = nodeNb; + huffNode[nodeNb].count = huffNode[lowS].count + huffNode[lowS-1].count; huffNode[lowS].parent = huffNode[lowS-1].parent = (U16)nodeNb; - nodeNb++; lowS-=2; - for (n=nodeNb; n<=nodeRoot; n++) huffNode[n].count = (U32)(1U<<30); + nodeNb++; lowS-=2; + for (n=nodeNb; n<=nodeRoot; n++) huffNode[n].count = (U32)(1U<<30); huffNode0[0].count = (U32)(1U<<31); /* fake entry, strong barrier */ - - /* create parents */ - while (nodeNb <= nodeRoot) { + + /* create parents */ + while (nodeNb <= nodeRoot) { int const n1 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++; int const n2 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++; - huffNode[nodeNb].count = huffNode[n1].count + huffNode[n2].count; + huffNode[nodeNb].count = huffNode[n1].count + huffNode[n2].count; huffNode[n1].parent = huffNode[n2].parent = (U16)nodeNb; - nodeNb++; - } - - /* distribute weights (unlimited tree height) */ - huffNode[nodeRoot].nbBits = 0; - for (n=nodeRoot-1; n>=STARTNODE; n--) - huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1; - for (n=0; n<=nonNullRank; n++) - huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1; - + nodeNb++; + } + + /* distribute weights (unlimited tree height) */ + huffNode[nodeRoot].nbBits = 0; + for (n=nodeRoot-1; n>=STARTNODE; n--) + huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1; + for (n=0; n<=nonNullRank; n++) + huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1; + return nonNullRank; } @@ -698,17 +698,17 @@ size_t HUF_buildCTable_wksp (HUF_CElt* CTable, const unsigned* count, U32 maxSym /* build tree */ nonNullRank = HUF_buildTree(huffNode, maxSymbolValue); - /* enforce maxTableLog */ + /* enforce maxTableLog */ maxNbBits = HUF_setMaxHeight(huffNode, (U32)nonNullRank, maxNbBits); if (maxNbBits > HUF_TABLELOG_MAX) return ERROR(GENERIC); /* check fit into table */ - + HUF_buildCTableFromTree(CTable, huffNode, nonNullRank, maxSymbolValue, maxNbBits); - - return maxNbBits; -} - + + return maxNbBits; +} + size_t HUF_estimateCompressedSize(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue) -{ +{ HUF_CElt const* ct = CTable + 1; size_t nbBits = 0; int s; @@ -716,8 +716,8 @@ size_t HUF_estimateCompressedSize(const HUF_CElt* CTable, const unsigned* count, nbBits += HUF_getNbBits(ct[s]) * count[s]; } return nbBits >> 3; -} - +} + int HUF_validateCTable(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue) { HUF_CElt const* ct = CTable + 1; int bad = 0; @@ -728,8 +728,8 @@ int HUF_validateCTable(const HUF_CElt* CTable, const unsigned* count, unsigned m return !bad; } -size_t HUF_compressBound(size_t size) { return HUF_COMPRESSBOUND(size); } - +size_t HUF_compressBound(size_t size) { return HUF_COMPRESSBOUND(size); } + /** HUF_CStream_t: * Huffman uses its own BIT_CStream_t implementation. * There are three major differences from BIT_CStream_t: @@ -896,7 +896,7 @@ HUF_encodeSymbol(HUF_CStream_t* bitCPtr, U32 symbol, const HUF_CElt* CTable, int { HUF_addBits(bitCPtr, CTable[symbol], idx, fast); } - + FORCE_INLINE_TEMPLATE void HUF_compress1X_usingCTable_internal_body_loop(HUF_CStream_t* bitC, const BYTE* ip, size_t srcSize, @@ -925,7 +925,7 @@ HUF_compress1X_usingCTable_internal_body_loop(HUF_CStream_t* bitC, n -= kUnroll; } assert(n % (2 * kUnroll) == 0); - + for (; n>0; n-= 2 * kUnroll) { /* Encode kUnroll symbols into the bitstream @ index 0. */ int u; @@ -948,7 +948,7 @@ HUF_compress1X_usingCTable_internal_body_loop(HUF_CStream_t* bitC, HUF_flushBits(bitC, kFastFlush); } assert(n == 0); - + } /** @@ -966,20 +966,20 @@ FORCE_INLINE_TEMPLATE size_t HUF_compress1X_usingCTable_internal_body(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable) -{ +{ U32 const tableLog = (U32)CTable[0]; HUF_CElt const* ct = CTable + 1; - const BYTE* ip = (const BYTE*) src; - BYTE* const ostart = (BYTE*)dst; - BYTE* const oend = ostart + dstSize; - BYTE* op = ostart; + const BYTE* ip = (const BYTE*) src; + BYTE* const ostart = (BYTE*)dst; + BYTE* const oend = ostart + dstSize; + BYTE* op = ostart; HUF_CStream_t bitC; - - /* init */ - if (dstSize < 8) return 0; /* not enough space to compress */ + + /* init */ + if (dstSize < 8) return 0; /* not enough space to compress */ { size_t const initErr = HUF_initCStream(&bitC, op, (size_t)(oend-op)); if (HUF_isError(initErr)) return 0; } - + if (dstSize < HUF_tightCompressBound(srcSize, (size_t)tableLog) || tableLog > 11) HUF_compress1X_usingCTable_internal_body_loop(&bitC, ip, srcSize, ct, /* kUnroll */ MEM_32bits() ? 2 : 4, /* kFast */ 0, /* kLastFast */ 0); else { @@ -1021,19 +1021,19 @@ HUF_compress1X_usingCTable_internal_body(void* dst, size_t dstSize, break; } } - } + } assert(bitC.ptr <= bitC.endPtr); - + return HUF_closeCStream(&bitC); -} - +} + #if DYNAMIC_BMI2 - + static BMI2_TARGET_ATTRIBUTE size_t HUF_compress1X_usingCTable_internal_bmi2(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable) -{ +{ return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable); } @@ -1084,56 +1084,56 @@ HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable, int bmi2) { - size_t const segmentSize = (srcSize+3)/4; /* first 3 segments */ - const BYTE* ip = (const BYTE*) src; - const BYTE* const iend = ip + srcSize; - BYTE* const ostart = (BYTE*) dst; - BYTE* const oend = ostart + dstSize; - BYTE* op = ostart; - - if (dstSize < 6 + 1 + 1 + 1 + 8) return 0; /* minimum space to compress successfully */ - if (srcSize < 12) return 0; /* no saving possible : too small input */ - op += 6; /* jumpTable */ - + size_t const segmentSize = (srcSize+3)/4; /* first 3 segments */ + const BYTE* ip = (const BYTE*) src; + const BYTE* const iend = ip + srcSize; + BYTE* const ostart = (BYTE*) dst; + BYTE* const oend = ostart + dstSize; + BYTE* op = ostart; + + if (dstSize < 6 + 1 + 1 + 1 + 8) return 0; /* minimum space to compress successfully */ + if (srcSize < 12) return 0; /* no saving possible : too small input */ + op += 6; /* jumpTable */ + assert(op <= oend); { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) ); if (cSize == 0 || cSize > 65535) return 0; - MEM_writeLE16(ostart, (U16)cSize); - op += cSize; - } - - ip += segmentSize; + MEM_writeLE16(ostart, (U16)cSize); + op += cSize; + } + + ip += segmentSize; assert(op <= oend); { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) ); if (cSize == 0 || cSize > 65535) return 0; - MEM_writeLE16(ostart+2, (U16)cSize); - op += cSize; - } - - ip += segmentSize; + MEM_writeLE16(ostart+2, (U16)cSize); + op += cSize; + } + + ip += segmentSize; assert(op <= oend); { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) ); if (cSize == 0 || cSize > 65535) return 0; - MEM_writeLE16(ostart+4, (U16)cSize); - op += cSize; - } - - ip += segmentSize; + MEM_writeLE16(ostart+4, (U16)cSize); + op += cSize; + } + + ip += segmentSize; assert(op <= oend); assert(ip <= iend); { CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, (size_t)(iend-ip), CTable, bmi2) ); if (cSize == 0 || cSize > 65535) return 0; - op += cSize; - } - + op += cSize; + } + return (size_t)(op-ostart); -} - +} + size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable) { return HUF_compress4X_usingCTable_bmi2(dst, dstSize, src, srcSize, CTable, /* bmi2 */ 0); } - + size_t HUF_compress4X_usingCTable_bmi2(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable, int bmi2) { return HUF_compress4X_usingCTable_internal(dst, dstSize, src, srcSize, CTable, bmi2); @@ -1182,24 +1182,24 @@ HUF_compress_internal (void* dst, size_t dstSize, void* workSpace, size_t wkspSize, HUF_CElt* oldHufTable, HUF_repeat* repeat, int preferRepeat, const int bmi2, unsigned suspectUncompressible) -{ +{ HUF_compress_tables_t* const table = (HUF_compress_tables_t*)HUF_alignUpWorkspace(workSpace, &wkspSize, ZSTD_ALIGNOF(size_t)); - BYTE* const ostart = (BYTE*)dst; - BYTE* const oend = ostart + dstSize; - BYTE* op = ostart; - + BYTE* const ostart = (BYTE*)dst; + BYTE* const oend = ostart + dstSize; + BYTE* op = ostart; + HUF_STATIC_ASSERT(sizeof(*table) + HUF_WORKSPACE_MAX_ALIGNMENT <= HUF_WORKSPACE_SIZE); - /* checks & inits */ + /* checks & inits */ if (wkspSize < sizeof(*table)) return ERROR(workSpace_tooSmall); if (!srcSize) return 0; /* Uncompressed */ if (!dstSize) return 0; /* cannot fit anything within dst budget */ - if (srcSize > HUF_BLOCKSIZE_MAX) return ERROR(srcSize_wrong); /* current block size limit */ - if (huffLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge); + if (srcSize > HUF_BLOCKSIZE_MAX) return ERROR(srcSize_wrong); /* current block size limit */ + if (huffLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge); if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge); - if (!maxSymbolValue) maxSymbolValue = HUF_SYMBOLVALUE_MAX; - if (!huffLog) huffLog = HUF_TABLELOG_DEFAULT; - + if (!maxSymbolValue) maxSymbolValue = HUF_SYMBOLVALUE_MAX; + if (!huffLog) huffLog = HUF_TABLELOG_DEFAULT; + /* Heuristic : If old table is valid, use it for small inputs */ if (preferRepeat && repeat && *repeat == HUF_repeat_valid) { return HUF_compressCTable_internal(ostart, op, oend, @@ -1222,12 +1222,12 @@ HUF_compress_internal (void* dst, size_t dstSize, if (largestTotal <= ((2 * SUSPECT_INCOMPRESSIBLE_SAMPLE_SIZE) >> 7)+4) return 0; /* heuristic : probably not compressible enough */ } - /* Scan input and build symbol stats */ + /* Scan input and build symbol stats */ { CHECK_V_F(largest, HIST_count_wksp (table->count, &maxSymbolValue, (const BYTE*)src, srcSize, table->wksps.hist_wksp, sizeof(table->wksps.hist_wksp)) ); if (largest == srcSize) { *ostart = ((const BYTE*)src)[0]; return 1; } /* single symbol, rle */ if (largest <= (srcSize >> 7)+4) return 0; /* heuristic : probably not compressible enough */ - } - + } + /* Check validity of previous table */ if ( repeat && *repeat == HUF_repeat_check @@ -1241,22 +1241,22 @@ HUF_compress_internal (void* dst, size_t dstSize, nbStreams, oldHufTable, bmi2); } - /* Build Huffman Tree */ - huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue); + /* Build Huffman Tree */ + huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue); { size_t const maxBits = HUF_buildCTable_wksp(table->CTable, table->count, maxSymbolValue, huffLog, &table->wksps.buildCTable_wksp, sizeof(table->wksps.buildCTable_wksp)); CHECK_F(maxBits); - huffLog = (U32)maxBits; - } + huffLog = (U32)maxBits; + } /* Zero unused symbols in CTable, so we can check it for validity */ { size_t const ctableSize = HUF_CTABLE_SIZE_ST(maxSymbolValue); size_t const unusedSize = sizeof(table->CTable) - ctableSize * sizeof(HUF_CElt); ZSTD_memset(table->CTable + ctableSize, 0, unusedSize); } - - /* Write table description header */ + + /* Write table description header */ { CHECK_V_F(hSize, HUF_writeCTable_wksp(op, dstSize, table->CTable, maxSymbolValue, huffLog, &table->wksps.writeCTable_wksp, sizeof(table->wksps.writeCTable_wksp)) ); /* Check if using previous huffman table is beneficial */ @@ -1271,17 +1271,17 @@ HUF_compress_internal (void* dst, size_t dstSize, /* Use the new huffman table */ if (hSize + 12ul >= srcSize) { return 0; } - op += hSize; + op += hSize; if (repeat) { *repeat = HUF_repeat_none; } if (oldHufTable) ZSTD_memcpy(oldHufTable, table->CTable, sizeof(table->CTable)); /* Save new table */ - } + } return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, nbStreams, table->CTable, bmi2); -} - - +} + + size_t HUF_compress1X_wksp (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned huffLog, @@ -1355,16 +1355,16 @@ size_t HUF_compress1X (void* dst, size_t dstSize, return HUF_compress1X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace)); } -size_t HUF_compress2 (void* dst, size_t dstSize, - const void* src, size_t srcSize, - unsigned maxSymbolValue, unsigned huffLog) -{ +size_t HUF_compress2 (void* dst, size_t dstSize, + const void* src, size_t srcSize, + unsigned maxSymbolValue, unsigned huffLog) +{ U64 workSpace[HUF_WORKSPACE_SIZE_U64]; return HUF_compress4X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace)); -} - -size_t HUF_compress (void* dst, size_t maxDstSize, const void* src, size_t srcSize) -{ +} + +size_t HUF_compress (void* dst, size_t maxDstSize, const void* src, size_t srcSize) +{ return HUF_compress2(dst, maxDstSize, src, srcSize, 255, HUF_TABLELOG_DEFAULT); -} +} #endif |