// ------------------------------------------------------------------------------- private static unsafe byte[] PickleV0(byte *target, int targetLength, int sourceLength) { int diff = sourceLength - targetLength; int llen = diff == 0 ? 0 : diff < 0x100 ? 1 : diff < 0x10000 ? 2 : 4; byte[] result = new byte[targetLength + 1 + llen]; fixed(byte *resultPtr = result) { int llenFlags = llen == 4 ? 3 : llen; // 2 bits byte flags = (byte)((llenFlags << 6) | CurrentVersion); LZ4MemoryHelper.Poke8(resultPtr + 0, flags); if (llen == 1) { LZ4MemoryHelper.Poke8(resultPtr + 1, (byte)diff); } else if (llen == 2) { LZ4MemoryHelper.Poke16(resultPtr + 1, (ushort)diff); } else if (llen == 4) { LZ4MemoryHelper.Poke32(resultPtr + 1, (uint)diff); } LZ4MemoryHelper.Move(resultPtr + llen + 1, target, targetLength); } return(result); }
private static uint ReverseCountPattern(byte *ip, byte *iLow, uint pattern) { byte *iStart = ip; while (ip >= iLow + 4) { if (LZ4MemoryHelper.Peek32(ip - 4) != pattern) { break; } ip -= 4; } { // works for any endianess byte *bytePtr = (byte *)&pattern + 3; while (ip > iLow) { if (ip[-1] != *bytePtr) { break; } ip--; bytePtr--; } } return((uint)(iStart - ip)); }
public static int SaveDictHC(CCtxT *streamHCPtr, byte *safeBuffer, int dictSize) { CCtxT *streamPtr = streamHCPtr; int prefixSize = (int)(streamPtr->End - (streamPtr->BaseP + streamPtr->DictLimit)); if (dictSize > 64 * KB) { dictSize = 64 * KB; } if (dictSize < 4) { dictSize = 0; } if (dictSize > prefixSize) { dictSize = prefixSize; } LZ4MemoryHelper.Move(safeBuffer, streamPtr->End - dictSize, dictSize); uint endIndex = (uint)(streamPtr->End - streamPtr->BaseP); streamPtr->End = safeBuffer + dictSize; streamPtr->BaseP = streamPtr->End - endIndex; streamPtr->DictLimit = endIndex - (uint)dictSize; streamPtr->LowLimit = endIndex - (uint)dictSize; if (streamPtr->NextToUpdate < streamPtr->DictLimit) { streamPtr->NextToUpdate = streamPtr->DictLimit; } return(dictSize); }
private static uint HashPosition(void *p, TableTypeT tableType) { #if !BIT32 if (tableType != TableTypeT.ByU16) { return(Hash5(ReadArch(p), tableType)); } #endif return(Hash4(LZ4MemoryHelper.Peek32(p), tableType)); }
public static void InitializeHC(CCtxT *hc4, byte *start) { LZ4MemoryHelper.Zero((byte *)hc4->HashTable, HashTableSizeHC * sizeof(uint)); LZ4MemoryHelper.Fill((byte *)hc4->ChainTable, 0xFF, MaxD * sizeof(ushort)); hc4->NextToUpdate = 64 * KB; hc4->BaseP = start - 64 * KB; hc4->End = start; hc4->DictBase = start - 64 * KB; hc4->DictLimit = 64 * KB; hc4->LowLimit = 64 * KB; }
protected static uint Count(byte *pIn, byte *pMatch, byte *pInLimit) { byte *pStart = pIn; if (pIn < pInLimit - (StepSize - 1)) { ulong diff = ReadArch(pMatch) ^ ReadArch(pIn); if (diff != 0) { return(NbCommonBytes(diff)); } pIn += StepSize; pMatch += StepSize; } while (pIn < pInLimit - (StepSize - 1)) { ulong diff = ReadArch(pMatch) ^ ReadArch(pIn); if (diff != 0) { return((uint)(pIn + NbCommonBytes(diff) - pStart)); } pIn += StepSize; pMatch += StepSize; } #if !BIT32 if ((pIn < pInLimit - 3) && (LZ4MemoryHelper.Peek32(pMatch) == LZ4MemoryHelper.Peek32(pIn))) { pIn += 4; pMatch += 4; } #endif if ((pIn < pInLimit - 1) && (LZ4MemoryHelper.Peek16(pMatch) == LZ4MemoryHelper.Peek16(pIn))) { pIn += 2; pMatch += 2; } if ((pIn < pInLimit) && (*pMatch == *pIn)) { pIn++; } return((uint)(pIn - pStart)); }
private static MatchT FindLongerMatch(CCtxT *ctx, byte *ip, byte *iHighLimit, int minLen, int nbSearches) { MatchT match; LZ4MemoryHelper.Zero((byte *)&match, sizeof(MatchT)); byte *matchPtr = null; int matchLength = InsertAndGetWiderMatch(ctx, ip, ip, iHighLimit, minLen, &matchPtr, &ip, nbSearches, 1); if (matchLength <= minLen) { return(match); } match.Len = matchLength; match.Off = (int)(ip - matchPtr); return(match); }
public static int SaveDict(StreamT *dict, byte *safeBuffer, int dictSize) { byte *previousDictEnd = dict->Dictionary + dict->DictSize; if ((uint)dictSize > 64 * KB) { dictSize = 64 * KB; } if ((uint)dictSize > dict->DictSize) { dictSize = (int)dict->DictSize; } LZ4MemoryHelper.Move(safeBuffer, previousDictEnd - dictSize, dictSize); dict->Dictionary = safeBuffer; dict->DictSize = (uint)dictSize; return(dictSize); }
public static unsafe byte[] Compress(byte *source, int sourceLength, LZ4CompressionLevel level = LZ4CompressionLevel.Level0) { if (sourceLength <= 0) { return(Array.Empty <byte>()); } int targetLength = sourceLength - 1; byte *target = (byte *)LZ4MemoryHelper.Alloc(sourceLength); try { int encodedLength = Encode(source, sourceLength, target, targetLength, level); return(encodedLength <= 0 ? PickleV0(source, sourceLength, sourceLength) : PickleV0(target, encodedLength, sourceLength)); } finally { LZ4MemoryHelper.Free(target); } }
public static int DecompressGeneric(byte *src, byte *dst, int srcSize, int outputSize, EndConditionDirective endOnInput, EarlyEndDirective partialDecoding, int targetOutputSize, DictDirective dict, byte *lowPrefix, byte *dictStart, int dictSize) { byte *ip = src; byte *iend = ip + srcSize; byte *op = dst; byte *oend = op + outputSize; byte *oexit = op + targetOutputSize; byte *dictEnd = dictStart + dictSize; bool safeDecode = endOnInput == EndConditionDirective.EndOnInputSize; bool checkOffset = safeDecode && dictSize < 64 * KB; if ((partialDecoding != EarlyEndDirective.Full) && (oexit > oend - MFLimit)) { oexit = oend - MFLimit; } if ((endOnInput == EndConditionDirective.EndOnInputSize) && (outputSize == 0)) { return(srcSize == 1 && *ip == 0 ? 0 : -1); } if ((endOnInput != EndConditionDirective.EndOnInputSize) && (outputSize == 0)) { return(*ip == 0 ? 1 : -1); } for (; ;) { int length; uint token = *ip++; if ((ip + 14 + 2 <= iend) && (op + 14 + 18 <= oend) && (token < 15 << MLBits) && ((token & MLMask) != 15)) { int ll = (int)(token >> MLBits); int off = LZ4MemoryHelper.Peek16(ip + ll); byte *matchPtr = op + ll - off; if (off >= 18 && matchPtr >= lowPrefix) { int ml = (int)((token & MLMask) + MinMatch); LZ4MemoryHelper.Copy16(op, ip); op += ll; ip += ll + 2; LZ4MemoryHelper.Copy18(op, matchPtr); op += ml; continue; } } if ((length = (int)(token >> MLBits)) == RunMask) { uint s; do { s = *ip++; length += (int)s; }while ( (endOnInput != EndConditionDirective.EndOnInputSize || ip < iend - RunMask) && (s == 255)); if (safeDecode && op + length < op) { goto _output_error; } if (safeDecode && ip + length < ip) { goto _output_error; } } byte *cpy = op + length; if ((endOnInput == EndConditionDirective.EndOnInputSize) && ((cpy > (partialDecoding == EarlyEndDirective.Partial ? oexit : oend - MFLimit)) || (ip + length > iend - (2 + 1 + LastLiterals))) || (endOnInput != EndConditionDirective.EndOnInputSize) && (cpy > oend - WildCopyLength)) { if (partialDecoding == EarlyEndDirective.Partial) { if (cpy > oend) { goto _output_error; } if ((endOnInput == EndConditionDirective.EndOnInputSize) && (ip + length > iend)) { goto _output_error; } } else { if ((endOnInput != EndConditionDirective.EndOnInputSize) && (cpy != oend)) { goto _output_error; } if ((endOnInput == EndConditionDirective.EndOnInputSize) && (ip + length != iend || cpy > oend)) { goto _output_error; } } LZ4MemoryHelper.Copy(op, ip, length); ip += length; op += length; break; } LZ4MemoryHelper.WildCopy(op, ip, cpy); ip += length; op = cpy; int offset = LZ4MemoryHelper.Peek16(ip); ip += 2; byte *match = op - offset; if (checkOffset && match + dictSize < lowPrefix) { goto _output_error; } LZ4MemoryHelper.Poke32(op, (uint)offset); length = (int)(token & MLMask); if (length == MLMask) { uint s; do { s = *ip++; if ((endOnInput == EndConditionDirective.EndOnInputSize) && (ip > iend - LastLiterals)) { goto _output_error; } length += (int)s; }while (s == 255); if (safeDecode && (op + length < op)) { goto _output_error; } } length += MinMatch; if ((dict == DictDirective.UsingExtDict) && (match < lowPrefix)) { if (op + length > oend - LastLiterals) { goto _output_error; } if (length <= lowPrefix - match) { LZ4MemoryHelper.Move(op, dictEnd - (lowPrefix - match), length); op += length; } else { int copySize = (int)(lowPrefix - match); int restSize = length - copySize; LZ4MemoryHelper.Copy(op, dictEnd - copySize, copySize); op += copySize; if (restSize > (int)(op - lowPrefix)) { byte *endOfMatch = op + restSize; byte *copyFrom = lowPrefix; while (op < endOfMatch) { *op++ = *copyFrom++; } } else { LZ4MemoryHelper.Copy(op, lowPrefix, restSize); op += restSize; } } continue; } cpy = op + length; if (offset < 8) { op[0] = match[0]; op[1] = match[1]; op[2] = match[2]; op[3] = match[3]; match += _inc32table[offset]; LZ4MemoryHelper.Copy(op + 4, match, 4); match -= _dec64table[offset]; } else { LZ4MemoryHelper.Copy8(op, match); match += 8; } op += 8; if (cpy > oend - 12) { byte *oCopyLimit = oend - (WildCopyLength - 1); if (cpy > oend - LastLiterals) { goto _output_error; } if (op < oCopyLimit) { LZ4MemoryHelper.WildCopy(op, match, oCopyLimit); match += oCopyLimit - op; op = oCopyLimit; } while (op < cpy) { *op++ = *match++; } } else { LZ4MemoryHelper.Copy8(op, match); if (length > 16) { LZ4MemoryHelper.WildCopy(op + 8, match + 8, cpy); } } op = cpy; // correction } // end of decoding if (endOnInput == EndConditionDirective.EndOnInputSize) { return((int)(op - dst)); // Nb of output bytes decoded } return((int)(ip - src)); // Nb of input bytes read // Overflow error detected _output_error: return((int)-(ip - src) - 1); }
private static void FreeCtx(CCtxT *context) { LZ4MemoryHelper.Free(context); }
private static int CompressDestSizeGeneric(StreamT *ctx, byte *src, byte *dst, int *srcSizePtr, int targetDstSize, TableTypeT tableType) { byte *ip = src; byte *srcBase = src; byte *lowLimit = src; byte *anchor = ip; byte *iend = ip + *srcSizePtr; byte *mflimit = iend - MFLimit; byte *matchlimit = iend - LastLiterals; byte *op = dst; byte *oend = op + targetDstSize; byte *oMaxLit = op + targetDstSize - 2 - 8 - 1; byte *oMaxMatch = op + targetDstSize - (LastLiterals + 1); byte *oMaxSeq = oMaxLit - 1; if (targetDstSize < 1) { return(0); } if (*srcSizePtr > MaxInputSize) { return(0); } if ((tableType == TableTypeT.ByU16) && (*srcSizePtr >= Limit64k)) { return(0); } if (*srcSizePtr < LZ4MinLength) { goto _last_literals; // Input too small, no compression (all literals) } *srcSizePtr = 0; PutPosition(ip, ctx->HashTable, tableType, srcBase); ip++; uint forwardH = HashPosition(ip, tableType); for (; ;) { byte *match; byte *token; { byte *forwardIp = ip; uint step = 1u; uint searchMatchNb = 1u << SkipTrigger; do { uint h = forwardH; ip = forwardIp; forwardIp += step; step = searchMatchNb++ >> SkipTrigger; if (forwardIp > mflimit) { goto _last_literals; } match = GetPositionOnHash(h, ctx->HashTable, tableType, srcBase); forwardH = HashPosition(forwardIp, tableType); PutPositionOnHash(ip, h, ctx->HashTable, tableType, srcBase); }while ((tableType != TableTypeT.ByU16) && (match + MaxDistance < ip) || (LZ4MemoryHelper.Peek32(match) != LZ4MemoryHelper.Peek32(ip))); } while ((ip > anchor) && (match > lowLimit) && (ip[-1] == match[-1])) { ip--; match--; } { uint litLength = (uint)(ip - anchor); token = op++; if (op + (litLength + 240) / 255 + litLength > oMaxLit) { op--; goto _last_literals; } if (litLength >= RunMask) { uint len = litLength - RunMask; *token = (byte)(RunMask << MLBits); for (; len >= 255; len -= 255) { *op++ = 255; } *op++ = (byte)len; } else { *token = (byte)(litLength << MLBits); } LZ4MemoryHelper.WildCopy(op, anchor, op + litLength); op += litLength; } _next_match: LZ4MemoryHelper.Poke16(op, (ushort)(ip - match)); op += 2; { int matchLength = (int)Count(ip + MinMatch, match + MinMatch, matchlimit); if (op + (matchLength + 240) / 255 > oMaxMatch) { matchLength = (int)(15 - 1 + (oMaxMatch - op) * 255); } ip += MinMatch + matchLength; if (matchLength >= MLMask) { *token += (byte)MLMask; matchLength -= (int)MLMask; while (matchLength >= 255) { matchLength -= 255; *op++ = 255; } *op++ = (byte)matchLength; } else { *token += (byte)matchLength; } } anchor = ip; if (ip > mflimit) { break; } if (op > oMaxSeq) { break; } PutPosition(ip - 2, ctx->HashTable, tableType, srcBase); match = GetPosition(ip, ctx->HashTable, tableType, srcBase); PutPosition(ip, ctx->HashTable, tableType, srcBase); if ((match + MaxDistance >= ip) && (LZ4MemoryHelper.Peek32(match) == LZ4MemoryHelper.Peek32(ip))) { token = op++; *token = 0; goto _next_match; } forwardH = HashPosition(++ip, tableType); } _last_literals: { int lastRunSize = (int)(iend - anchor); if (op + 1 + (lastRunSize + 240) / 255 + lastRunSize > oend) { lastRunSize = (int)(oend - op) - 1; lastRunSize -= (lastRunSize + 240) / 255; } ip = anchor + lastRunSize; if (lastRunSize >= RunMask) { long accumulator = lastRunSize - RunMask; *op++ = (byte)(RunMask << MLBits); for (; accumulator >= 255; accumulator -= 255) { *op++ = 255; } *op++ = (byte)accumulator; } else { *op++ = (byte)(lastRunSize << MLBits); } LZ4MemoryHelper.Copy(op, anchor, lastRunSize); op += lastRunSize; } *srcSizePtr = (int)(ip - src); return((int)(op - dst)); }
private static int EncodeSequence(byte **ip, byte **op, byte **anchor, int matchLength, byte *match, LimitedOutputDirective limit, byte *oend) { byte *token = (*op)++; size_t length = (size_t)(*ip - *anchor); if ((limit != LimitedOutputDirective.NoLimit) && (*op + (length >> 8) + length + (2 + 1 + LastLiterals) > oend)) { return(1); } if (length >= RunMask) { size_t len = length - RunMask; * token = (byte)(RunMask << MLBits); for (; len >= 255; len -= 255) { *(*op)++ = 255; } *(*op)++ = (byte)len; } else { *token = (byte)(length << MLBits); } LZ4MemoryHelper.WildCopy(*op, *anchor, (*op) + length); *op += length; LZ4MemoryHelper.Poke16(*op, (ushort)(*ip - match)); *op += 2; length = (size_t)(matchLength - MinMatch); if ((limit != LimitedOutputDirective.NoLimit) && (*op + (length >> 8) + (1 + LastLiterals) > oend)) { return(1); } if (length >= MLMask) { *token += (byte)MLMask; length -= MLMask; for (; length >= 510; length -= 510) { *(*op)++ = 255; *(*op)++ = 255; } if (length >= 255) { length -= 255; *(*op)++ = 255; } *(*op)++ = (byte)length; } else { *token += (byte)(length); } *ip += matchLength; *anchor = *ip; return(0); }
private static int InsertAndGetWiderMatch(CCtxT *hc4, byte *ip, byte *iLowLimit, byte *iHighLimit, int longest, byte **matchpos, byte **startpos, int maxNbAttempts, int patternAnalysis) { ushort *chainTable = hc4->ChainTable; uint * hashTable = hc4->HashTable; byte * basep = hc4->BaseP; uint dictLimit = hc4->DictLimit; byte * lowPrefixPtr = basep + dictLimit; uint lowLimit = hc4->LowLimit + 64 * KB > (uint)(ip - basep) ? hc4->LowLimit : (uint)(ip - basep) - MaxDistance; byte * dictBase = hc4->DictBase; int delta = (int)(ip - iLowLimit); int nbAttempts = maxNbAttempts; uint pattern = LZ4MemoryHelper.Peek32(ip); RepeatStateE repeat = RepeatStateE.RepUntested; int srcPatternLength = 0; // First Match Insert(hc4, ip); uint matchIndex = hashTable[HashPtr(ip)]; while ((matchIndex >= lowLimit) && (nbAttempts != 0)) { nbAttempts--; if (matchIndex >= dictLimit) { byte *matchPtr = basep + matchIndex; if (*(iLowLimit + longest) == *(matchPtr - delta + longest)) { if (LZ4MemoryHelper.Peek32(matchPtr) == pattern) { int mlt = MinMatch + (int)Count(ip + MinMatch, matchPtr + MinMatch, iHighLimit); int back = 0; while ((ip + back > iLowLimit) && (matchPtr + back > lowPrefixPtr) && (ip[back - 1] == matchPtr[back - 1])) { back--; } mlt -= back; if (mlt > longest) { longest = mlt; *matchpos = matchPtr + back; *startpos = ip + back; } } } } else { byte *matchPtr = dictBase + matchIndex; if (LZ4MemoryHelper.Peek32(matchPtr) == pattern) { int back = 0; byte *vLimit = ip + (dictLimit - matchIndex); if (vLimit > iHighLimit) { vLimit = iHighLimit; } int mlt = MinMatch + (int)Count(ip + MinMatch, matchPtr + MinMatch, vLimit); if ((ip + mlt == vLimit) && (vLimit < iHighLimit)) { mlt += (int)Count(ip + mlt, basep + dictLimit, iHighLimit); } while ((ip + back > iLowLimit) && (matchIndex + back > lowLimit) && (ip[back - 1] == matchPtr[back - 1])) { back--; } mlt -= back; if (mlt > longest) { longest = mlt; *matchpos = basep + matchIndex + back; *startpos = ip + back; } } } { ushort nextOffset = DeltaNextU16(chainTable, (ushort)matchIndex); matchIndex -= nextOffset; if ((patternAnalysis != 0) && (nextOffset == 1)) { // may be a repeated pattern if (repeat == RepeatStateE.RepUntested) { if (((pattern & 0xFFFF) == pattern >> 16) & ((pattern & 0xFF) == pattern >> 24)) { repeat = RepeatStateE.RepConfirmed; srcPatternLength = (int)CountPattern(ip + 4, iHighLimit, pattern) + 4; } else { repeat = RepeatStateE.RepNot; } } if ((repeat == RepeatStateE.RepConfirmed) && (matchIndex >= dictLimit)) { byte *matchPtr = basep + matchIndex; if (LZ4MemoryHelper.Peek32(matchPtr) == pattern) { int forwardPatternLength = (int)CountPattern(matchPtr + sizeof(uint), iHighLimit, pattern) + sizeof(uint); byte *maxLowPtr = lowPrefixPtr + MaxDistance >= ip ? lowPrefixPtr : ip - MaxDistance; int backLength = (int)ReverseCountPattern(matchPtr, maxLowPtr, pattern); int currentSegmentLength = backLength + forwardPatternLength; if ((currentSegmentLength >= srcPatternLength) && (forwardPatternLength <= srcPatternLength)) { matchIndex += (uint)(forwardPatternLength - srcPatternLength); } else { matchIndex -= (uint)backLength; } } } } } } return(longest); }
private static uint HashPtr(void *ptr) { return((LZ4MemoryHelper.Peek32(ptr) * 2654435761U) >> (MinMatch * 8 - HashLogHC)); }
public static int CompressGeneric(StreamT *cctx, byte *source, byte *dest, int inputSize, int maxOutputSize, LimitedOutputDirective outputLimited, TableTypeT tableType, DictDirective dict, DictIssueDirective dictIssue, uint acceleration) { byte *ip = source; byte *ibase; byte *lowLimit; byte *lowRefLimit = ip - cctx->DictSize; byte *dictionary = cctx->Dictionary; byte *dictEnd = dictionary + cctx->DictSize; long dictDelta = dictEnd - source; byte *anchor = source; byte *iend = ip + inputSize; byte *mflimit = iend - MFLimit; byte *matchlimit = iend - LastLiterals; byte *op = dest; byte *olimit = op + maxOutputSize; if (inputSize > MaxInputSize) { return(0); } switch (dict) { case DictDirective.WithPrefix64k: ibase = source - cctx->CurrentOffset; lowLimit = source - cctx->DictSize; break; case DictDirective.UsingExtDict: ibase = source - cctx->CurrentOffset; lowLimit = source; break; default: ibase = source; lowLimit = source; break; } if ((tableType == TableTypeT.ByU16) && (inputSize >= Limit64k)) { return(0); } if (inputSize < LZ4MinLength) { goto _last_literals; } PutPosition(ip, cctx->HashTable, tableType, ibase); ip++; uint forwardH = HashPosition(ip, tableType); for (; ;) { long refDelta = 0L; byte *match; byte *token; { byte *forwardIp = ip; uint step = 1u; uint searchMatchNb = acceleration << SkipTrigger; do { uint h = forwardH; ip = forwardIp; forwardIp += step; step = searchMatchNb++ >> SkipTrigger; if (forwardIp > mflimit) { goto _last_literals; } match = GetPositionOnHash(h, cctx->HashTable, tableType, ibase); if (dict == DictDirective.UsingExtDict) { if (match < source) { refDelta = dictDelta; lowLimit = dictionary; } else { refDelta = 0; lowLimit = source; } } forwardH = HashPosition(forwardIp, tableType); PutPositionOnHash(ip, h, cctx->HashTable, tableType, ibase); }while ((dictIssue == DictIssueDirective.DictSmall) && (match < lowRefLimit) || (tableType != TableTypeT.ByU16) && (match + MaxDistance < ip) || (LZ4MemoryHelper.Peek32(match + refDelta) != LZ4MemoryHelper.Peek32(ip))); } while ((ip > anchor) && (match + refDelta > lowLimit) && (ip[-1] == match[refDelta - 1])) { ip--; match--; } { uint litLength = (uint)(ip - anchor); token = op++; if ((outputLimited == LimitedOutputDirective.LimitedOutput) && (op + litLength + (2 + 1 + LastLiterals) + litLength / 255 > olimit)) { return(0); } if (litLength >= RunMask) { int len = (int)(litLength - RunMask); *token = (byte)(RunMask << MLBits); for (; len >= 255; len -= 255) { *op++ = 255; } *op++ = (byte)len; } else { *token = (byte)(litLength << MLBits); } LZ4MemoryHelper.WildCopy(op, anchor, op + litLength); op += litLength; } _next_match: LZ4MemoryHelper.Poke16(op, (ushort)(ip - match)); op += 2; { uint matchCode; if ((dict == DictDirective.UsingExtDict) && (lowLimit == dictionary)) { match += refDelta; byte *limit = ip + (dictEnd - match); if (limit > matchlimit) { limit = matchlimit; } matchCode = Count(ip + MinMatch, match + MinMatch, limit); ip += MinMatch + matchCode; if (ip == limit) { uint more = Count(ip, source, matchlimit); matchCode += more; ip += more; } } else { matchCode = Count(ip + MinMatch, match + MinMatch, matchlimit); ip += MinMatch + matchCode; } if ((outputLimited == LimitedOutputDirective.LimitedOutput) && (op + (1 + LastLiterals) + (matchCode >> 8) > olimit)) { return(0); } if (matchCode >= MLMask) { *token += (byte)MLMask; matchCode -= MLMask; LZ4MemoryHelper.Poke32(op, 0xFFFFFFFF); while (matchCode >= 4 * 255) { op += 4; LZ4MemoryHelper.Poke32(op, 0xFFFFFFFF); matchCode -= 4 * 255; } op += matchCode / 255; *op++ = (byte)(matchCode % 255); } else { *token += (byte)matchCode; } } anchor = ip; if (ip > mflimit) { break; } PutPosition(ip - 2, cctx->HashTable, tableType, ibase); match = GetPosition(ip, cctx->HashTable, tableType, ibase); if (dict == DictDirective.UsingExtDict) { if (match < source) { refDelta = dictDelta; lowLimit = dictionary; } else { refDelta = 0; lowLimit = source; } } PutPosition(ip, cctx->HashTable, tableType, ibase); if ((dictIssue != DictIssueDirective.DictSmall || match >= lowRefLimit) && (match + MaxDistance >= ip) && (LZ4MemoryHelper.Peek32(match + refDelta) == LZ4MemoryHelper.Peek32(ip))) { token = op++; *token = 0; goto _next_match; } forwardH = HashPosition(++ip, tableType); } _last_literals: { int lastRun = (int)(iend - anchor); if ((outputLimited == LimitedOutputDirective.LimitedOutput) && (op - dest + lastRun + 1 + (lastRun + 255 - RunMask) / 255 > (uint)maxOutputSize)) { return(0); } if (lastRun >= RunMask) { int accumulator = (int)(lastRun - RunMask); *op++ = (byte)(RunMask << MLBits); for (; accumulator >= 255; accumulator -= 255) { *op++ = 255; } *op++ = (byte)accumulator; } else { *op++ = (byte)(lastRun << MLBits); } LZ4MemoryHelper.Copy(op, anchor, lastRun); op += lastRun; } return((int)(op - dest)); }
public static void ResetStream(StreamT *state) { LZ4MemoryHelper.Zero((byte *)state, sizeof(StreamT)); }
// ----------------------------------------------------------------- private static int CompressOptimal(CCtxT *ctx, byte *source, byte *dst, int *srcSizePtr, int dstCapacity, int nbSearches, size_t sufficientLen, LimitedOutputDirective limit, int fullUpdate) { const int TrailingLiterals = 3; OptimalT *opt = stackalloc OptimalT[OptNum + TrailingLiterals]; byte *ip = source; byte *anchor = ip; byte *iend = ip + *srcSizePtr; byte *mflimit = iend - MFLimit; byte *matchlimit = iend - LastLiterals; byte *op = dst; byte *opSaved; byte *oend = op + dstCapacity; *srcSizePtr = 0; if (limit == LimitedOutputDirective.LimitedDestSize) { oend -= LastLiterals; } if (sufficientLen >= OptNum) { sufficientLen = OptNum - 1; } // Main Loop while (ip < mflimit) { int llen = (int)(ip - anchor); int bestMlen, bestOff; int cur, lastMatchPos; MatchT firstMatch = FindLongerMatch(ctx, ip, matchlimit, MinMatch - 1, nbSearches); if (firstMatch.Len == 0) { ip++; continue; } if ((size_t)firstMatch.Len > sufficientLen) { int firstML = firstMatch.Len; byte *matchPos = ip - firstMatch.Off; opSaved = op; if (EncodeSequence(&ip, &op, &anchor, firstML, matchPos, limit, oend) != 0) { goto _dest_overflow; } continue; } // set prices for first positions (literals) { int rPos; for (rPos = 0; rPos < MinMatch; rPos++) { int cost = LiteralsPrice(llen + rPos); opt[rPos].Mlen = 1; opt[rPos].Off = 0; opt[rPos].Litlen = llen + rPos; opt[rPos].Price = cost; } } // set prices using initial match { int mlen = MinMatch; int matchML = firstMatch.Len; // necessarily < sufficient_len < LZ4_OPT_NUM int offset = firstMatch.Off; for (; mlen <= matchML; mlen++) { int cost = SequencePrice(llen, mlen); opt[mlen].Mlen = mlen; opt[mlen].Off = offset; opt[mlen].Litlen = llen; opt[mlen].Price = cost; } } lastMatchPos = firstMatch.Len; { int addLit; for (addLit = 1; addLit <= TrailingLiterals; addLit++) { opt[lastMatchPos + addLit].Mlen = 1; opt[lastMatchPos + addLit].Off = 0; opt[lastMatchPos + addLit].Litlen = addLit; opt[lastMatchPos + addLit].Price = opt[lastMatchPos].Price + LiteralsPrice(addLit); } } // check further positions for (cur = 1; cur < lastMatchPos; cur++) { byte *curPtr = ip + cur; if (curPtr >= mflimit) { break; } if (fullUpdate != 0) { if ((opt[cur + 1].Price <= opt[cur].Price) && (opt[cur + MinMatch].Price < opt[cur].Price + 3)) { continue; } } else { if (opt[cur + 1].Price <= opt[cur].Price) { continue; } } MatchT newMatch = FindLongerMatch( ctx, curPtr, matchlimit, fullUpdate != 0 ? MinMatch - 1 : lastMatchPos - cur, nbSearches); if (newMatch.Len == 0) { continue; } if ((size_t)newMatch.Len > sufficientLen || newMatch.Len + cur >= OptNum) { // immediate encoding bestMlen = newMatch.Len; bestOff = newMatch.Off; lastMatchPos = cur + 1; goto encode; } // before match : set price with literals at beginning { int baseLitlen = opt[cur].Litlen; for (int litlen = 1; litlen < MinMatch; litlen++) { int price = opt[cur].Price - LiteralsPrice(baseLitlen) + LiteralsPrice(baseLitlen + litlen); int pos = cur + litlen; if (price >= opt[pos].Price) { continue; } opt[pos].Mlen = 1; opt[pos].Off = 0; opt[pos].Litlen = baseLitlen + litlen; opt[pos].Price = price; } } // set prices using match at position = cur { int matchML = newMatch.Len; int ml = MinMatch; for (; ml <= matchML; ml++) { int pos = cur + ml; int offset = newMatch.Off; int price; int ll; if (opt[cur].Mlen == 1) { ll = opt[cur].Litlen; price = ((cur > ll) ? opt[cur - ll].Price : 0) + SequencePrice(ll, ml); } else { ll = 0; price = opt[cur].Price + SequencePrice(0, ml); } if ((pos > lastMatchPos + TrailingLiterals) || (price <= opt[pos].Price)) { if (ml == matchML && lastMatchPos < pos) { lastMatchPos = pos; } opt[pos].Mlen = ml; opt[pos].Off = offset; opt[pos].Litlen = ll; opt[pos].Price = price; } } } { int addLit; for (addLit = 1; addLit <= TrailingLiterals; addLit++) { opt[lastMatchPos + addLit].Mlen = 1; opt[lastMatchPos + addLit].Off = 0; opt[lastMatchPos + addLit].Litlen = addLit; opt[lastMatchPos + addLit].Price = opt[lastMatchPos].Price + LiteralsPrice(addLit); } } } bestMlen = opt[lastMatchPos].Mlen; bestOff = opt[lastMatchPos].Off; cur = lastMatchPos - bestMlen; encode: // cur, last_match_pos, best_mlen, best_off must be set { int candidate_pos = cur; int selected_matchLength = bestMlen; int selected_offset = bestOff; while (true) { int next_matchLength = opt[candidate_pos].Mlen; int next_offset = opt[candidate_pos].Off; 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; } candidate_pos -= next_matchLength; } } { int rPos = 0; while (rPos < lastMatchPos) { int ml = opt[rPos].Mlen; int offset = opt[rPos].Off; if (ml == 1) { ip++; rPos++; continue; } rPos += ml; opSaved = op; if (EncodeSequence(&ip, &op, &anchor, ml, ip - offset, limit, oend) != 0) { goto _dest_overflow; } } } } _last_literals: { size_t lastRunSize = (size_t)(iend - anchor); size_t litLength = (lastRunSize + 255 - RunMask) / 255; size_t totalSize = 1 + litLength + lastRunSize; if (limit == LimitedOutputDirective.LimitedDestSize) { oend += LastLiterals; } if (limit != 0 && op + totalSize > oend) { if (limit == LimitedOutputDirective.LimitedOutput) { return(0); } lastRunSize = (size_t)(oend - op) - 1; litLength = (lastRunSize + 255 - RunMask) / 255; lastRunSize -= litLength; } ip = anchor + lastRunSize; if (lastRunSize >= RunMask) { size_t accumulator = lastRunSize - RunMask; * op++ = (byte)(RunMask << MLBits); for (; accumulator >= 255; accumulator -= 255) { *op++ = 255; } *op++ = (byte)accumulator; } else { *op++ = (byte)(lastRunSize << MLBits); } LZ4MemoryHelper.Copy(op, anchor, (int)lastRunSize); op += lastRunSize; } // End *srcSizePtr = (int)(ip - source); return((int)(op - dst)); _dest_overflow: if (limit != LimitedOutputDirective.LimitedDestSize) { return(0); } op = opSaved; // restore correct out pointer goto _last_literals; }
private static int CompressHashChain(CCtxT *ctx, byte *source, byte *dest, int *srcSizePtr, int maxOutputSize, uint maxNbAttempts, LimitedOutputDirective limit) { int inputSize = *srcSizePtr; int patternAnalysis = maxNbAttempts > 64 ? 1 : 0; // levels 8+ byte *ip = source; byte *anchor = ip; byte *iend = ip + inputSize; byte *mflimit = iend - MFLimit; byte *matchlimit = (iend - LastLiterals); byte *optr; byte *op = dest; byte *oend = op + maxOutputSize; byte *refPos = null; byte *start2 = null; byte *ref2 = null; byte *start3 = null; byte *ref3 = null; // init *srcSizePtr = 0; if (limit == LimitedOutputDirective.LimitedDestSize) { oend -= LastLiterals; // Hack for support LZ4 format restriction } if (inputSize < LZ4MinLength) { goto _last_literals; // Input too small, no compression (all literals) } // Main Loop while (ip < mflimit) { int ml = InsertAndFindBestMatch( ctx, ip, matchlimit, &refPos, (int)maxNbAttempts, patternAnalysis); if (ml < MinMatch) { ip++; continue; } // saved, in case we would skip too much byte *start0 = ip; byte *ref0 = refPos; int ml0 = ml; _Search2: int ml2; if (ip + ml < mflimit) { ml2 = InsertAndGetWiderMatch( ctx, ip + ml - 2, ip + 0, matchlimit, ml, &ref2, &start2, (int)maxNbAttempts, patternAnalysis); } else { ml2 = ml; } if (ml2 == ml) { // No better match optr = op; if (EncodeSequence(&ip, &op, &anchor, ml, refPos, limit, oend) != 0) { goto _dest_overflow; } continue; } if (start0 < ip) { if (start2 < ip + ml0) { // empirical ip = start0; refPos = ref0; ml = ml0; } } // Here, start0==ip if (start2 - ip < 3) { // First Match too small : removed ml = ml2; ip = start2; refPos = ref2; goto _Search2; } _Search3: // At this stage, we have : // ml2 > ml1, and // ip1+3 <= ip2 (usually < ip1+ml1) if ((start2 - ip) < OptimalML) { int newMl = ml; if (newMl > OptimalML) { newMl = OptimalML; } if (ip + newMl > start2 + ml2 - MinMatch) { newMl = (int)(start2 - ip) + ml2 - MinMatch; } int correction = newMl - (int)(start2 - ip); if (correction > 0) { start2 += correction; ref2 += correction; ml2 -= correction; } } // Now, we have start2 = ip+new_ml, with new_ml = min(ml, OPTIMAL_ML=18) int ml3; if (start2 + ml2 < mflimit) { ml3 = InsertAndGetWiderMatch( ctx, start2 + ml2 - 3, start2, matchlimit, ml2, &ref3, &start3, (int)maxNbAttempts, patternAnalysis); } else { ml3 = ml2; } if (ml3 == ml2) { // No better match : 2 sequences to encode // ip & ref are known; Now for ml if (start2 < ip + ml) { ml = (int)(start2 - ip); } // Now, encode 2 sequences optr = op; if (EncodeSequence(&ip, &op, &anchor, ml, refPos, limit, oend) != 0) { goto _dest_overflow; } ip = start2; optr = op; if (EncodeSequence(&ip, &op, &anchor, ml2, ref2, limit, oend) != 0) { goto _dest_overflow; } continue; } if (start3 < ip + ml + 3) { // Not enough space for match 2 : remove it if (start3 >= (ip + ml)) { // can write Seq1 immediately ==> Seq2 is removed, so Seq3 becomes Seq1 if (start2 < ip + ml) { int correction = (int)(ip + ml - start2); start2 += correction; ref2 += correction; ml2 -= correction; if (ml2 < MinMatch) { start2 = start3; ref2 = ref3; ml2 = ml3; } } optr = op; if (EncodeSequence(&ip, &op, &anchor, ml, refPos, limit, oend) != 0) { goto _dest_overflow; } ip = start3; refPos = ref3; ml = ml3; start0 = start2; ref0 = ref2; ml0 = ml2; goto _Search2; } start2 = start3; ref2 = ref3; ml2 = ml3; goto _Search3; } // OK, now we have 3 ascending matches; let's write at least the first one // ip & ref are known; Now for ml if (start2 < ip + ml) { if (start2 - ip < (int)MLMask) { if (ml > OptimalML) { ml = OptimalML; } if (ip + ml > start2 + ml2 - MinMatch) { ml = (int)(start2 - ip) + ml2 - MinMatch; } int correction = ml - (int)(start2 - ip); if (correction > 0) { start2 += correction; ref2 += correction; ml2 -= correction; } } else { ml = (int)(start2 - ip); } } optr = op; if (EncodeSequence(&ip, &op, &anchor, ml, refPos, limit, oend) != 0) { goto _dest_overflow; } ip = start2; refPos = ref2; ml = ml2; start2 = start3; ref2 = ref3; ml2 = ml3; goto _Search3; } _last_literals: // Encode last literals { size_t lastRunSize = (size_t)(iend - anchor); // literals size_t litLength = (lastRunSize + 255 - RunMask) / 255; size_t totalSize = 1 + litLength + lastRunSize; if (limit == LimitedOutputDirective.LimitedDestSize) { oend += LastLiterals; // restore correct value } if ((limit != 0) && (op + totalSize > oend)) { if (limit == LimitedOutputDirective.LimitedOutput) { return(0); // Check output limit } // adapt lastRunSize to fill 'dest' lastRunSize = (size_t)(oend - op) - 1; litLength = (lastRunSize + 255 - RunMask) / 255; lastRunSize -= litLength; } ip = anchor + lastRunSize; if (lastRunSize >= RunMask) { size_t accumulator = lastRunSize - RunMask; *op++ = (byte)(RunMask << MLBits); for (; accumulator >= 255; accumulator -= 255) { *op++ = 255; } *op++ = (byte)accumulator; } else { *op++ = (byte)(lastRunSize << MLBits); } LZ4MemoryHelper.Copy(op, anchor, (int)lastRunSize); op += lastRunSize; } // End *srcSizePtr = (int)(ip - source); return((int)(op - dest)); _dest_overflow: if (limit != LimitedOutputDirective.LimitedDestSize) { return(0); } op = optr; // restore correct out pointer goto _last_literals; }
private static CCtxT *AllocCtx() { return((CCtxT *)LZ4MemoryHelper.Alloc(sizeof(CCtxT))); }