diff options
author | danlark <danlark@yandex-team.ru> | 2022-02-10 16:46:08 +0300 |
---|---|---|
committer | Daniil Cherednik <dcherednik@yandex-team.ru> | 2022-02-10 16:46:08 +0300 |
commit | 3426a9bc7f169ae9da54cef557ad2a33f6e8eee0 (patch) | |
tree | 26154e1e9990f1bb4525d3e3fb5b6dac2c2c1da2 /contrib/libs/lz4/lz4hc.c | |
parent | cb68f224c46a8ee52ac3fdd2a32534b8bb8dc134 (diff) | |
download | ydb-3426a9bc7f169ae9da54cef557ad2a33f6e8eee0.tar.gz |
Restoring authorship annotation for <danlark@yandex-team.ru>. Commit 1 of 2.
Diffstat (limited to 'contrib/libs/lz4/lz4hc.c')
-rw-r--r-- | contrib/libs/lz4/lz4hc.c | 1860 |
1 files changed, 930 insertions, 930 deletions
diff --git a/contrib/libs/lz4/lz4hc.c b/contrib/libs/lz4/lz4hc.c index a556d47920..6e5ea27f1d 100644 --- a/contrib/libs/lz4/lz4hc.c +++ b/contrib/libs/lz4/lz4hc.c @@ -1,6 +1,6 @@ /* LZ4 HC - High Compression Mode of LZ4 - Copyright (C) 2011-2017, Yann Collet. + Copyright (C) 2011-2017, Yann Collet. BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) @@ -38,18 +38,18 @@ * Tuning Parameter ***************************************/ -/*! HEAPMODE : - * Select how default compression function will allocate workplace memory, - * in stack (0:fastest), or in heap (1:requires malloc()). - * Since workplace is rather large, heap mode is recommended. +/*! HEAPMODE : + * Select how default compression function will allocate workplace memory, + * in stack (0:fastest), or in heap (1:requires malloc()). + * Since workplace is rather large, heap mode is recommended. */ #ifndef LZ4HC_HEAPMODE # define LZ4HC_HEAPMODE 1 #endif -/*=== Dependency ===*/ -#define LZ4_HC_STATIC_LINKING_ONLY +/*=== Dependency ===*/ +#define LZ4_HC_STATIC_LINKING_ONLY #include "lz4hc.h" @@ -62,28 +62,28 @@ #endif #define LZ4_COMMONDEFS_ONLY -#ifndef LZ4_SRC_INCLUDED -#include "lz4.c" /* LZ4_count, constants, mem */ -#endif +#ifndef LZ4_SRC_INCLUDED +#include "lz4.c" /* LZ4_count, constants, mem */ +#endif /*=== Enums ===*/ typedef enum { noDictCtx, usingDictCtxHc } dictCtx_directive; -/*=== Constants ===*/ +/*=== Constants ===*/ #define OPTIMAL_ML (int)((ML_MASK-1)+MINMATCH) -#define LZ4_OPT_NUM (1<<12) +#define LZ4_OPT_NUM (1<<12) -/*=== Macros ===*/ -#define MIN(a,b) ( (a) < (b) ? (a) : (b) ) -#define MAX(a,b) ( (a) > (b) ? (a) : (b) ) -#define HASH_FUNCTION(i) (((i) * 2654435761U) >> ((MINMATCH*8)-LZ4HC_HASH_LOG)) -#define DELTANEXTMAXD(p) chainTable[(p) & LZ4HC_MAXD_MASK] /* flexible, LZ4HC_MAXD dependent */ -#define DELTANEXTU16(table, pos) table[(U16)(pos)] /* faster */ -/* Make fields passed to, and updated by LZ4HC_encodeSequence explicit */ -#define UPDATABLE(ip, op, anchor) &ip, &op, &anchor +/*=== Macros ===*/ +#define MIN(a,b) ( (a) < (b) ? (a) : (b) ) +#define MAX(a,b) ( (a) > (b) ? (a) : (b) ) +#define HASH_FUNCTION(i) (((i) * 2654435761U) >> ((MINMATCH*8)-LZ4HC_HASH_LOG)) +#define DELTANEXTMAXD(p) chainTable[(p) & LZ4HC_MAXD_MASK] /* flexible, LZ4HC_MAXD dependent */ +#define DELTANEXTU16(table, pos) table[(U16)(pos)] /* faster */ +/* Make fields passed to, and updated by LZ4HC_encodeSequence explicit */ +#define UPDATABLE(ip, op, anchor) &ip, &op, &anchor static U32 LZ4HC_hashPtr(const void* ptr) { return HASH_FUNCTION(LZ4_read32(ptr)); } @@ -91,31 +91,31 @@ static U32 LZ4HC_hashPtr(const void* ptr) { return HASH_FUNCTION(LZ4_read32(ptr) /************************************** * HC Compression **************************************/ -static void LZ4HC_clearTables (LZ4HC_CCtx_internal* hc4) +static void LZ4HC_clearTables (LZ4HC_CCtx_internal* hc4) { MEM_INIT(hc4->hashTable, 0, sizeof(hc4->hashTable)); MEM_INIT(hc4->chainTable, 0xFF, sizeof(hc4->chainTable)); -} - -static void LZ4HC_init_internal (LZ4HC_CCtx_internal* hc4, const BYTE* start) -{ - uptrval startingOffset = (uptrval)(hc4->end - hc4->base); - if (startingOffset > 1 GB) { - LZ4HC_clearTables(hc4); - startingOffset = 0; - } - startingOffset += 64 KB; - hc4->nextToUpdate = (U32) startingOffset; - hc4->base = start - startingOffset; +} + +static void LZ4HC_init_internal (LZ4HC_CCtx_internal* hc4, const BYTE* start) +{ + uptrval startingOffset = (uptrval)(hc4->end - hc4->base); + if (startingOffset > 1 GB) { + LZ4HC_clearTables(hc4); + startingOffset = 0; + } + startingOffset += 64 KB; + hc4->nextToUpdate = (U32) startingOffset; + hc4->base = start - startingOffset; hc4->end = start; - hc4->dictBase = start - startingOffset; - hc4->dictLimit = (U32) startingOffset; - hc4->lowLimit = (U32) startingOffset; + hc4->dictBase = start - startingOffset; + hc4->dictLimit = (U32) startingOffset; + hc4->lowLimit = (U32) startingOffset; } /* Update chains up to ip (excluded) */ -LZ4_FORCE_INLINE void LZ4HC_Insert (LZ4HC_CCtx_internal* hc4, const BYTE* ip) +LZ4_FORCE_INLINE void LZ4HC_Insert (LZ4HC_CCtx_internal* hc4, const BYTE* ip) { U16* const chainTable = hc4->chainTable; U32* const hashTable = hc4->hashTable; @@ -126,8 +126,8 @@ LZ4_FORCE_INLINE void LZ4HC_Insert (LZ4HC_CCtx_internal* hc4, const BYTE* ip) while (idx < target) { U32 const h = LZ4HC_hashPtr(base+idx); size_t delta = idx - hashTable[h]; - if (delta>LZ4_DISTANCE_MAX) delta = LZ4_DISTANCE_MAX; - DELTANEXTU16(chainTable, idx) = (U16)delta; + if (delta>LZ4_DISTANCE_MAX) delta = LZ4_DISTANCE_MAX; + DELTANEXTU16(chainTable, idx) = (U16)delta; hashTable[h] = idx; idx++; } @@ -135,22 +135,22 @@ LZ4_FORCE_INLINE void LZ4HC_Insert (LZ4HC_CCtx_internal* hc4, const BYTE* ip) hc4->nextToUpdate = target; } -/** LZ4HC_countBack() : - * @return : negative value, nb of common bytes before ip/match */ -LZ4_FORCE_INLINE -int LZ4HC_countBack(const BYTE* const ip, const BYTE* const match, - const BYTE* const iMin, const BYTE* const mMin) -{ - int back = 0; - int const min = (int)MAX(iMin - ip, mMin - match); - assert(min <= 0); - assert(ip >= iMin); assert((size_t)(ip-iMin) < (1U<<31)); - assert(match >= mMin); assert((size_t)(match - mMin) < (1U<<31)); - while ( (back > min) - && (ip[back-1] == match[back-1]) ) - back--; - return back; -} +/** LZ4HC_countBack() : + * @return : negative value, nb of common bytes before ip/match */ +LZ4_FORCE_INLINE +int LZ4HC_countBack(const BYTE* const ip, const BYTE* const match, + const BYTE* const iMin, const BYTE* const mMin) +{ + int back = 0; + int const min = (int)MAX(iMin - ip, mMin - match); + assert(min <= 0); + assert(ip >= iMin); assert((size_t)(ip-iMin) < (1U<<31)); + assert(match >= mMin); assert((size_t)(match - mMin) < (1U<<31)); + while ( (back > min) + && (ip[back-1] == match[back-1]) ) + back--; + return back; +} #if defined(_MSC_VER) # define LZ4HC_rotl32(x,r) _rotl(x,r) @@ -166,59 +166,59 @@ static U32 LZ4HC_rotatePattern(size_t const rotate, U32 const pattern) return LZ4HC_rotl32(pattern, (int)bitsToRotate); } -/* LZ4HC_countPattern() : - * pattern32 must be a sample of repetitive pattern of length 1, 2 or 4 (but not 3!) */ -static unsigned -LZ4HC_countPattern(const BYTE* ip, const BYTE* const iEnd, U32 const pattern32) +/* LZ4HC_countPattern() : + * pattern32 must be a sample of repetitive pattern of length 1, 2 or 4 (but not 3!) */ +static unsigned +LZ4HC_countPattern(const BYTE* ip, const BYTE* const iEnd, U32 const pattern32) { - const BYTE* const iStart = ip; + const BYTE* const iStart = ip; reg_t const pattern = (sizeof(pattern)==8) ? (reg_t)pattern32 + (((reg_t)pattern32) << (sizeof(pattern)*4)) : pattern32; - while (likely(ip < iEnd-(sizeof(pattern)-1))) { - reg_t const diff = LZ4_read_ARCH(ip) ^ pattern; - if (!diff) { ip+=sizeof(pattern); continue; } - ip += LZ4_NbCommonBytes(diff); - return (unsigned)(ip - iStart); - } - - if (LZ4_isLittleEndian()) { - reg_t patternByte = pattern; - while ((ip<iEnd) && (*ip == (BYTE)patternByte)) { - ip++; patternByte >>= 8; - } - } else { /* big endian */ - U32 bitOffset = (sizeof(pattern)*8) - 8; - while (ip < iEnd) { - BYTE const byte = (BYTE)(pattern >> bitOffset); - if (*ip != byte) break; - ip ++; bitOffset -= 8; + while (likely(ip < iEnd-(sizeof(pattern)-1))) { + reg_t const diff = LZ4_read_ARCH(ip) ^ pattern; + if (!diff) { ip+=sizeof(pattern); continue; } + ip += LZ4_NbCommonBytes(diff); + return (unsigned)(ip - iStart); + } + + if (LZ4_isLittleEndian()) { + reg_t patternByte = pattern; + while ((ip<iEnd) && (*ip == (BYTE)patternByte)) { + ip++; patternByte >>= 8; } + } else { /* big endian */ + U32 bitOffset = (sizeof(pattern)*8) - 8; + while (ip < iEnd) { + BYTE const byte = (BYTE)(pattern >> bitOffset); + if (*ip != byte) break; + ip ++; bitOffset -= 8; + } } - return (unsigned)(ip - iStart); -} - -/* LZ4HC_reverseCountPattern() : - * pattern must be a sample of repetitive pattern of length 1, 2 or 4 (but not 3!) - * read using natural platform endianess */ -static unsigned -LZ4HC_reverseCountPattern(const BYTE* ip, const BYTE* const iLow, U32 pattern) -{ - const BYTE* const iStart = ip; - - while (likely(ip >= iLow+4)) { - if (LZ4_read32(ip-4) != pattern) break; - ip -= 4; - } - { const BYTE* bytePtr = (const BYTE*)(&pattern) + 3; /* works for any endianess */ - while (likely(ip>iLow)) { - if (ip[-1] != *bytePtr) break; - ip--; bytePtr--; - } } - return (unsigned)(iStart - ip); + return (unsigned)(ip - iStart); } +/* LZ4HC_reverseCountPattern() : + * pattern must be a sample of repetitive pattern of length 1, 2 or 4 (but not 3!) + * read using natural platform endianess */ +static unsigned +LZ4HC_reverseCountPattern(const BYTE* ip, const BYTE* const iLow, U32 pattern) +{ + const BYTE* const iStart = ip; + + while (likely(ip >= iLow+4)) { + if (LZ4_read32(ip-4) != pattern) break; + ip -= 4; + } + { const BYTE* bytePtr = (const BYTE*)(&pattern) + 3; /* works for any endianess */ + while (likely(ip>iLow)) { + if (ip[-1] != *bytePtr) break; + ip--; bytePtr--; + } } + return (unsigned)(iStart - ip); +} + /* LZ4HC_protectDictEnd() : * Checks if the match is in the last 3 bytes of the dictionary, so reading the * 4 byte MINMATCH would overflow. @@ -229,11 +229,11 @@ static int LZ4HC_protectDictEnd(U32 const dictLimit, U32 const matchIndex) return ((U32)((dictLimit - 1) - matchIndex) >= 3); } -typedef enum { rep_untested, rep_not, rep_confirmed } repeat_state_e; -typedef enum { favorCompressionRatio=0, favorDecompressionSpeed } HCfavor_e; - -LZ4_FORCE_INLINE int -LZ4HC_InsertAndGetWiderMatch ( +typedef enum { rep_untested, rep_not, rep_confirmed } repeat_state_e; +typedef enum { favorCompressionRatio=0, favorDecompressionSpeed } HCfavor_e; + +LZ4_FORCE_INLINE int +LZ4HC_InsertAndGetWiderMatch ( LZ4HC_CCtx_internal* hc4, const BYTE* const ip, const BYTE* const iLowLimit, @@ -241,116 +241,116 @@ LZ4HC_InsertAndGetWiderMatch ( int longest, const BYTE** matchpos, const BYTE** startpos, - const int maxNbAttempts, - const int patternAnalysis, - const int chainSwap, - const dictCtx_directive dict, - const HCfavor_e favorDecSpeed) + const int maxNbAttempts, + const int patternAnalysis, + const int chainSwap, + const dictCtx_directive dict, + const HCfavor_e favorDecSpeed) { U16* const chainTable = hc4->chainTable; U32* const HashTable = hc4->hashTable; - const LZ4HC_CCtx_internal * const dictCtx = hc4->dictCtx; + const LZ4HC_CCtx_internal * const dictCtx = hc4->dictCtx; const BYTE* const base = hc4->base; const U32 dictLimit = hc4->dictLimit; const BYTE* const lowPrefixPtr = base + dictLimit; - const U32 ipIndex = (U32)(ip - base); + const U32 ipIndex = (U32)(ip - base); const U32 lowestMatchIndex = (hc4->lowLimit + (LZ4_DISTANCE_MAX + 1) > ipIndex) ? hc4->lowLimit : ipIndex - LZ4_DISTANCE_MAX; const BYTE* const dictBase = hc4->dictBase; - int const lookBackLength = (int)(ip-iLowLimit); + int const lookBackLength = (int)(ip-iLowLimit); int nbAttempts = maxNbAttempts; - U32 matchChainPos = 0; - U32 const pattern = LZ4_read32(ip); - U32 matchIndex; - repeat_state_e repeat = rep_untested; - size_t srcPatternLength = 0; + U32 matchChainPos = 0; + U32 const pattern = LZ4_read32(ip); + U32 matchIndex; + repeat_state_e repeat = rep_untested; + size_t srcPatternLength = 0; - DEBUGLOG(7, "LZ4HC_InsertAndGetWiderMatch"); + DEBUGLOG(7, "LZ4HC_InsertAndGetWiderMatch"); /* First Match */ LZ4HC_Insert(hc4, ip); matchIndex = HashTable[LZ4HC_hashPtr(ip)]; - DEBUGLOG(7, "First match at index %u / %u (lowestMatchIndex)", - matchIndex, lowestMatchIndex); + DEBUGLOG(7, "First match at index %u / %u (lowestMatchIndex)", + matchIndex, lowestMatchIndex); while ((matchIndex>=lowestMatchIndex) && (nbAttempts>0)) { - int matchLength=0; + int matchLength=0; nbAttempts--; - assert(matchIndex < ipIndex); - if (favorDecSpeed && (ipIndex - matchIndex < 8)) { - /* do nothing */ - } else if (matchIndex >= dictLimit) { /* within current Prefix */ - const BYTE* const matchPtr = base + matchIndex; - assert(matchPtr >= lowPrefixPtr); - assert(matchPtr < ip); - assert(longest >= 1); - if (LZ4_read16(iLowLimit + longest - 1) == LZ4_read16(matchPtr - lookBackLength + longest - 1)) { - if (LZ4_read32(matchPtr) == pattern) { - int const back = lookBackLength ? LZ4HC_countBack(ip, matchPtr, iLowLimit, lowPrefixPtr) : 0; - matchLength = MINMATCH + (int)LZ4_count(ip+MINMATCH, matchPtr+MINMATCH, iHighLimit); - matchLength -= back; - if (matchLength > longest) { - longest = matchLength; - *matchpos = matchPtr + back; - *startpos = ip + back; - } } } - } else { /* lowestMatchIndex <= matchIndex < dictLimit */ - const BYTE* const matchPtr = dictBase + matchIndex; - if (LZ4_read32(matchPtr) == pattern) { - const BYTE* const dictStart = dictBase + hc4->lowLimit; - int back = 0; - const BYTE* vLimit = ip + (dictLimit - matchIndex); - if (vLimit > iHighLimit) vLimit = iHighLimit; - matchLength = (int)LZ4_count(ip+MINMATCH, matchPtr+MINMATCH, vLimit) + MINMATCH; - if ((ip+matchLength == vLimit) && (vLimit < iHighLimit)) - matchLength += LZ4_count(ip+matchLength, lowPrefixPtr, iHighLimit); - back = lookBackLength ? LZ4HC_countBack(ip, matchPtr, iLowLimit, dictStart) : 0; - matchLength -= back; - if (matchLength > longest) { - longest = matchLength; - *matchpos = base + matchIndex + back; /* virtual pos, relative to ip, to retrieve offset */ - *startpos = ip + back; - } } } - - if (chainSwap && matchLength==longest) { /* better match => select a better chain */ - assert(lookBackLength==0); /* search forward only */ - if (matchIndex + (U32)longest <= ipIndex) { + assert(matchIndex < ipIndex); + if (favorDecSpeed && (ipIndex - matchIndex < 8)) { + /* do nothing */ + } else if (matchIndex >= dictLimit) { /* within current Prefix */ + const BYTE* const matchPtr = base + matchIndex; + assert(matchPtr >= lowPrefixPtr); + assert(matchPtr < ip); + assert(longest >= 1); + if (LZ4_read16(iLowLimit + longest - 1) == LZ4_read16(matchPtr - lookBackLength + longest - 1)) { + if (LZ4_read32(matchPtr) == pattern) { + int const back = lookBackLength ? LZ4HC_countBack(ip, matchPtr, iLowLimit, lowPrefixPtr) : 0; + matchLength = MINMATCH + (int)LZ4_count(ip+MINMATCH, matchPtr+MINMATCH, iHighLimit); + matchLength -= back; + if (matchLength > longest) { + longest = matchLength; + *matchpos = matchPtr + back; + *startpos = ip + back; + } } } + } else { /* lowestMatchIndex <= matchIndex < dictLimit */ + const BYTE* const matchPtr = dictBase + matchIndex; + if (LZ4_read32(matchPtr) == pattern) { + const BYTE* const dictStart = dictBase + hc4->lowLimit; + int back = 0; + const BYTE* vLimit = ip + (dictLimit - matchIndex); + if (vLimit > iHighLimit) vLimit = iHighLimit; + matchLength = (int)LZ4_count(ip+MINMATCH, matchPtr+MINMATCH, vLimit) + MINMATCH; + if ((ip+matchLength == vLimit) && (vLimit < iHighLimit)) + matchLength += LZ4_count(ip+matchLength, lowPrefixPtr, iHighLimit); + back = lookBackLength ? LZ4HC_countBack(ip, matchPtr, iLowLimit, dictStart) : 0; + matchLength -= back; + if (matchLength > longest) { + longest = matchLength; + *matchpos = base + matchIndex + back; /* virtual pos, relative to ip, to retrieve offset */ + *startpos = ip + back; + } } } + + if (chainSwap && matchLength==longest) { /* better match => select a better chain */ + assert(lookBackLength==0); /* search forward only */ + if (matchIndex + (U32)longest <= ipIndex) { int const kTrigger = 4; - U32 distanceToNextMatch = 1; + U32 distanceToNextMatch = 1; int const end = longest - MINMATCH + 1; int step = 1; int accel = 1 << kTrigger; - int pos; + int pos; for (pos = 0; pos < end; pos += step) { - U32 const candidateDist = DELTANEXTU16(chainTable, matchIndex + (U32)pos); + U32 const candidateDist = DELTANEXTU16(chainTable, matchIndex + (U32)pos); step = (accel++ >> kTrigger); - if (candidateDist > distanceToNextMatch) { - distanceToNextMatch = candidateDist; - matchChainPos = (U32)pos; + if (candidateDist > distanceToNextMatch) { + distanceToNextMatch = candidateDist; + matchChainPos = (U32)pos; accel = 1 << kTrigger; } } - if (distanceToNextMatch > 1) { - if (distanceToNextMatch > matchIndex) break; /* avoid overflow */ - matchIndex -= distanceToNextMatch; - continue; - } } } - - { U32 const distNextMatch = DELTANEXTU16(chainTable, matchIndex); - if (patternAnalysis && distNextMatch==1 && matchChainPos==0) { - U32 const matchCandidateIdx = matchIndex-1; - /* may be a repeated pattern */ - if (repeat == rep_untested) { - if ( ((pattern & 0xFFFF) == (pattern >> 16)) - & ((pattern & 0xFF) == (pattern >> 24)) ) { - repeat = rep_confirmed; - srcPatternLength = LZ4HC_countPattern(ip+sizeof(pattern), iHighLimit, pattern) + sizeof(pattern); - } else { - repeat = rep_not; - } } + if (distanceToNextMatch > 1) { + if (distanceToNextMatch > matchIndex) break; /* avoid overflow */ + matchIndex -= distanceToNextMatch; + continue; + } } } + + { U32 const distNextMatch = DELTANEXTU16(chainTable, matchIndex); + if (patternAnalysis && distNextMatch==1 && matchChainPos==0) { + U32 const matchCandidateIdx = matchIndex-1; + /* may be a repeated pattern */ + if (repeat == rep_untested) { + if ( ((pattern & 0xFFFF) == (pattern >> 16)) + & ((pattern & 0xFF) == (pattern >> 24)) ) { + repeat = rep_confirmed; + srcPatternLength = LZ4HC_countPattern(ip+sizeof(pattern), iHighLimit, pattern) + sizeof(pattern); + } else { + repeat = rep_not; + } } if ( (repeat == rep_confirmed) && (matchCandidateIdx >= lowestMatchIndex) && LZ4HC_protectDictEnd(dictLimit, matchCandidateIdx) ) { const int extDict = matchCandidateIdx < dictLimit; const BYTE* const matchPtr = (extDict ? dictBase : base) + matchCandidateIdx; - if (LZ4_read32(matchPtr) == pattern) { /* good candidate */ + if (LZ4_read32(matchPtr) == pattern) { /* good candidate */ const BYTE* const dictStart = dictBase + hc4->lowLimit; const BYTE* const iLimit = extDict ? dictBase + dictLimit : iHighLimit; size_t forwardPatternLength = LZ4HC_countPattern(matchPtr+sizeof(pattern), iLimit, pattern) + sizeof(pattern); @@ -379,7 +379,7 @@ LZ4HC_InsertAndGetWiderMatch ( /* Can only happen if started in the prefix */ assert(newMatchIndex >= dictLimit - 3 && newMatchIndex < dictLimit && !extDict); matchIndex = dictLimit; - } + } } else { U32 const newMatchIndex = matchCandidateIdx - (U32)backLength; /* farthest position in current segment, will find a match of length currentSegmentLength + maybe some back */ if (!LZ4HC_protectDictEnd(dictLimit, newMatchIndex)) { @@ -401,96 +401,96 @@ LZ4HC_InsertAndGetWiderMatch ( if (distToNextPattern > matchIndex) break; /* avoid overflow */ matchIndex -= distToNextPattern; } } } } } - continue; - } } - } } /* PA optimization */ - - /* follow current chain */ - matchIndex -= DELTANEXTU16(chainTable, matchIndex + matchChainPos); - - } /* while ((matchIndex>=lowestMatchIndex) && (nbAttempts)) */ - - if ( dict == usingDictCtxHc + continue; + } } + } } /* PA optimization */ + + /* follow current chain */ + matchIndex -= DELTANEXTU16(chainTable, matchIndex + matchChainPos); + + } /* while ((matchIndex>=lowestMatchIndex) && (nbAttempts)) */ + + if ( dict == usingDictCtxHc && nbAttempts > 0 - && ipIndex - lowestMatchIndex < LZ4_DISTANCE_MAX) { - size_t const dictEndOffset = (size_t)(dictCtx->end - dictCtx->base); - U32 dictMatchIndex = dictCtx->hashTable[LZ4HC_hashPtr(ip)]; - assert(dictEndOffset <= 1 GB); - matchIndex = dictMatchIndex + lowestMatchIndex - (U32)dictEndOffset; - while (ipIndex - matchIndex <= LZ4_DISTANCE_MAX && nbAttempts--) { - const BYTE* const matchPtr = dictCtx->base + dictMatchIndex; - - if (LZ4_read32(matchPtr) == pattern) { - int mlt; - int back = 0; - const BYTE* vLimit = ip + (dictEndOffset - dictMatchIndex); + && ipIndex - lowestMatchIndex < LZ4_DISTANCE_MAX) { + size_t const dictEndOffset = (size_t)(dictCtx->end - dictCtx->base); + U32 dictMatchIndex = dictCtx->hashTable[LZ4HC_hashPtr(ip)]; + assert(dictEndOffset <= 1 GB); + matchIndex = dictMatchIndex + lowestMatchIndex - (U32)dictEndOffset; + while (ipIndex - matchIndex <= LZ4_DISTANCE_MAX && nbAttempts--) { + const BYTE* const matchPtr = dictCtx->base + dictMatchIndex; + + if (LZ4_read32(matchPtr) == pattern) { + int mlt; + int back = 0; + const BYTE* vLimit = ip + (dictEndOffset - dictMatchIndex); if (vLimit > iHighLimit) vLimit = iHighLimit; - mlt = (int)LZ4_count(ip+MINMATCH, matchPtr+MINMATCH, vLimit) + MINMATCH; - back = lookBackLength ? LZ4HC_countBack(ip, matchPtr, iLowLimit, dictCtx->base + dictCtx->dictLimit) : 0; + mlt = (int)LZ4_count(ip+MINMATCH, matchPtr+MINMATCH, vLimit) + MINMATCH; + back = lookBackLength ? LZ4HC_countBack(ip, matchPtr, iLowLimit, dictCtx->base + dictCtx->dictLimit) : 0; mlt -= back; - if (mlt > longest) { - longest = mlt; - *matchpos = base + matchIndex + back; - *startpos = ip + back; - } } - - { U32 const nextOffset = DELTANEXTU16(dictCtx->chainTable, dictMatchIndex); - dictMatchIndex -= nextOffset; - matchIndex -= nextOffset; - } } } - + if (mlt > longest) { + longest = mlt; + *matchpos = base + matchIndex + back; + *startpos = ip + back; + } } + + { U32 const nextOffset = DELTANEXTU16(dictCtx->chainTable, dictMatchIndex); + dictMatchIndex -= nextOffset; + matchIndex -= nextOffset; + } } } + return longest; } -LZ4_FORCE_INLINE -int LZ4HC_InsertAndFindBestMatch(LZ4HC_CCtx_internal* const hc4, /* Index table will be updated */ - const BYTE* const ip, const BYTE* const iLimit, - const BYTE** matchpos, - const int maxNbAttempts, - const int patternAnalysis, - const dictCtx_directive dict) -{ - const BYTE* uselessPtr = ip; - /* note : LZ4HC_InsertAndGetWiderMatch() is able to modify the starting position of a match (*startpos), - * but this won't be the case here, as we define iLowLimit==ip, - * so LZ4HC_InsertAndGetWiderMatch() won't be allowed to search past ip */ - return LZ4HC_InsertAndGetWiderMatch(hc4, ip, ip, iLimit, MINMATCH-1, matchpos, &uselessPtr, maxNbAttempts, patternAnalysis, 0 /*chainSwap*/, dict, favorCompressionRatio); -} - -/* LZ4HC_encodeSequence() : - * @return : 0 if ok, - * 1 if buffer issue detected */ -LZ4_FORCE_INLINE int LZ4HC_encodeSequence ( +LZ4_FORCE_INLINE +int LZ4HC_InsertAndFindBestMatch(LZ4HC_CCtx_internal* const hc4, /* Index table will be updated */ + const BYTE* const ip, const BYTE* const iLimit, + const BYTE** matchpos, + const int maxNbAttempts, + const int patternAnalysis, + const dictCtx_directive dict) +{ + const BYTE* uselessPtr = ip; + /* note : LZ4HC_InsertAndGetWiderMatch() is able to modify the starting position of a match (*startpos), + * but this won't be the case here, as we define iLowLimit==ip, + * so LZ4HC_InsertAndGetWiderMatch() won't be allowed to search past ip */ + return LZ4HC_InsertAndGetWiderMatch(hc4, ip, ip, iLimit, MINMATCH-1, matchpos, &uselessPtr, maxNbAttempts, patternAnalysis, 0 /*chainSwap*/, dict, favorCompressionRatio); +} + +/* LZ4HC_encodeSequence() : + * @return : 0 if ok, + * 1 if buffer issue detected */ +LZ4_FORCE_INLINE int LZ4HC_encodeSequence ( const BYTE** _ip, BYTE** _op, const BYTE** _anchor, int matchLength, const BYTE* const match, - limitedOutput_directive limit, + limitedOutput_directive limit, BYTE* oend) { #define ip (*_ip) #define op (*_op) #define anchor (*_anchor) - size_t length; + size_t length; BYTE* const token = op++; -#if defined(LZ4_DEBUG) && (LZ4_DEBUG >= 6) - static const BYTE* start = NULL; - static U32 totalCost = 0; +#if defined(LZ4_DEBUG) && (LZ4_DEBUG >= 6) + static const BYTE* start = NULL; + static U32 totalCost = 0; U32 const pos = (start==NULL) ? 0 : (U32)(anchor - start); U32 const ll = (U32)(ip - anchor); - U32 const llAdd = (ll>=15) ? ((ll-15) / 255) + 1 : 0; - U32 const mlAdd = (matchLength>=19) ? ((matchLength-19) / 255) + 1 : 0; - U32 const cost = 1 + llAdd + ll + 2 + mlAdd; + U32 const llAdd = (ll>=15) ? ((ll-15) / 255) + 1 : 0; + U32 const mlAdd = (matchLength>=19) ? ((matchLength-19) / 255) + 1 : 0; + U32 const cost = 1 + llAdd + ll + 2 + mlAdd; if (start==NULL) start = anchor; /* only works for single segment */ - /* g_debuglog_enable = (pos >= 2228) & (pos <= 2262); */ + /* g_debuglog_enable = (pos >= 2228) & (pos <= 2262); */ DEBUGLOG(6, "pos:%7u -- literals:%4u, match:%4i, offset:%5u, cost:%4u + %5u", - pos, + pos, (U32)(ip - anchor), matchLength, (U32)(ip-match), - cost, totalCost); - totalCost += cost; + cost, totalCost); + totalCost += cost; #endif /* Encode Literal length */ @@ -502,14 +502,14 @@ LZ4_FORCE_INLINE int LZ4HC_encodeSequence ( (int)length, (int)(oend - op)); return 1; } - if (length >= RUN_MASK) { - size_t len = length - RUN_MASK; - *token = (RUN_MASK << ML_BITS); + if (length >= RUN_MASK) { + size_t len = length - RUN_MASK; + *token = (RUN_MASK << ML_BITS); for(; len >= 255 ; len -= 255) *op++ = 255; *op++ = (BYTE)len; - } else { - *token = (BYTE)(length << ML_BITS); - } + } else { + *token = (BYTE)(length << ML_BITS); + } /* Copy Literals */ LZ4_wildCopy8(op, anchor, op + length); @@ -520,13 +520,13 @@ LZ4_FORCE_INLINE int LZ4HC_encodeSequence ( LZ4_writeLE16(op, (U16)(ip - match)); op += 2; /* Encode MatchLength */ - assert(matchLength >= MINMATCH); - length = (size_t)matchLength - MINMATCH; + assert(matchLength >= MINMATCH); + length = (size_t)matchLength - MINMATCH; if (limit && (op + (length / 255) + (1 + LASTLITERALS) > oend)) { DEBUGLOG(6, "Not enough room to write match length"); return 1; /* Check output limit */ } - if (length >= ML_MASK) { + if (length >= ML_MASK) { *token += ML_MASK; length -= ML_MASK; for(; length >= 510 ; length -= 510) { *op++ = 255; *op++ = 255; } @@ -546,33 +546,33 @@ LZ4_FORCE_INLINE int LZ4HC_encodeSequence ( #undef op #undef anchor -LZ4_FORCE_INLINE int LZ4HC_compress_hashChain ( +LZ4_FORCE_INLINE int LZ4HC_compress_hashChain ( LZ4HC_CCtx_internal* const ctx, const char* const source, char* const dest, - int* srcSizePtr, + int* srcSizePtr, int const maxOutputSize, int maxNbAttempts, - const limitedOutput_directive limit, - const dictCtx_directive dict + const limitedOutput_directive limit, + const dictCtx_directive dict ) { - const int inputSize = *srcSizePtr; - const int patternAnalysis = (maxNbAttempts > 128); /* levels 9+ */ - + const int inputSize = *srcSizePtr; + const int patternAnalysis = (maxNbAttempts > 128); /* levels 9+ */ + const BYTE* ip = (const BYTE*) source; const BYTE* anchor = ip; const BYTE* const iend = ip + inputSize; const BYTE* const mflimit = iend - MFLIMIT; const BYTE* const matchlimit = (iend - LASTLITERALS); - BYTE* optr = (BYTE*) dest; + BYTE* optr = (BYTE*) dest; BYTE* op = (BYTE*) dest; - BYTE* oend = op + maxOutputSize; + BYTE* oend = op + maxOutputSize; - int ml0, ml, ml2, ml3; - const BYTE* start0; - const BYTE* ref0; + int ml0, ml, ml2, ml3; + const BYTE* start0; + const BYTE* ref0; const BYTE* ref = NULL; const BYTE* start2 = NULL; const BYTE* ref2 = NULL; @@ -580,37 +580,37 @@ LZ4_FORCE_INLINE int LZ4HC_compress_hashChain ( const BYTE* ref3 = NULL; /* init */ - *srcSizePtr = 0; - if (limit == fillOutput) oend -= LASTLITERALS; /* Hack for support LZ4 format restriction */ + *srcSizePtr = 0; + if (limit == fillOutput) oend -= LASTLITERALS; /* Hack for support LZ4 format restriction */ if (inputSize < LZ4_minLength) goto _last_literals; /* Input too small, no compression (all literals) */ /* Main Loop */ - while (ip <= mflimit) { - ml = LZ4HC_InsertAndFindBestMatch(ctx, ip, matchlimit, &ref, maxNbAttempts, patternAnalysis, dict); - if (ml<MINMATCH) { ip++; continue; } + while (ip <= mflimit) { + ml = LZ4HC_InsertAndFindBestMatch(ctx, ip, matchlimit, &ref, maxNbAttempts, patternAnalysis, dict); + if (ml<MINMATCH) { ip++; continue; } /* saved, in case we would skip too much */ - start0 = ip; ref0 = ref; ml0 = ml; + start0 = ip; ref0 = ref; ml0 = ml; _Search2: - if (ip+ml <= mflimit) { - ml2 = LZ4HC_InsertAndGetWiderMatch(ctx, - ip + ml - 2, ip + 0, matchlimit, ml, &ref2, &start2, - maxNbAttempts, patternAnalysis, 0, dict, favorCompressionRatio); - } else { - ml2 = ml; - } - - if (ml2 == ml) { /* No better match => encode ML1 */ - optr = op; - if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, limit, oend)) goto _dest_overflow; + if (ip+ml <= mflimit) { + ml2 = LZ4HC_InsertAndGetWiderMatch(ctx, + ip + ml - 2, ip + 0, matchlimit, ml, &ref2, &start2, + maxNbAttempts, patternAnalysis, 0, dict, favorCompressionRatio); + } else { + ml2 = ml; + } + + if (ml2 == ml) { /* No better match => encode ML1 */ + optr = op; + if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, limit, oend)) goto _dest_overflow; continue; } - if (start0 < ip) { /* first match was skipped at least once */ - if (start2 < ip + ml0) { /* squeezing ML1 between ML0(original ML1) and ML2 */ - ip = start0; ref = ref0; ml = ml0; /* restore initial ML1 */ - } } + if (start0 < ip) { /* first match was skipped at least once */ + if (start2 < ip + ml0) { /* squeezing ML1 between ML0(original ML1) and ML2 */ + ip = start0; ref = ref0; ml = ml0; /* restore initial ML1 */ + } } /* Here, start0==ip */ if ((start2 - ip) < 3) { /* First Match too small : removed */ @@ -621,9 +621,9 @@ _Search2: } _Search3: - /* At this stage, we have : - * ml2 > ml1, and - * ip1+3 <= ip2 (usually < ip1+ml1) */ + /* At this stage, we have : + * ml2 > ml1, and + * ip1+3 <= ip2 (usually < ip1+ml1) */ if ((start2 - ip) < OPTIMAL_ML) { int correction; int new_ml = ml; @@ -638,22 +638,22 @@ _Search3: } /* Now, we have start2 = ip+new_ml, with new_ml = min(ml, OPTIMAL_ML=18) */ - if (start2 + ml2 <= mflimit) { - ml3 = LZ4HC_InsertAndGetWiderMatch(ctx, - start2 + ml2 - 3, start2, matchlimit, ml2, &ref3, &start3, - maxNbAttempts, patternAnalysis, 0, dict, favorCompressionRatio); - } else { - ml3 = ml2; - } + if (start2 + ml2 <= mflimit) { + ml3 = LZ4HC_InsertAndGetWiderMatch(ctx, + start2 + ml2 - 3, start2, matchlimit, ml2, &ref3, &start3, + maxNbAttempts, patternAnalysis, 0, dict, favorCompressionRatio); + } else { + ml3 = ml2; + } - if (ml3 == ml2) { /* No better match => encode ML1 and ML2 */ + if (ml3 == ml2) { /* No better match => encode ML1 and ML2 */ /* ip & ref are known; Now for ml */ if (start2 < ip+ml) ml = (int)(start2 - ip); /* Now, encode 2 sequences */ - optr = op; - if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, limit, oend)) goto _dest_overflow; + optr = op; + if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, limit, oend)) goto _dest_overflow; ip = start2; - optr = op; + optr = op; if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml2, ref2, limit, oend)) { ml = ml2; ref = ref2; @@ -676,8 +676,8 @@ _Search3: } } - optr = op; - if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, limit, oend)) goto _dest_overflow; + optr = op; + if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, limit, oend)) goto _dest_overflow; ip = start3; ref = ref3; ml = ml3; @@ -695,12 +695,12 @@ _Search3: } /* - * OK, now we have 3 ascending matches; - * let's write the first one ML1. - * ip & ref are known; Now decide ml. + * OK, now we have 3 ascending matches; + * let's write the first one ML1. + * ip & ref are known; Now decide ml. */ if (start2 < ip+ml) { - if ((start2 - ip) < OPTIMAL_ML) { + if ((start2 - ip) < OPTIMAL_ML) { int correction; if (ml > OPTIMAL_ML) ml = OPTIMAL_ML; if (ip + ml > start2 + ml2 - MINMATCH) ml = (int)(start2 - ip) + ml2 - MINMATCH; @@ -714,60 +714,60 @@ _Search3: ml = (int)(start2 - ip); } } - optr = op; - if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, limit, oend)) goto _dest_overflow; + optr = op; + if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, limit, oend)) goto _dest_overflow; - /* ML2 becomes ML1 */ - ip = start2; ref = ref2; ml = ml2; + /* ML2 becomes ML1 */ + ip = start2; ref = ref2; ml = ml2; - /* ML3 becomes ML2 */ - start2 = start3; ref2 = ref3; ml2 = ml3; + /* ML3 becomes ML2 */ + start2 = start3; ref2 = ref3; ml2 = ml3; - /* let's find a new ML3 */ + /* let's find a new ML3 */ goto _Search3; } -_last_literals: +_last_literals: /* Encode Last Literals */ - { size_t lastRunSize = (size_t)(iend - anchor); /* literals */ + { size_t lastRunSize = (size_t)(iend - anchor); /* literals */ size_t llAdd = (lastRunSize + 255 - RUN_MASK) / 255; size_t const totalSize = 1 + llAdd + lastRunSize; - if (limit == fillOutput) oend += LASTLITERALS; /* restore correct value */ - if (limit && (op + totalSize > oend)) { + if (limit == fillOutput) oend += LASTLITERALS; /* restore correct value */ + if (limit && (op + totalSize > oend)) { if (limit == limitedOutput) return 0; - /* adapt lastRunSize to fill 'dest' */ + /* adapt lastRunSize to fill 'dest' */ lastRunSize = (size_t)(oend - op) - 1 /*token*/; llAdd = (lastRunSize + 256 - RUN_MASK) / 256; lastRunSize -= llAdd; - } + } DEBUGLOG(6, "Final literal run : %i literals", (int)lastRunSize); ip = anchor + lastRunSize; /* can be != iend if limit==fillOutput */ - - if (lastRunSize >= RUN_MASK) { - size_t accumulator = lastRunSize - RUN_MASK; - *op++ = (RUN_MASK << ML_BITS); - for(; accumulator >= 255 ; accumulator -= 255) *op++ = 255; - *op++ = (BYTE) accumulator; - } else { - *op++ = (BYTE)(lastRunSize << ML_BITS); - } - memcpy(op, anchor, lastRunSize); - op += lastRunSize; + + if (lastRunSize >= RUN_MASK) { + size_t accumulator = lastRunSize - RUN_MASK; + *op++ = (RUN_MASK << ML_BITS); + for(; accumulator >= 255 ; accumulator -= 255) *op++ = 255; + *op++ = (BYTE) accumulator; + } else { + *op++ = (BYTE)(lastRunSize << ML_BITS); + } + memcpy(op, anchor, lastRunSize); + op += lastRunSize; } /* End */ - *srcSizePtr = (int) (((const char*)ip) - source); + *srcSizePtr = (int) (((const char*)ip) - source); return (int) (((char*)op)-dest); -_dest_overflow: - if (limit == fillOutput) { +_dest_overflow: + if (limit == fillOutput) { /* Assumption : ip, anchor, ml and ref must be set correctly */ size_t const ll = (size_t)(ip - anchor); size_t const ll_addbytes = (ll + 240) / 255; size_t const ll_totalCost = 1 + ll_addbytes + ll; BYTE* const maxLitPos = oend - 3; /* 2 for offset, 1 for token */ DEBUGLOG(6, "Last sequence overflowing"); - op = optr; /* restore correct out pointer */ + op = optr; /* restore correct out pointer */ if (op + ll_totalCost <= maxLitPos) { /* ll validated; now adjust match length */ size_t const bytesLeftForMl = (size_t)(maxLitPos - (op+ll_totalCost)); @@ -777,318 +777,318 @@ _dest_overflow: if ((oend + LASTLITERALS) - (op + ll_totalCost + 2) - 1 + ml >= MFLIMIT) { LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, notLimited, oend); } } - goto _last_literals; + goto _last_literals; } /* compression failed */ - return 0; + return 0; } - -static int LZ4HC_compress_optimal( LZ4HC_CCtx_internal* ctx, - const char* const source, char* dst, - int* srcSizePtr, int dstCapacity, - int const nbSearches, size_t sufficient_len, - const limitedOutput_directive limit, int const fullUpdate, - const dictCtx_directive dict, + +static int LZ4HC_compress_optimal( LZ4HC_CCtx_internal* ctx, + const char* const source, char* dst, + int* srcSizePtr, int dstCapacity, + int const nbSearches, size_t sufficient_len, + const limitedOutput_directive limit, int const fullUpdate, + const dictCtx_directive dict, const HCfavor_e favorDecSpeed); - - -LZ4_FORCE_INLINE int LZ4HC_compress_generic_internal ( + + +LZ4_FORCE_INLINE int LZ4HC_compress_generic_internal ( LZ4HC_CCtx_internal* const ctx, - const char* const src, - char* const dst, - int* const srcSizePtr, - int const dstCapacity, - int cLevel, - const limitedOutput_directive limit, - const dictCtx_directive dict + const char* const src, + char* const dst, + int* const srcSizePtr, + int const dstCapacity, + int cLevel, + const limitedOutput_directive limit, + const dictCtx_directive dict ) { - typedef enum { lz4hc, lz4opt } lz4hc_strat_e; - typedef struct { - lz4hc_strat_e strat; + typedef enum { lz4hc, lz4opt } lz4hc_strat_e; + typedef struct { + lz4hc_strat_e strat; int nbSearches; - U32 targetLength; - } cParams_t; - static const cParams_t clTable[LZ4HC_CLEVEL_MAX+1] = { - { lz4hc, 2, 16 }, /* 0, unused */ - { lz4hc, 2, 16 }, /* 1, unused */ - { lz4hc, 2, 16 }, /* 2, unused */ - { lz4hc, 4, 16 }, /* 3 */ - { lz4hc, 8, 16 }, /* 4 */ - { lz4hc, 16, 16 }, /* 5 */ - { lz4hc, 32, 16 }, /* 6 */ - { lz4hc, 64, 16 }, /* 7 */ - { lz4hc, 128, 16 }, /* 8 */ - { lz4hc, 256, 16 }, /* 9 */ - { lz4opt, 96, 64 }, /*10==LZ4HC_CLEVEL_OPT_MIN*/ - { lz4opt, 512,128 }, /*11 */ - { lz4opt,16384,LZ4_OPT_NUM }, /* 12==LZ4HC_CLEVEL_MAX */ - }; - + U32 targetLength; + } cParams_t; + static const cParams_t clTable[LZ4HC_CLEVEL_MAX+1] = { + { lz4hc, 2, 16 }, /* 0, unused */ + { lz4hc, 2, 16 }, /* 1, unused */ + { lz4hc, 2, 16 }, /* 2, unused */ + { lz4hc, 4, 16 }, /* 3 */ + { lz4hc, 8, 16 }, /* 4 */ + { lz4hc, 16, 16 }, /* 5 */ + { lz4hc, 32, 16 }, /* 6 */ + { lz4hc, 64, 16 }, /* 7 */ + { lz4hc, 128, 16 }, /* 8 */ + { lz4hc, 256, 16 }, /* 9 */ + { lz4opt, 96, 64 }, /*10==LZ4HC_CLEVEL_OPT_MIN*/ + { lz4opt, 512,128 }, /*11 */ + { lz4opt,16384,LZ4_OPT_NUM }, /* 12==LZ4HC_CLEVEL_MAX */ + }; + DEBUGLOG(4, "LZ4HC_compress_generic(ctx=%p, src=%p, srcSize=%d, limit=%d)", ctx, src, *srcSizePtr, limit); - - if (limit == fillOutput && dstCapacity < 1) return 0; /* Impossible to store anything */ - if ((U32)*srcSizePtr > (U32)LZ4_MAX_INPUT_SIZE) return 0; /* Unsupported input size (too large or negative) */ - - ctx->end += *srcSizePtr; - if (cLevel < 1) cLevel = LZ4HC_CLEVEL_DEFAULT; /* note : convention is different from lz4frame, maybe something to review */ - cLevel = MIN(LZ4HC_CLEVEL_MAX, cLevel); - { cParams_t const cParam = clTable[cLevel]; - HCfavor_e const favor = ctx->favorDecSpeed ? favorDecompressionSpeed : favorCompressionRatio; - int result; - - if (cParam.strat == lz4hc) { - result = LZ4HC_compress_hashChain(ctx, - src, dst, srcSizePtr, dstCapacity, - cParam.nbSearches, limit, dict); - } else { - assert(cParam.strat == lz4opt); - result = LZ4HC_compress_optimal(ctx, - src, dst, srcSizePtr, dstCapacity, + + if (limit == fillOutput && dstCapacity < 1) return 0; /* Impossible to store anything */ + if ((U32)*srcSizePtr > (U32)LZ4_MAX_INPUT_SIZE) return 0; /* Unsupported input size (too large or negative) */ + + ctx->end += *srcSizePtr; + if (cLevel < 1) cLevel = LZ4HC_CLEVEL_DEFAULT; /* note : convention is different from lz4frame, maybe something to review */ + cLevel = MIN(LZ4HC_CLEVEL_MAX, cLevel); + { cParams_t const cParam = clTable[cLevel]; + HCfavor_e const favor = ctx->favorDecSpeed ? favorDecompressionSpeed : favorCompressionRatio; + int result; + + if (cParam.strat == lz4hc) { + result = LZ4HC_compress_hashChain(ctx, + src, dst, srcSizePtr, dstCapacity, + cParam.nbSearches, limit, dict); + } else { + assert(cParam.strat == lz4opt); + result = LZ4HC_compress_optimal(ctx, + src, dst, srcSizePtr, dstCapacity, cParam.nbSearches, cParam.targetLength, limit, - cLevel == LZ4HC_CLEVEL_MAX, /* ultra mode */ - dict, favor); + cLevel == LZ4HC_CLEVEL_MAX, /* ultra mode */ + dict, favor); } - if (result <= 0) ctx->dirty = 1; - return result; + if (result <= 0) ctx->dirty = 1; + return result; } } -static void LZ4HC_setExternalDict(LZ4HC_CCtx_internal* ctxPtr, const BYTE* newBlock); - -static int -LZ4HC_compress_generic_noDictCtx ( - LZ4HC_CCtx_internal* const ctx, - const char* const src, - char* const dst, - int* const srcSizePtr, - int const dstCapacity, - int cLevel, - limitedOutput_directive limit - ) -{ - assert(ctx->dictCtx == NULL); - return LZ4HC_compress_generic_internal(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit, noDictCtx); -} - -static int -LZ4HC_compress_generic_dictCtx ( - LZ4HC_CCtx_internal* const ctx, - const char* const src, - char* const dst, - int* const srcSizePtr, - int const dstCapacity, - int cLevel, - limitedOutput_directive limit - ) -{ - const size_t position = (size_t)(ctx->end - ctx->base) - ctx->lowLimit; - assert(ctx->dictCtx != NULL); - if (position >= 64 KB) { - ctx->dictCtx = NULL; - return LZ4HC_compress_generic_noDictCtx(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit); - } else if (position == 0 && *srcSizePtr > 4 KB) { - memcpy(ctx, ctx->dictCtx, sizeof(LZ4HC_CCtx_internal)); - LZ4HC_setExternalDict(ctx, (const BYTE *)src); - ctx->compressionLevel = (short)cLevel; - return LZ4HC_compress_generic_noDictCtx(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit); - } else { - return LZ4HC_compress_generic_internal(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit, usingDictCtxHc); - } -} - -static int -LZ4HC_compress_generic ( - LZ4HC_CCtx_internal* const ctx, - const char* const src, - char* const dst, - int* const srcSizePtr, - int const dstCapacity, - int cLevel, - limitedOutput_directive limit - ) -{ - if (ctx->dictCtx == NULL) { - return LZ4HC_compress_generic_noDictCtx(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit); - } else { - return LZ4HC_compress_generic_dictCtx(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit); - } -} - - -int LZ4_sizeofStateHC(void) { return (int)sizeof(LZ4_streamHC_t); } - -static size_t LZ4_streamHC_t_alignment(void) +static void LZ4HC_setExternalDict(LZ4HC_CCtx_internal* ctxPtr, const BYTE* newBlock); + +static int +LZ4HC_compress_generic_noDictCtx ( + LZ4HC_CCtx_internal* const ctx, + const char* const src, + char* const dst, + int* const srcSizePtr, + int const dstCapacity, + int cLevel, + limitedOutput_directive limit + ) +{ + assert(ctx->dictCtx == NULL); + return LZ4HC_compress_generic_internal(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit, noDictCtx); +} + +static int +LZ4HC_compress_generic_dictCtx ( + LZ4HC_CCtx_internal* const ctx, + const char* const src, + char* const dst, + int* const srcSizePtr, + int const dstCapacity, + int cLevel, + limitedOutput_directive limit + ) { + const size_t position = (size_t)(ctx->end - ctx->base) - ctx->lowLimit; + assert(ctx->dictCtx != NULL); + if (position >= 64 KB) { + ctx->dictCtx = NULL; + return LZ4HC_compress_generic_noDictCtx(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit); + } else if (position == 0 && *srcSizePtr > 4 KB) { + memcpy(ctx, ctx->dictCtx, sizeof(LZ4HC_CCtx_internal)); + LZ4HC_setExternalDict(ctx, (const BYTE *)src); + ctx->compressionLevel = (short)cLevel; + return LZ4HC_compress_generic_noDictCtx(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit); + } else { + return LZ4HC_compress_generic_internal(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit, usingDictCtxHc); + } +} + +static int +LZ4HC_compress_generic ( + LZ4HC_CCtx_internal* const ctx, + const char* const src, + char* const dst, + int* const srcSizePtr, + int const dstCapacity, + int cLevel, + limitedOutput_directive limit + ) +{ + if (ctx->dictCtx == NULL) { + return LZ4HC_compress_generic_noDictCtx(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit); + } else { + return LZ4HC_compress_generic_dictCtx(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit); + } +} + + +int LZ4_sizeofStateHC(void) { return (int)sizeof(LZ4_streamHC_t); } + +static size_t LZ4_streamHC_t_alignment(void) +{ #if LZ4_ALIGN_TEST typedef struct { char c; LZ4_streamHC_t t; } t_a; return sizeof(t_a) - sizeof(LZ4_streamHC_t); #else return 1; /* effectively disabled */ #endif -} - -/* state is presumed correctly initialized, - * in which case its size and alignment have already been validate */ -int LZ4_compress_HC_extStateHC_fastReset (void* state, const char* src, char* dst, int srcSize, int dstCapacity, int compressionLevel) -{ - LZ4HC_CCtx_internal* const ctx = &((LZ4_streamHC_t*)state)->internal_donotuse; +} + +/* state is presumed correctly initialized, + * in which case its size and alignment have already been validate */ +int LZ4_compress_HC_extStateHC_fastReset (void* state, const char* src, char* dst, int srcSize, int dstCapacity, int compressionLevel) +{ + LZ4HC_CCtx_internal* const ctx = &((LZ4_streamHC_t*)state)->internal_donotuse; if (!LZ4_isAligned(state, LZ4_streamHC_t_alignment())) return 0; - LZ4_resetStreamHC_fast((LZ4_streamHC_t*)state, compressionLevel); - LZ4HC_init_internal (ctx, (const BYTE*)src); - if (dstCapacity < LZ4_compressBound(srcSize)) - return LZ4HC_compress_generic (ctx, src, dst, &srcSize, dstCapacity, compressionLevel, limitedOutput); + LZ4_resetStreamHC_fast((LZ4_streamHC_t*)state, compressionLevel); + LZ4HC_init_internal (ctx, (const BYTE*)src); + if (dstCapacity < LZ4_compressBound(srcSize)) + return LZ4HC_compress_generic (ctx, src, dst, &srcSize, dstCapacity, compressionLevel, limitedOutput); else - return LZ4HC_compress_generic (ctx, src, dst, &srcSize, dstCapacity, compressionLevel, notLimited); -} - -int LZ4_compress_HC_extStateHC (void* state, const char* src, char* dst, int srcSize, int dstCapacity, int compressionLevel) -{ - LZ4_streamHC_t* const ctx = LZ4_initStreamHC(state, sizeof(*ctx)); - if (ctx==NULL) return 0; /* init failure */ - return LZ4_compress_HC_extStateHC_fastReset(state, src, dst, srcSize, dstCapacity, compressionLevel); + return LZ4HC_compress_generic (ctx, src, dst, &srcSize, dstCapacity, compressionLevel, notLimited); } -int LZ4_compress_HC(const char* src, char* dst, int srcSize, int dstCapacity, int compressionLevel) +int LZ4_compress_HC_extStateHC (void* state, const char* src, char* dst, int srcSize, int dstCapacity, int compressionLevel) { + LZ4_streamHC_t* const ctx = LZ4_initStreamHC(state, sizeof(*ctx)); + if (ctx==NULL) return 0; /* init failure */ + return LZ4_compress_HC_extStateHC_fastReset(state, src, dst, srcSize, dstCapacity, compressionLevel); +} + +int LZ4_compress_HC(const char* src, char* dst, int srcSize, int dstCapacity, int compressionLevel) +{ #if defined(LZ4HC_HEAPMODE) && LZ4HC_HEAPMODE==1 - LZ4_streamHC_t* const statePtr = (LZ4_streamHC_t*)ALLOC(sizeof(LZ4_streamHC_t)); + LZ4_streamHC_t* const statePtr = (LZ4_streamHC_t*)ALLOC(sizeof(LZ4_streamHC_t)); #else LZ4_streamHC_t state; LZ4_streamHC_t* const statePtr = &state; #endif - int const cSize = LZ4_compress_HC_extStateHC(statePtr, src, dst, srcSize, dstCapacity, compressionLevel); + int const cSize = LZ4_compress_HC_extStateHC(statePtr, src, dst, srcSize, dstCapacity, compressionLevel); #if defined(LZ4HC_HEAPMODE) && LZ4HC_HEAPMODE==1 - FREEMEM(statePtr); + FREEMEM(statePtr); #endif return cSize; } -/* state is presumed sized correctly (>= sizeof(LZ4_streamHC_t)) */ -int LZ4_compress_HC_destSize(void* state, const char* source, char* dest, int* sourceSizePtr, int targetDestSize, int cLevel) -{ - LZ4_streamHC_t* const ctx = LZ4_initStreamHC(state, sizeof(*ctx)); - if (ctx==NULL) return 0; /* init failure */ - LZ4HC_init_internal(&ctx->internal_donotuse, (const BYTE*) source); - LZ4_setCompressionLevel(ctx, cLevel); - return LZ4HC_compress_generic(&ctx->internal_donotuse, source, dest, sourceSizePtr, targetDestSize, cLevel, fillOutput); -} - +/* state is presumed sized correctly (>= sizeof(LZ4_streamHC_t)) */ +int LZ4_compress_HC_destSize(void* state, const char* source, char* dest, int* sourceSizePtr, int targetDestSize, int cLevel) +{ + LZ4_streamHC_t* const ctx = LZ4_initStreamHC(state, sizeof(*ctx)); + if (ctx==NULL) return 0; /* init failure */ + LZ4HC_init_internal(&ctx->internal_donotuse, (const BYTE*) source); + LZ4_setCompressionLevel(ctx, cLevel); + return LZ4HC_compress_generic(&ctx->internal_donotuse, source, dest, sourceSizePtr, targetDestSize, cLevel, fillOutput); +} + /************************************** * Streaming Functions **************************************/ /* allocation */ -LZ4_streamHC_t* LZ4_createStreamHC(void) -{ +LZ4_streamHC_t* LZ4_createStreamHC(void) +{ LZ4_streamHC_t* const state = (LZ4_streamHC_t*)ALLOC_AND_ZERO(sizeof(LZ4_streamHC_t)); if (state == NULL) return NULL; LZ4_setCompressionLevel(state, LZ4HC_CLEVEL_DEFAULT); return state; -} - -int LZ4_freeStreamHC (LZ4_streamHC_t* LZ4_streamHCPtr) +} + +int LZ4_freeStreamHC (LZ4_streamHC_t* LZ4_streamHCPtr) +{ + DEBUGLOG(4, "LZ4_freeStreamHC(%p)", LZ4_streamHCPtr); + if (!LZ4_streamHCPtr) return 0; /* support free on NULL */ + FREEMEM(LZ4_streamHCPtr); + return 0; +} + + +LZ4_streamHC_t* LZ4_initStreamHC (void* buffer, size_t size) { - DEBUGLOG(4, "LZ4_freeStreamHC(%p)", LZ4_streamHCPtr); - if (!LZ4_streamHCPtr) return 0; /* support free on NULL */ - FREEMEM(LZ4_streamHCPtr); - return 0; -} - - -LZ4_streamHC_t* LZ4_initStreamHC (void* buffer, size_t size) -{ - LZ4_streamHC_t* const LZ4_streamHCPtr = (LZ4_streamHC_t*)buffer; + LZ4_streamHC_t* const LZ4_streamHCPtr = (LZ4_streamHC_t*)buffer; /* if compilation fails here, LZ4_STREAMHCSIZE must be increased */ LZ4_STATIC_ASSERT(sizeof(LZ4HC_CCtx_internal) <= LZ4_STREAMHCSIZE); DEBUGLOG(4, "LZ4_initStreamHC(%p, %u)", buffer, (unsigned)size); /* check conditions */ - if (buffer == NULL) return NULL; - if (size < sizeof(LZ4_streamHC_t)) return NULL; + if (buffer == NULL) return NULL; + if (size < sizeof(LZ4_streamHC_t)) return NULL; if (!LZ4_isAligned(buffer, LZ4_streamHC_t_alignment())) return NULL; /* init */ { LZ4HC_CCtx_internal* const hcstate = &(LZ4_streamHCPtr->internal_donotuse); MEM_INIT(hcstate, 0, sizeof(*hcstate)); } - LZ4_setCompressionLevel(LZ4_streamHCPtr, LZ4HC_CLEVEL_DEFAULT); - return LZ4_streamHCPtr; -} - -/* just a stub */ -void LZ4_resetStreamHC (LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel) -{ - LZ4_initStreamHC(LZ4_streamHCPtr, sizeof(*LZ4_streamHCPtr)); - LZ4_setCompressionLevel(LZ4_streamHCPtr, compressionLevel); + LZ4_setCompressionLevel(LZ4_streamHCPtr, LZ4HC_CLEVEL_DEFAULT); + return LZ4_streamHCPtr; } -void LZ4_resetStreamHC_fast (LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel) +/* just a stub */ +void LZ4_resetStreamHC (LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel) { - DEBUGLOG(4, "LZ4_resetStreamHC_fast(%p, %d)", LZ4_streamHCPtr, compressionLevel); - if (LZ4_streamHCPtr->internal_donotuse.dirty) { - LZ4_initStreamHC(LZ4_streamHCPtr, sizeof(*LZ4_streamHCPtr)); - } else { - /* preserve end - base : can trigger clearTable's threshold */ + LZ4_initStreamHC(LZ4_streamHCPtr, sizeof(*LZ4_streamHCPtr)); + LZ4_setCompressionLevel(LZ4_streamHCPtr, compressionLevel); +} + +void LZ4_resetStreamHC_fast (LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel) +{ + DEBUGLOG(4, "LZ4_resetStreamHC_fast(%p, %d)", LZ4_streamHCPtr, compressionLevel); + if (LZ4_streamHCPtr->internal_donotuse.dirty) { + LZ4_initStreamHC(LZ4_streamHCPtr, sizeof(*LZ4_streamHCPtr)); + } else { + /* preserve end - base : can trigger clearTable's threshold */ if (LZ4_streamHCPtr->internal_donotuse.end != NULL) { LZ4_streamHCPtr->internal_donotuse.end -= (uptrval)LZ4_streamHCPtr->internal_donotuse.base; } else { assert(LZ4_streamHCPtr->internal_donotuse.base == NULL); } - LZ4_streamHCPtr->internal_donotuse.base = NULL; - LZ4_streamHCPtr->internal_donotuse.dictCtx = NULL; - } - LZ4_setCompressionLevel(LZ4_streamHCPtr, compressionLevel); -} - -void LZ4_setCompressionLevel(LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel) -{ - DEBUGLOG(5, "LZ4_setCompressionLevel(%p, %d)", LZ4_streamHCPtr, compressionLevel); - if (compressionLevel < 1) compressionLevel = LZ4HC_CLEVEL_DEFAULT; - if (compressionLevel > LZ4HC_CLEVEL_MAX) compressionLevel = LZ4HC_CLEVEL_MAX; - LZ4_streamHCPtr->internal_donotuse.compressionLevel = (short)compressionLevel; -} - -void LZ4_favorDecompressionSpeed(LZ4_streamHC_t* LZ4_streamHCPtr, int favor) -{ - LZ4_streamHCPtr->internal_donotuse.favorDecSpeed = (favor!=0); -} - -/* LZ4_loadDictHC() : - * LZ4_streamHCPtr is presumed properly initialized */ -int LZ4_loadDictHC (LZ4_streamHC_t* LZ4_streamHCPtr, - const char* dictionary, int dictSize) -{ - LZ4HC_CCtx_internal* const ctxPtr = &LZ4_streamHCPtr->internal_donotuse; + LZ4_streamHCPtr->internal_donotuse.base = NULL; + LZ4_streamHCPtr->internal_donotuse.dictCtx = NULL; + } + LZ4_setCompressionLevel(LZ4_streamHCPtr, compressionLevel); +} + +void LZ4_setCompressionLevel(LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel) +{ + DEBUGLOG(5, "LZ4_setCompressionLevel(%p, %d)", LZ4_streamHCPtr, compressionLevel); + if (compressionLevel < 1) compressionLevel = LZ4HC_CLEVEL_DEFAULT; + if (compressionLevel > LZ4HC_CLEVEL_MAX) compressionLevel = LZ4HC_CLEVEL_MAX; + LZ4_streamHCPtr->internal_donotuse.compressionLevel = (short)compressionLevel; +} + +void LZ4_favorDecompressionSpeed(LZ4_streamHC_t* LZ4_streamHCPtr, int favor) +{ + LZ4_streamHCPtr->internal_donotuse.favorDecSpeed = (favor!=0); +} + +/* LZ4_loadDictHC() : + * LZ4_streamHCPtr is presumed properly initialized */ +int LZ4_loadDictHC (LZ4_streamHC_t* LZ4_streamHCPtr, + const char* dictionary, int dictSize) +{ + LZ4HC_CCtx_internal* const ctxPtr = &LZ4_streamHCPtr->internal_donotuse; DEBUGLOG(4, "LZ4_loadDictHC(ctx:%p, dict:%p, dictSize:%d)", LZ4_streamHCPtr, dictionary, dictSize); - assert(LZ4_streamHCPtr != NULL); + assert(LZ4_streamHCPtr != NULL); if (dictSize > 64 KB) { - dictionary += (size_t)dictSize - 64 KB; + dictionary += (size_t)dictSize - 64 KB; dictSize = 64 KB; } - /* need a full initialization, there are bad side-effects when using resetFast() */ - { int const cLevel = ctxPtr->compressionLevel; - LZ4_initStreamHC(LZ4_streamHCPtr, sizeof(*LZ4_streamHCPtr)); - LZ4_setCompressionLevel(LZ4_streamHCPtr, cLevel); - } - LZ4HC_init_internal (ctxPtr, (const BYTE*)dictionary); + /* need a full initialization, there are bad side-effects when using resetFast() */ + { int const cLevel = ctxPtr->compressionLevel; + LZ4_initStreamHC(LZ4_streamHCPtr, sizeof(*LZ4_streamHCPtr)); + LZ4_setCompressionLevel(LZ4_streamHCPtr, cLevel); + } + LZ4HC_init_internal (ctxPtr, (const BYTE*)dictionary); ctxPtr->end = (const BYTE*)dictionary + dictSize; - if (dictSize >= 4) LZ4HC_Insert (ctxPtr, ctxPtr->end-3); + if (dictSize >= 4) LZ4HC_Insert (ctxPtr, ctxPtr->end-3); return dictSize; } -void LZ4_attach_HC_dictionary(LZ4_streamHC_t *working_stream, const LZ4_streamHC_t *dictionary_stream) { - working_stream->internal_donotuse.dictCtx = dictionary_stream != NULL ? &(dictionary_stream->internal_donotuse) : NULL; -} +void LZ4_attach_HC_dictionary(LZ4_streamHC_t *working_stream, const LZ4_streamHC_t *dictionary_stream) { + working_stream->internal_donotuse.dictCtx = dictionary_stream != NULL ? &(dictionary_stream->internal_donotuse) : NULL; +} /* compression */ static void LZ4HC_setExternalDict(LZ4HC_CCtx_internal* ctxPtr, const BYTE* newBlock) { - DEBUGLOG(4, "LZ4HC_setExternalDict(%p, %p)", ctxPtr, newBlock); - if (ctxPtr->end >= ctxPtr->base + ctxPtr->dictLimit + 4) - LZ4HC_Insert (ctxPtr, ctxPtr->end-3); /* Referencing remaining dictionary content */ + DEBUGLOG(4, "LZ4HC_setExternalDict(%p, %p)", ctxPtr, newBlock); + if (ctxPtr->end >= ctxPtr->base + ctxPtr->dictLimit + 4) + LZ4HC_Insert (ctxPtr, ctxPtr->end-3); /* Referencing remaining dictionary content */ /* Only one memory segment for extDict, so any previous extDict is lost at this stage */ ctxPtr->lowLimit = ctxPtr->dictLimit; @@ -1108,12 +1108,12 @@ LZ4_compressHC_continue_generic (LZ4_streamHC_t* LZ4_streamHCPtr, int* srcSizePtr, int dstCapacity, limitedOutput_directive limit) { - LZ4HC_CCtx_internal* const ctxPtr = &LZ4_streamHCPtr->internal_donotuse; + LZ4HC_CCtx_internal* const ctxPtr = &LZ4_streamHCPtr->internal_donotuse; DEBUGLOG(5, "LZ4_compressHC_continue_generic(ctx=%p, src=%p, srcSize=%d, limit=%d)", LZ4_streamHCPtr, src, *srcSizePtr, limit); - assert(ctxPtr != NULL); + assert(ctxPtr != NULL); /* auto-init if forgotten */ - if (ctxPtr->base == NULL) LZ4HC_init_internal (ctxPtr, (const BYTE*) src); + if (ctxPtr->base == NULL) LZ4HC_init_internal (ctxPtr, (const BYTE*) src); /* Check overflow */ if ((size_t)(ctxPtr->end - ctxPtr->base) > 2 GB) { @@ -1123,37 +1123,37 @@ LZ4_compressHC_continue_generic (LZ4_streamHC_t* LZ4_streamHCPtr, } /* Check if blocks follow each other */ - if ((const BYTE*)src != ctxPtr->end) - LZ4HC_setExternalDict(ctxPtr, (const BYTE*)src); + if ((const BYTE*)src != ctxPtr->end) + LZ4HC_setExternalDict(ctxPtr, (const BYTE*)src); /* Check overlapping input/dictionary space */ - { const BYTE* sourceEnd = (const BYTE*) src + *srcSizePtr; + { const BYTE* sourceEnd = (const BYTE*) src + *srcSizePtr; const BYTE* const dictBegin = ctxPtr->dictBase + ctxPtr->lowLimit; const BYTE* const dictEnd = ctxPtr->dictBase + ctxPtr->dictLimit; - if ((sourceEnd > dictBegin) && ((const BYTE*)src < dictEnd)) { + if ((sourceEnd > dictBegin) && ((const BYTE*)src < dictEnd)) { if (sourceEnd > dictEnd) sourceEnd = dictEnd; ctxPtr->lowLimit = (U32)(sourceEnd - ctxPtr->dictBase); if (ctxPtr->dictLimit - ctxPtr->lowLimit < 4) ctxPtr->lowLimit = ctxPtr->dictLimit; } } - return LZ4HC_compress_generic (ctxPtr, src, dst, srcSizePtr, dstCapacity, ctxPtr->compressionLevel, limit); + return LZ4HC_compress_generic (ctxPtr, src, dst, srcSizePtr, dstCapacity, ctxPtr->compressionLevel, limit); } -int LZ4_compress_HC_continue (LZ4_streamHC_t* LZ4_streamHCPtr, const char* src, char* dst, int srcSize, int dstCapacity) +int LZ4_compress_HC_continue (LZ4_streamHC_t* LZ4_streamHCPtr, const char* src, char* dst, int srcSize, int dstCapacity) { - if (dstCapacity < LZ4_compressBound(srcSize)) - return LZ4_compressHC_continue_generic (LZ4_streamHCPtr, src, dst, &srcSize, dstCapacity, limitedOutput); + if (dstCapacity < LZ4_compressBound(srcSize)) + return LZ4_compressHC_continue_generic (LZ4_streamHCPtr, src, dst, &srcSize, dstCapacity, limitedOutput); else - return LZ4_compressHC_continue_generic (LZ4_streamHCPtr, src, dst, &srcSize, dstCapacity, notLimited); -} - -int LZ4_compress_HC_continue_destSize (LZ4_streamHC_t* LZ4_streamHCPtr, const char* src, char* dst, int* srcSizePtr, int targetDestSize) -{ - return LZ4_compressHC_continue_generic(LZ4_streamHCPtr, src, dst, srcSizePtr, targetDestSize, fillOutput); + return LZ4_compressHC_continue_generic (LZ4_streamHCPtr, src, dst, &srcSize, dstCapacity, notLimited); } +int LZ4_compress_HC_continue_destSize (LZ4_streamHC_t* LZ4_streamHCPtr, const char* src, char* dst, int* srcSizePtr, int targetDestSize) +{ + return LZ4_compressHC_continue_generic(LZ4_streamHCPtr, src, dst, srcSizePtr, targetDestSize, fillOutput); +} - + + /* LZ4_saveDictHC : * save history content * into a user-provided buffer @@ -1174,8 +1174,8 @@ int LZ4_saveDictHC (LZ4_streamHC_t* LZ4_streamHCPtr, char* safeBuffer, int dictS { U32 const endIndex = (U32)(streamPtr->end - streamPtr->base); streamPtr->end = (const BYTE*)safeBuffer + dictSize; streamPtr->base = streamPtr->end - endIndex; - streamPtr->dictLimit = endIndex - (U32)dictSize; - streamPtr->lowLimit = endIndex - (U32)dictSize; + streamPtr->dictLimit = endIndex - (U32)dictSize; + streamPtr->lowLimit = endIndex - (U32)dictSize; if (streamPtr->nextToUpdate < streamPtr->dictLimit) streamPtr->nextToUpdate = streamPtr->dictLimit; } @@ -1183,13 +1183,13 @@ int LZ4_saveDictHC (LZ4_streamHC_t* LZ4_streamHCPtr, char* safeBuffer, int dictS } -/*************************************************** +/*************************************************** * Deprecated Functions -***************************************************/ - +***************************************************/ + /* These functions currently generate deprecation warnings */ - -/* Wrappers for deprecated compression functions */ + +/* Wrappers for deprecated compression functions */ int LZ4_compressHC(const char* src, char* dst, int srcSize) { return LZ4_compress_HC (src, dst, srcSize, LZ4_compressBound(srcSize), 0); } int LZ4_compressHC_limitedOutput(const char* src, char* dst, int srcSize, int maxDstSize) { return LZ4_compress_HC(src, dst, srcSize, maxDstSize, 0); } int LZ4_compressHC2(const char* src, char* dst, int srcSize, int cLevel) { return LZ4_compress_HC (src, dst, srcSize, LZ4_compressBound(srcSize), cLevel); } @@ -1205,389 +1205,389 @@ int LZ4_compressHC_limitedOutput_continue (LZ4_streamHC_t* ctx, const char* src, /* Deprecated streaming functions */ int LZ4_sizeofStreamStateHC(void) { return LZ4_STREAMHCSIZE; } -/* state is presumed correctly sized, aka >= sizeof(LZ4_streamHC_t) - * @return : 0 on success, !=0 if error */ +/* state is presumed correctly sized, aka >= sizeof(LZ4_streamHC_t) + * @return : 0 on success, !=0 if error */ int LZ4_resetStreamStateHC(void* state, char* inputBuffer) { - LZ4_streamHC_t* const hc4 = LZ4_initStreamHC(state, sizeof(*hc4)); - if (hc4 == NULL) return 1; /* init failed */ - LZ4HC_init_internal (&hc4->internal_donotuse, (const BYTE*)inputBuffer); + LZ4_streamHC_t* const hc4 = LZ4_initStreamHC(state, sizeof(*hc4)); + if (hc4 == NULL) return 1; /* init failed */ + LZ4HC_init_internal (&hc4->internal_donotuse, (const BYTE*)inputBuffer); return 0; } -void* LZ4_createHC (const char* inputBuffer) +void* LZ4_createHC (const char* inputBuffer) { - LZ4_streamHC_t* const hc4 = LZ4_createStreamHC(); + LZ4_streamHC_t* const hc4 = LZ4_createStreamHC(); if (hc4 == NULL) return NULL; /* not enough memory */ - LZ4HC_init_internal (&hc4->internal_donotuse, (const BYTE*)inputBuffer); + LZ4HC_init_internal (&hc4->internal_donotuse, (const BYTE*)inputBuffer); return hc4; } -int LZ4_freeHC (void* LZ4HC_Data) -{ - if (!LZ4HC_Data) return 0; /* support free on NULL */ - FREEMEM(LZ4HC_Data); - return 0; -} +int LZ4_freeHC (void* LZ4HC_Data) +{ + if (!LZ4HC_Data) return 0; /* support free on NULL */ + FREEMEM(LZ4HC_Data); + return 0; +} -int LZ4_compressHC2_continue (void* LZ4HC_Data, const char* src, char* dst, int srcSize, int cLevel) +int LZ4_compressHC2_continue (void* LZ4HC_Data, const char* src, char* dst, int srcSize, int cLevel) { - return LZ4HC_compress_generic (&((LZ4_streamHC_t*)LZ4HC_Data)->internal_donotuse, src, dst, &srcSize, 0, cLevel, notLimited); + return LZ4HC_compress_generic (&((LZ4_streamHC_t*)LZ4HC_Data)->internal_donotuse, src, dst, &srcSize, 0, cLevel, notLimited); } -int LZ4_compressHC2_limitedOutput_continue (void* LZ4HC_Data, const char* src, char* dst, int srcSize, int dstCapacity, int cLevel) +int LZ4_compressHC2_limitedOutput_continue (void* LZ4HC_Data, const char* src, char* dst, int srcSize, int dstCapacity, int cLevel) { - return LZ4HC_compress_generic (&((LZ4_streamHC_t*)LZ4HC_Data)->internal_donotuse, src, dst, &srcSize, dstCapacity, cLevel, limitedOutput); + return LZ4HC_compress_generic (&((LZ4_streamHC_t*)LZ4HC_Data)->internal_donotuse, src, dst, &srcSize, dstCapacity, cLevel, limitedOutput); } char* LZ4_slideInputBufferHC(void* LZ4HC_Data) { - LZ4_streamHC_t *ctx = (LZ4_streamHC_t*)LZ4HC_Data; - const BYTE *bufferStart = ctx->internal_donotuse.base + ctx->internal_donotuse.lowLimit; - LZ4_resetStreamHC_fast(ctx, ctx->internal_donotuse.compressionLevel); - /* avoid const char * -> char * conversion warning :( */ - return (char *)(uptrval)bufferStart; + LZ4_streamHC_t *ctx = (LZ4_streamHC_t*)LZ4HC_Data; + const BYTE *bufferStart = ctx->internal_donotuse.base + ctx->internal_donotuse.lowLimit; + LZ4_resetStreamHC_fast(ctx, ctx->internal_donotuse.compressionLevel); + /* avoid const char * -> char * conversion warning :( */ + return (char *)(uptrval)bufferStart; } - - -/* ================================================ - * LZ4 Optimal parser (levels [LZ4HC_CLEVEL_OPT_MIN - LZ4HC_CLEVEL_MAX]) - * ===============================================*/ -typedef struct { - int price; - int off; - int mlen; - int litlen; -} LZ4HC_optimal_t; - -/* price in bytes */ -LZ4_FORCE_INLINE int LZ4HC_literalsPrice(int const litlen) -{ - int price = litlen; - assert(litlen >= 0); - if (litlen >= (int)RUN_MASK) - price += 1 + ((litlen-(int)RUN_MASK) / 255); - return price; -} - - -/* requires mlen >= MINMATCH */ -LZ4_FORCE_INLINE int LZ4HC_sequencePrice(int litlen, int mlen) -{ - int price = 1 + 2 ; /* token + 16-bit offset */ - assert(litlen >= 0); - assert(mlen >= MINMATCH); - - price += LZ4HC_literalsPrice(litlen); - - if (mlen >= (int)(ML_MASK+MINMATCH)) - price += 1 + ((mlen-(int)(ML_MASK+MINMATCH)) / 255); - - return price; -} - - -typedef struct { - int off; - int len; -} LZ4HC_match_t; - -LZ4_FORCE_INLINE LZ4HC_match_t -LZ4HC_FindLongerMatch(LZ4HC_CCtx_internal* const ctx, - const BYTE* ip, const BYTE* const iHighLimit, - int minLen, int nbSearches, - const dictCtx_directive dict, - const HCfavor_e favorDecSpeed) -{ - LZ4HC_match_t match = { 0 , 0 }; - const BYTE* matchPtr = NULL; - /* note : LZ4HC_InsertAndGetWiderMatch() is able to modify the starting position of a match (*startpos), - * but this won't be the case here, as we define iLowLimit==ip, - * so LZ4HC_InsertAndGetWiderMatch() won't be allowed to search past ip */ - int matchLength = LZ4HC_InsertAndGetWiderMatch(ctx, ip, ip, iHighLimit, minLen, &matchPtr, &ip, nbSearches, 1 /*patternAnalysis*/, 1 /*chainSwap*/, dict, favorDecSpeed); - if (matchLength <= minLen) return match; - if (favorDecSpeed) { - if ((matchLength>18) & (matchLength<=36)) matchLength=18; /* favor shortcut */ - } - match.len = matchLength; - match.off = (int)(ip-matchPtr); - return match; -} - - -static int LZ4HC_compress_optimal ( LZ4HC_CCtx_internal* ctx, - const char* const source, - char* dst, - int* srcSizePtr, - int dstCapacity, - int const nbSearches, - size_t sufficient_len, - const limitedOutput_directive limit, - int const fullUpdate, - const dictCtx_directive dict, - const HCfavor_e favorDecSpeed) -{ + + +/* ================================================ + * LZ4 Optimal parser (levels [LZ4HC_CLEVEL_OPT_MIN - LZ4HC_CLEVEL_MAX]) + * ===============================================*/ +typedef struct { + int price; + int off; + int mlen; + int litlen; +} LZ4HC_optimal_t; + +/* price in bytes */ +LZ4_FORCE_INLINE int LZ4HC_literalsPrice(int const litlen) +{ + int price = litlen; + assert(litlen >= 0); + if (litlen >= (int)RUN_MASK) + price += 1 + ((litlen-(int)RUN_MASK) / 255); + return price; +} + + +/* requires mlen >= MINMATCH */ +LZ4_FORCE_INLINE int LZ4HC_sequencePrice(int litlen, int mlen) +{ + int price = 1 + 2 ; /* token + 16-bit offset */ + assert(litlen >= 0); + assert(mlen >= MINMATCH); + + price += LZ4HC_literalsPrice(litlen); + + if (mlen >= (int)(ML_MASK+MINMATCH)) + price += 1 + ((mlen-(int)(ML_MASK+MINMATCH)) / 255); + + return price; +} + + +typedef struct { + int off; + int len; +} LZ4HC_match_t; + +LZ4_FORCE_INLINE LZ4HC_match_t +LZ4HC_FindLongerMatch(LZ4HC_CCtx_internal* const ctx, + const BYTE* ip, const BYTE* const iHighLimit, + int minLen, int nbSearches, + const dictCtx_directive dict, + const HCfavor_e favorDecSpeed) +{ + LZ4HC_match_t match = { 0 , 0 }; + const BYTE* matchPtr = NULL; + /* note : LZ4HC_InsertAndGetWiderMatch() is able to modify the starting position of a match (*startpos), + * but this won't be the case here, as we define iLowLimit==ip, + * so LZ4HC_InsertAndGetWiderMatch() won't be allowed to search past ip */ + int matchLength = LZ4HC_InsertAndGetWiderMatch(ctx, ip, ip, iHighLimit, minLen, &matchPtr, &ip, nbSearches, 1 /*patternAnalysis*/, 1 /*chainSwap*/, dict, favorDecSpeed); + if (matchLength <= minLen) return match; + if (favorDecSpeed) { + if ((matchLength>18) & (matchLength<=36)) matchLength=18; /* favor shortcut */ + } + match.len = matchLength; + match.off = (int)(ip-matchPtr); + return match; +} + + +static int LZ4HC_compress_optimal ( LZ4HC_CCtx_internal* ctx, + const char* const source, + char* dst, + int* srcSizePtr, + int dstCapacity, + int const nbSearches, + size_t sufficient_len, + const limitedOutput_directive limit, + int const fullUpdate, + const dictCtx_directive dict, + const HCfavor_e favorDecSpeed) +{ int retval = 0; -#define TRAILING_LITERALS 3 +#define TRAILING_LITERALS 3 #ifdef LZ4HC_HEAPMODE LZ4HC_optimal_t* const opt = (LZ4HC_optimal_t*)ALLOC(sizeof(LZ4HC_optimal_t) * (LZ4_OPT_NUM + TRAILING_LITERALS)); #else - LZ4HC_optimal_t opt[LZ4_OPT_NUM + TRAILING_LITERALS]; /* ~64 KB, which is a bit large for stack... */ + LZ4HC_optimal_t opt[LZ4_OPT_NUM + TRAILING_LITERALS]; /* ~64 KB, which is a bit large for stack... */ #endif - - const BYTE* ip = (const BYTE*) source; - const BYTE* anchor = ip; - const BYTE* const iend = ip + *srcSizePtr; - const BYTE* const mflimit = iend - MFLIMIT; - const BYTE* const matchlimit = iend - LASTLITERALS; - BYTE* op = (BYTE*) dst; - BYTE* opSaved = (BYTE*) dst; - BYTE* oend = op + dstCapacity; + + const BYTE* ip = (const BYTE*) source; + const BYTE* anchor = ip; + const BYTE* const iend = ip + *srcSizePtr; + const BYTE* const mflimit = iend - MFLIMIT; + const BYTE* const matchlimit = iend - LASTLITERALS; + BYTE* op = (BYTE*) dst; + BYTE* opSaved = (BYTE*) dst; + BYTE* oend = op + dstCapacity; int ovml = MINMATCH; /* overflow - last sequence */ const BYTE* ovref = NULL; - - /* init */ + + /* init */ #ifdef LZ4HC_HEAPMODE if (opt == NULL) goto _return_label; #endif - DEBUGLOG(5, "LZ4HC_compress_optimal(dst=%p, dstCapa=%u)", dst, (unsigned)dstCapacity); - *srcSizePtr = 0; - if (limit == fillOutput) oend -= LASTLITERALS; /* Hack for support LZ4 format restriction */ - if (sufficient_len >= LZ4_OPT_NUM) sufficient_len = LZ4_OPT_NUM-1; - - /* Main Loop */ - while (ip <= mflimit) { - int const llen = (int)(ip - anchor); - int best_mlen, best_off; - int cur, last_match_pos = 0; - - LZ4HC_match_t const firstMatch = LZ4HC_FindLongerMatch(ctx, ip, matchlimit, MINMATCH-1, nbSearches, dict, favorDecSpeed); - if (firstMatch.len==0) { ip++; continue; } - - if ((size_t)firstMatch.len > sufficient_len) { - /* good enough solution : immediate encoding */ - int const firstML = firstMatch.len; - const BYTE* const matchPos = ip - firstMatch.off; - opSaved = op; + DEBUGLOG(5, "LZ4HC_compress_optimal(dst=%p, dstCapa=%u)", dst, (unsigned)dstCapacity); + *srcSizePtr = 0; + if (limit == fillOutput) oend -= LASTLITERALS; /* Hack for support LZ4 format restriction */ + if (sufficient_len >= LZ4_OPT_NUM) sufficient_len = LZ4_OPT_NUM-1; + + /* Main Loop */ + while (ip <= mflimit) { + int const llen = (int)(ip - anchor); + int best_mlen, best_off; + int cur, last_match_pos = 0; + + LZ4HC_match_t const firstMatch = LZ4HC_FindLongerMatch(ctx, ip, matchlimit, MINMATCH-1, nbSearches, dict, favorDecSpeed); + if (firstMatch.len==0) { ip++; continue; } + + if ((size_t)firstMatch.len > sufficient_len) { + /* good enough solution : immediate encoding */ + int const firstML = firstMatch.len; + const BYTE* const matchPos = ip - firstMatch.off; + opSaved = op; if ( LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), firstML, matchPos, limit, oend) ) { /* updates ip, op and anchor */ ovml = firstML; ovref = matchPos; - goto _dest_overflow; - } - continue; - } - - /* set prices for first positions (literals) */ - { int rPos; - for (rPos = 0 ; rPos < MINMATCH ; rPos++) { - int const cost = LZ4HC_literalsPrice(llen + rPos); - opt[rPos].mlen = 1; - opt[rPos].off = 0; - opt[rPos].litlen = llen + rPos; - opt[rPos].price = cost; - DEBUGLOG(7, "rPos:%3i => price:%3i (litlen=%i) -- initial setup", - rPos, cost, opt[rPos].litlen); - } } - /* set prices using initial match */ - { int mlen = MINMATCH; - int const matchML = firstMatch.len; /* necessarily < sufficient_len < LZ4_OPT_NUM */ - int const offset = firstMatch.off; - assert(matchML < LZ4_OPT_NUM); - for ( ; mlen <= matchML ; mlen++) { - int const cost = LZ4HC_sequencePrice(llen, mlen); - opt[mlen].mlen = mlen; - opt[mlen].off = offset; - opt[mlen].litlen = llen; - opt[mlen].price = cost; - DEBUGLOG(7, "rPos:%3i => price:%3i (matchlen=%i) -- initial setup", - mlen, cost, mlen); - } } - last_match_pos = firstMatch.len; - { int addLit; - for (addLit = 1; addLit <= TRAILING_LITERALS; addLit ++) { - opt[last_match_pos+addLit].mlen = 1; /* literal */ - opt[last_match_pos+addLit].off = 0; - opt[last_match_pos+addLit].litlen = addLit; - opt[last_match_pos+addLit].price = opt[last_match_pos].price + LZ4HC_literalsPrice(addLit); - DEBUGLOG(7, "rPos:%3i => price:%3i (litlen=%i) -- initial setup", - last_match_pos+addLit, opt[last_match_pos+addLit].price, addLit); - } } - - /* check further positions */ - for (cur = 1; cur < last_match_pos; cur++) { - const BYTE* const curPtr = ip + cur; - LZ4HC_match_t newMatch; - - if (curPtr > mflimit) break; - DEBUGLOG(7, "rPos:%u[%u] vs [%u]%u", - cur, opt[cur].price, opt[cur+1].price, cur+1); - if (fullUpdate) { - /* not useful to search here if next position has same (or lower) cost */ - if ( (opt[cur+1].price <= opt[cur].price) - /* in some cases, next position has same cost, but cost rises sharply after, so a small match would still be beneficial */ - && (opt[cur+MINMATCH].price < opt[cur].price + 3/*min seq price*/) ) - continue; - } else { - /* not useful to search here if next position has same (or lower) cost */ - if (opt[cur+1].price <= opt[cur].price) continue; + goto _dest_overflow; } - - DEBUGLOG(7, "search at rPos:%u", cur); - if (fullUpdate) - newMatch = LZ4HC_FindLongerMatch(ctx, curPtr, matchlimit, MINMATCH-1, nbSearches, dict, favorDecSpeed); - else - /* only test matches of minimum length; slightly faster, but misses a few bytes */ - newMatch = LZ4HC_FindLongerMatch(ctx, curPtr, matchlimit, last_match_pos - cur, nbSearches, dict, favorDecSpeed); - if (!newMatch.len) continue; - - if ( ((size_t)newMatch.len > sufficient_len) - || (newMatch.len + cur >= LZ4_OPT_NUM) ) { - /* immediate encoding */ - best_mlen = newMatch.len; - best_off = newMatch.off; - last_match_pos = cur + 1; - goto encode; - } - - /* before match : set price with literals at beginning */ - { int const baseLitlen = opt[cur].litlen; - int litlen; - for (litlen = 1; litlen < MINMATCH; litlen++) { - int const price = opt[cur].price - LZ4HC_literalsPrice(baseLitlen) + LZ4HC_literalsPrice(baseLitlen+litlen); - int const pos = cur + litlen; - if (price < opt[pos].price) { - opt[pos].mlen = 1; /* literal */ - opt[pos].off = 0; - opt[pos].litlen = baseLitlen+litlen; - opt[pos].price = price; - DEBUGLOG(7, "rPos:%3i => price:%3i (litlen=%i)", - pos, price, opt[pos].litlen); - } } } - - /* set prices using match at position = cur */ - { int const matchML = newMatch.len; - int ml = MINMATCH; - - assert(cur + newMatch.len < LZ4_OPT_NUM); - for ( ; ml <= matchML ; ml++) { - int const pos = cur + ml; - int const offset = newMatch.off; - int price; - int ll; - DEBUGLOG(7, "testing price rPos %i (last_match_pos=%i)", - pos, last_match_pos); - if (opt[cur].mlen == 1) { - ll = opt[cur].litlen; - price = ((cur > ll) ? opt[cur - ll].price : 0) - + LZ4HC_sequencePrice(ll, ml); - } else { - ll = 0; - price = opt[cur].price + LZ4HC_sequencePrice(0, ml); - } - - assert((U32)favorDecSpeed <= 1); - if (pos > last_match_pos+TRAILING_LITERALS - || price <= opt[pos].price - (int)favorDecSpeed) { - DEBUGLOG(7, "rPos:%3i => price:%3i (matchlen=%i)", - pos, price, ml); - assert(pos < LZ4_OPT_NUM); - if ( (ml == matchML) /* last pos of last match */ - && (last_match_pos < pos) ) - last_match_pos = pos; - opt[pos].mlen = ml; - opt[pos].off = offset; - opt[pos].litlen = ll; - opt[pos].price = price; - } } } - /* complete following positions with literals */ - { int addLit; - for (addLit = 1; addLit <= TRAILING_LITERALS; addLit ++) { - opt[last_match_pos+addLit].mlen = 1; /* literal */ - opt[last_match_pos+addLit].off = 0; - opt[last_match_pos+addLit].litlen = addLit; - opt[last_match_pos+addLit].price = opt[last_match_pos].price + LZ4HC_literalsPrice(addLit); - DEBUGLOG(7, "rPos:%3i => price:%3i (litlen=%i)", last_match_pos+addLit, opt[last_match_pos+addLit].price, addLit); - } } - } /* for (cur = 1; cur <= last_match_pos; cur++) */ - - assert(last_match_pos < LZ4_OPT_NUM + TRAILING_LITERALS); - best_mlen = opt[last_match_pos].mlen; - best_off = opt[last_match_pos].off; - cur = last_match_pos - best_mlen; - + continue; + } + + /* set prices for first positions (literals) */ + { int rPos; + for (rPos = 0 ; rPos < MINMATCH ; rPos++) { + int const cost = LZ4HC_literalsPrice(llen + rPos); + opt[rPos].mlen = 1; + opt[rPos].off = 0; + opt[rPos].litlen = llen + rPos; + opt[rPos].price = cost; + DEBUGLOG(7, "rPos:%3i => price:%3i (litlen=%i) -- initial setup", + rPos, cost, opt[rPos].litlen); + } } + /* set prices using initial match */ + { int mlen = MINMATCH; + int const matchML = firstMatch.len; /* necessarily < sufficient_len < LZ4_OPT_NUM */ + int const offset = firstMatch.off; + assert(matchML < LZ4_OPT_NUM); + for ( ; mlen <= matchML ; mlen++) { + int const cost = LZ4HC_sequencePrice(llen, mlen); + opt[mlen].mlen = mlen; + opt[mlen].off = offset; + opt[mlen].litlen = llen; + opt[mlen].price = cost; + DEBUGLOG(7, "rPos:%3i => price:%3i (matchlen=%i) -- initial setup", + mlen, cost, mlen); + } } + last_match_pos = firstMatch.len; + { int addLit; + for (addLit = 1; addLit <= TRAILING_LITERALS; addLit ++) { + opt[last_match_pos+addLit].mlen = 1; /* literal */ + opt[last_match_pos+addLit].off = 0; + opt[last_match_pos+addLit].litlen = addLit; + opt[last_match_pos+addLit].price = opt[last_match_pos].price + LZ4HC_literalsPrice(addLit); + DEBUGLOG(7, "rPos:%3i => price:%3i (litlen=%i) -- initial setup", + last_match_pos+addLit, opt[last_match_pos+addLit].price, addLit); + } } + + /* check further positions */ + for (cur = 1; cur < last_match_pos; cur++) { + const BYTE* const curPtr = ip + cur; + LZ4HC_match_t newMatch; + + if (curPtr > mflimit) break; + DEBUGLOG(7, "rPos:%u[%u] vs [%u]%u", + cur, opt[cur].price, opt[cur+1].price, cur+1); + if (fullUpdate) { + /* not useful to search here if next position has same (or lower) cost */ + if ( (opt[cur+1].price <= opt[cur].price) + /* in some cases, next position has same cost, but cost rises sharply after, so a small match would still be beneficial */ + && (opt[cur+MINMATCH].price < opt[cur].price + 3/*min seq price*/) ) + continue; + } else { + /* not useful to search here if next position has same (or lower) cost */ + if (opt[cur+1].price <= opt[cur].price) continue; + } + + DEBUGLOG(7, "search at rPos:%u", cur); + if (fullUpdate) + newMatch = LZ4HC_FindLongerMatch(ctx, curPtr, matchlimit, MINMATCH-1, nbSearches, dict, favorDecSpeed); + else + /* only test matches of minimum length; slightly faster, but misses a few bytes */ + newMatch = LZ4HC_FindLongerMatch(ctx, curPtr, matchlimit, last_match_pos - cur, nbSearches, dict, favorDecSpeed); + if (!newMatch.len) continue; + + if ( ((size_t)newMatch.len > sufficient_len) + || (newMatch.len + cur >= LZ4_OPT_NUM) ) { + /* immediate encoding */ + best_mlen = newMatch.len; + best_off = newMatch.off; + last_match_pos = cur + 1; + goto encode; + } + + /* before match : set price with literals at beginning */ + { int const baseLitlen = opt[cur].litlen; + int litlen; + for (litlen = 1; litlen < MINMATCH; litlen++) { + int const price = opt[cur].price - LZ4HC_literalsPrice(baseLitlen) + LZ4HC_literalsPrice(baseLitlen+litlen); + int const pos = cur + litlen; + if (price < opt[pos].price) { + opt[pos].mlen = 1; /* literal */ + opt[pos].off = 0; + opt[pos].litlen = baseLitlen+litlen; + opt[pos].price = price; + DEBUGLOG(7, "rPos:%3i => price:%3i (litlen=%i)", + pos, price, opt[pos].litlen); + } } } + + /* set prices using match at position = cur */ + { int const matchML = newMatch.len; + int ml = MINMATCH; + + assert(cur + newMatch.len < LZ4_OPT_NUM); + for ( ; ml <= matchML ; ml++) { + int const pos = cur + ml; + int const offset = newMatch.off; + int price; + int ll; + DEBUGLOG(7, "testing price rPos %i (last_match_pos=%i)", + pos, last_match_pos); + if (opt[cur].mlen == 1) { + ll = opt[cur].litlen; + price = ((cur > ll) ? opt[cur - ll].price : 0) + + LZ4HC_sequencePrice(ll, ml); + } else { + ll = 0; + price = opt[cur].price + LZ4HC_sequencePrice(0, ml); + } + + assert((U32)favorDecSpeed <= 1); + if (pos > last_match_pos+TRAILING_LITERALS + || price <= opt[pos].price - (int)favorDecSpeed) { + DEBUGLOG(7, "rPos:%3i => price:%3i (matchlen=%i)", + pos, price, ml); + assert(pos < LZ4_OPT_NUM); + if ( (ml == matchML) /* last pos of last match */ + && (last_match_pos < pos) ) + last_match_pos = pos; + opt[pos].mlen = ml; + opt[pos].off = offset; + opt[pos].litlen = ll; + opt[pos].price = price; + } } } + /* complete following positions with literals */ + { int addLit; + for (addLit = 1; addLit <= TRAILING_LITERALS; addLit ++) { + opt[last_match_pos+addLit].mlen = 1; /* literal */ + opt[last_match_pos+addLit].off = 0; + opt[last_match_pos+addLit].litlen = addLit; + opt[last_match_pos+addLit].price = opt[last_match_pos].price + LZ4HC_literalsPrice(addLit); + DEBUGLOG(7, "rPos:%3i => price:%3i (litlen=%i)", last_match_pos+addLit, opt[last_match_pos+addLit].price, addLit); + } } + } /* for (cur = 1; cur <= last_match_pos; cur++) */ + + assert(last_match_pos < LZ4_OPT_NUM + TRAILING_LITERALS); + best_mlen = opt[last_match_pos].mlen; + best_off = opt[last_match_pos].off; + cur = last_match_pos - best_mlen; + encode: /* cur, last_match_pos, best_mlen, best_off must be set */ - assert(cur < LZ4_OPT_NUM); - assert(last_match_pos >= 1); /* == 1 when only one candidate */ - DEBUGLOG(6, "reverse traversal, looking for shortest path (last_match_pos=%i)", last_match_pos); - { int candidate_pos = cur; - int selected_matchLength = best_mlen; - int selected_offset = best_off; - while (1) { /* from end to beginning */ - int const next_matchLength = opt[candidate_pos].mlen; /* can be 1, means literal */ - int const next_offset = opt[candidate_pos].off; - DEBUGLOG(7, "pos %i: sequence length %i", candidate_pos, selected_matchLength); - opt[candidate_pos].mlen = selected_matchLength; - opt[candidate_pos].off = selected_offset; - selected_matchLength = next_matchLength; - selected_offset = next_offset; - if (next_matchLength > candidate_pos) break; /* last match elected, first match to encode */ - assert(next_matchLength > 0); /* can be 1, means literal */ - candidate_pos -= next_matchLength; - } } - - /* encode all recorded sequences in order */ - { int rPos = 0; /* relative position (to ip) */ - while (rPos < last_match_pos) { - int const ml = opt[rPos].mlen; - int const offset = opt[rPos].off; - if (ml == 1) { ip++; rPos++; continue; } /* literal; note: can end up with several literals, in which case, skip them */ - rPos += ml; - assert(ml >= MINMATCH); - assert((offset >= 1) && (offset <= LZ4_DISTANCE_MAX)); - opSaved = op; + assert(cur < LZ4_OPT_NUM); + assert(last_match_pos >= 1); /* == 1 when only one candidate */ + DEBUGLOG(6, "reverse traversal, looking for shortest path (last_match_pos=%i)", last_match_pos); + { int candidate_pos = cur; + int selected_matchLength = best_mlen; + int selected_offset = best_off; + while (1) { /* from end to beginning */ + int const next_matchLength = opt[candidate_pos].mlen; /* can be 1, means literal */ + int const next_offset = opt[candidate_pos].off; + DEBUGLOG(7, "pos %i: sequence length %i", candidate_pos, selected_matchLength); + opt[candidate_pos].mlen = selected_matchLength; + opt[candidate_pos].off = selected_offset; + selected_matchLength = next_matchLength; + selected_offset = next_offset; + if (next_matchLength > candidate_pos) break; /* last match elected, first match to encode */ + assert(next_matchLength > 0); /* can be 1, means literal */ + candidate_pos -= next_matchLength; + } } + + /* encode all recorded sequences in order */ + { int rPos = 0; /* relative position (to ip) */ + while (rPos < last_match_pos) { + int const ml = opt[rPos].mlen; + int const offset = opt[rPos].off; + if (ml == 1) { ip++; rPos++; continue; } /* literal; note: can end up with several literals, in which case, skip them */ + rPos += ml; + assert(ml >= MINMATCH); + assert((offset >= 1) && (offset <= LZ4_DISTANCE_MAX)); + opSaved = op; if ( LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ip - offset, limit, oend) ) { /* updates ip, op and anchor */ ovml = ml; ovref = ip - offset; - goto _dest_overflow; + goto _dest_overflow; } } } - } /* while (ip <= mflimit) */ - + } /* while (ip <= mflimit) */ + _last_literals: - /* Encode Last Literals */ - { size_t lastRunSize = (size_t)(iend - anchor); /* literals */ + /* Encode Last Literals */ + { size_t lastRunSize = (size_t)(iend - anchor); /* literals */ size_t llAdd = (lastRunSize + 255 - RUN_MASK) / 255; size_t const totalSize = 1 + llAdd + lastRunSize; - if (limit == fillOutput) oend += LASTLITERALS; /* restore correct value */ - if (limit && (op + totalSize > oend)) { + if (limit == fillOutput) oend += LASTLITERALS; /* restore correct value */ + if (limit && (op + totalSize > oend)) { if (limit == limitedOutput) { /* Check output limit */ retval = 0; goto _return_label; } - /* adapt lastRunSize to fill 'dst' */ + /* adapt lastRunSize to fill 'dst' */ lastRunSize = (size_t)(oend - op) - 1 /*token*/; llAdd = (lastRunSize + 256 - RUN_MASK) / 256; lastRunSize -= llAdd; - } + } DEBUGLOG(6, "Final literal run : %i literals", (int)lastRunSize); ip = anchor + lastRunSize; /* can be != iend if limit==fillOutput */ - - if (lastRunSize >= RUN_MASK) { - size_t accumulator = lastRunSize - RUN_MASK; - *op++ = (RUN_MASK << ML_BITS); - for(; accumulator >= 255 ; accumulator -= 255) *op++ = 255; - *op++ = (BYTE) accumulator; - } else { - *op++ = (BYTE)(lastRunSize << ML_BITS); - } - memcpy(op, anchor, lastRunSize); - op += lastRunSize; - } - - /* End */ - *srcSizePtr = (int) (((const char*)ip) - source); + + if (lastRunSize >= RUN_MASK) { + size_t accumulator = lastRunSize - RUN_MASK; + *op++ = (RUN_MASK << ML_BITS); + for(; accumulator >= 255 ; accumulator -= 255) *op++ = 255; + *op++ = (BYTE) accumulator; + } else { + *op++ = (BYTE)(lastRunSize << ML_BITS); + } + memcpy(op, anchor, lastRunSize); + op += lastRunSize; + } + + /* End */ + *srcSizePtr = (int) (((const char*)ip) - source); retval = (int) ((char*)op-dst); goto _return_label; - + _dest_overflow: if (limit == fillOutput) { /* Assumption : ip, anchor, ovml and ovref must be set correctly */ |