From b148474ef84e6062178a8cb3ed2bdbdb098b633b Mon Sep 17 00:00:00 2001 From: Steeve Morin Date: Sun, 13 Jan 2013 20:20:55 +0100 Subject: [PATCH] Updating to lz4 r88 --- src/lz4.c | 933 ++++++++++++++++++++++++++++------------------------ src/lz4.h | 94 +++--- src/lz4hc.c | 799 ++++++++++++++++++++++++-------------------- 3 files changed, 982 insertions(+), 844 deletions(-) diff --git a/src/lz4.c b/src/lz4.c index 06e2829..a35f12b 100644 --- a/src/lz4.c +++ b/src/lz4.c @@ -34,31 +34,24 @@ //************************************** // Tuning parameters //************************************** -// COMPRESSIONLEVEL : -// Increasing this value improves compression ratio -// Lowering this value reduces memory usage -// Reduced memory usage typically improves speed, due to cache effect (ex : L1 32KB for Intel, L1 64KB for AMD) -// Memory usage formula : N->2^(N+2) Bytes (examples : 12 -> 16KB ; 17 -> 512KB) -#define COMPRESSIONLEVEL 12 - -// NOTCOMPRESSIBLE_CONFIRMATION : +// MEMORY_USAGE : +// Memory usage formula : N->2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; etc.) +// Increasing memory usage improves compression ratio +// Reduced memory usage can improve speed, due to cache effect +// Default value is 14, for 16KB, which nicely fits into Intel x86 L1 cache +#define MEMORY_USAGE 14 + +// NOTCOMPRESSIBLE_DETECTIONLEVEL : // Decreasing this value will make the algorithm skip faster data segments considered "incompressible" // This may decrease compression ratio dramatically, but will be faster on incompressible data // Increasing this value will make the algorithm search more before declaring a segment "incompressible" // This could improve compression a bit, but will be slower on incompressible data // The default value (6) is recommended -#define NOTCOMPRESSIBLE_CONFIRMATION 6 - -// LZ4_COMPRESSMIN : -// Compression function will *fail* if it is not successful at compressing input by at least LZ4_COMPRESSMIN bytes -// Since the compression function stops working prematurely, it results in a speed gain -// The output however is unusable. Compression function result will be zero. -// Default : 0 = disabled -#define LZ4_COMPRESSMIN 0 +#define NOTCOMPRESSIBLE_DETECTIONLEVEL 6 // BIG_ENDIAN_NATIVE_BUT_INCOMPATIBLE : -// This will provide a boost to performance for big endian cpu, but the resulting compressed stream will be incompatible with little-endian CPU. -// You can set this option to 1 in situations where data will stay within closed environment +// This will provide a small boost to performance for big endian cpu, but the resulting compressed stream will be incompatible with little-endian CPU. +// You can set this option to 1 in situations where data will remain within closed environment // This option is useless on Little_Endian CPU (such as x86) //#define BIG_ENDIAN_NATIVE_BUT_INCOMPATIBLE 1 @@ -75,8 +68,18 @@ #endif // Little Endian or Big Endian ? -// Note : overwrite the below #define if you know your architecture endianess -#if (defined(__BIG_ENDIAN__) || defined(__BIG_ENDIAN) || defined(_BIG_ENDIAN) || defined(_ARCH_PPC) || defined(__PPC__) || defined(__PPC) || defined(PPC) || defined(__powerpc__) || defined(__powerpc) || defined(powerpc) || ((defined(__BYTE_ORDER__)&&(__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__))) ) +// Overwrite the #define below if you know your architecture endianess +#if defined (__GLIBC__) +# include +# if (__BYTE_ORDER == __BIG_ENDIAN) +# define LZ4_BIG_ENDIAN 1 +# endif +#elif (defined(__BIG_ENDIAN__) || defined(__BIG_ENDIAN) || defined(_BIG_ENDIAN)) && !(defined(__LITTLE_ENDIAN__) || defined(__LITTLE_ENDIAN) || defined(_LITTLE_ENDIAN)) +# define LZ4_BIG_ENDIAN 1 +#elif defined(__sparc) || defined(__sparc__) \ + || defined(__ppc__) || defined(_POWER) || defined(__powerpc__) || defined(_ARCH_PPC) || defined(__PPC__) || defined(__PPC) || defined(PPC) || defined(__powerpc__) || defined(__powerpc) || defined(powerpc) \ + || defined(__hpux) || defined(__hppa) \ + || defined(_MIPSEB) || defined(__s390__) # define LZ4_BIG_ENDIAN 1 #else // Little Endian assumed. PDP Endian and other very rare endian format are unsupported. @@ -84,7 +87,7 @@ // Unaligned memory access is automatically enabled for "common" CPU, such as x86. // For others CPU, the compiler will be more cautious, and insert extra code to ensure aligned access is respected -// If you know your target CPU supports unaligned memory access, you may want to force this option manually to improve performance +// If you know your target CPU supports unaligned memory access, you want to force this option manually to improve performance #if defined(__ARM_FEATURE_UNALIGNED) # define LZ4_FORCE_UNALIGNED_ACCESS 1 #endif @@ -98,7 +101,7 @@ //************************************** // Compiler Options //************************************** -#if __STDC_VERSION__ >= 199901L // C99 +#if __STDC_VERSION__ >= 199901L // C99 /* "restrict" is a known keyword */ #else # define restrict // Disable restrict @@ -107,7 +110,7 @@ #define GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__) #ifdef _MSC_VER // Visual Studio -# define inline __forceinline // Visual is not C99, but supports some kind of inline +# include // For Visual 2005 # if LZ4_ARCH64 // 64-bit # pragma intrinsic(_BitScanForward64) // For Visual 2005 # pragma intrinsic(_BitScanReverse64) // For Visual 2005 @@ -181,11 +184,11 @@ typedef struct _U64_S { U64 v; } U64_S; //************************************** #define MINMATCH 4 -#define HASH_LOG COMPRESSIONLEVEL +#define HASH_LOG (MEMORY_USAGE-2) #define HASHTABLESIZE (1 << HASH_LOG) #define HASH_MASK (HASHTABLESIZE - 1) -#define SKIPSTRENGTH (NOTCOMPRESSIBLE_CONFIRMATION>2?NOTCOMPRESSIBLE_CONFIRMATION:2) +#define SKIPSTRENGTH (NOTCOMPRESSIBLE_DETECTIONLEVEL>2?NOTCOMPRESSIBLE_DETECTIONLEVEL:2) #define STACKLIMIT 13 #define HEAPMODE (HASH_LOG>STACKLIMIT) // Defines if memory is allocated into the stack (local variable), or into the heap (malloc()). #define COPYLENGTH 8 @@ -196,8 +199,8 @@ typedef struct _U64_S { U64 v; } U64_S; #define MAXD_LOG 16 #define MAX_DISTANCE ((1 << MAXD_LOG) - 1) -#define ML_BITS 4 -#define ML_MASK ((1U<= 304) && !defined(LZ4_FORCE_SW_BITCOUNT) return (__builtin_clzll(val) >> 3); #else - int r; - if (!(val>>32)) { r=4; } else { r=0; val>>=32; } - if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; } - r += (!val); - return r; + int r; + if (!(val>>32)) { r=4; } else { r=0; val>>=32; } + if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; } + r += (!val); + return r; #endif #else #if defined(_MSC_VER) && !defined(LZ4_FORCE_SW_BITCOUNT) @@ -281,15 +284,15 @@ inline static int LZ4_NbCommonBytes (register U64 val) #elif defined(__GNUC__) && (GCC_VERSION >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT) return (__builtin_ctzll(val) >> 3); #else - static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2, 0, 3, 1, 3, 1, 4, 2, 7, 0, 2, 3, 6, 1, 5, 3, 5, 1, 3, 4, 4, 2, 5, 6, 7, 7, 0, 1, 2, 3, 3, 4, 6, 2, 6, 5, 5, 3, 4, 5, 6, 7, 1, 2, 4, 6, 4, 4, 5, 7, 2, 6, 5, 7, 6, 7, 7 }; - return DeBruijnBytePos[((U64)((val & -val) * 0x0218A392CDABBD3F)) >> 58]; + static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2, 0, 3, 1, 3, 1, 4, 2, 7, 0, 2, 3, 6, 1, 5, 3, 5, 1, 3, 4, 4, 2, 5, 6, 7, 7, 0, 1, 2, 3, 3, 4, 6, 2, 6, 5, 5, 3, 4, 5, 6, 7, 1, 2, 4, 6, 4, 4, 5, 7, 2, 6, 5, 7, 6, 7, 7 }; + return DeBruijnBytePos[((U64)((val & -val) * 0x0218A392CDABBD3F)) >> 58]; #endif #endif } #else -inline static int LZ4_NbCommonBytes (register U32 val) +static inline int LZ4_NbCommonBytes (register U32 val) { #if defined(LZ4_BIG_ENDIAN) #if defined(_MSC_VER) && !defined(LZ4_FORCE_SW_BITCOUNT) @@ -299,10 +302,10 @@ inline static int LZ4_NbCommonBytes (register U32 val) #elif defined(__GNUC__) && (GCC_VERSION >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT) return (__builtin_clz(val) >> 3); #else - int r; - if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; } - r += (!val); - return r; + int r; + if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; } + r += (!val); + return r; #endif #else #if defined(_MSC_VER) && !defined(LZ4_FORCE_SW_BITCOUNT) @@ -312,8 +315,8 @@ inline static int LZ4_NbCommonBytes (register U32 val) #elif defined(__GNUC__) && (GCC_VERSION >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT) return (__builtin_ctz(val) >> 3); #else - static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0, 3, 2, 2, 1, 3, 2, 0, 1, 3, 3, 1, 2, 2, 2, 2, 0, 3, 1, 2, 0, 1, 0, 1, 1 }; - return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27]; + static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0, 3, 2, 2, 1, 3, 2, 0, 1, 3, 3, 1, 2, 2, 2, 2, 0, 3, 1, 2, 0, 1, 0, 1, 1 }; + return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27]; #endif #endif } @@ -321,154 +324,173 @@ inline static int LZ4_NbCommonBytes (register U32 val) #endif -//**************************** -// Public functions -//**************************** - -int LZ4_compressBound(int isize) -{ - return (isize + (isize/255) + 16); -} - - //****************************** // Compression functions //****************************** -int LZ4_compressCtx(void** ctx, - const char* source, - char* dest, - int isize) +// LZ4_compressCtx : +// ----------------- +// Compress 'isize' bytes from 'source' into an output buffer 'dest' of maximum size 'maxOutputSize'. +// If it cannot achieve it, compression will stop, and result of the function will be zero. +// return : the number of bytes written in buffer 'dest', or 0 if the compression fails + +static inline int LZ4_compressCtx(void** ctx, + const char* source, + char* dest, + int isize, + int maxOutputSize) { #if HEAPMODE - struct refTables *srt = (struct refTables *) (*ctx); - HTYPE* HashTable; + struct refTables *srt = (struct refTables *) (*ctx); + HTYPE* HashTable; #else - HTYPE HashTable[HASHTABLESIZE] = {0}; + HTYPE HashTable[HASHTABLESIZE] = {0}; #endif - const BYTE* ip = (BYTE*) source; - INITBASE(base); - const BYTE* anchor = ip; - const BYTE* const iend = ip + isize; - const BYTE* const mflimit = iend - MFLIMIT; + const BYTE* ip = (BYTE*) source; + INITBASE(base); + const BYTE* anchor = ip; + const BYTE* const iend = ip + isize; + const BYTE* const mflimit = iend - MFLIMIT; #define matchlimit (iend - LASTLITERALS) - BYTE* op = (BYTE*) dest; + BYTE* op = (BYTE*) dest; + BYTE* const oend = op + maxOutputSize; - int len, length; - const int skipStrength = SKIPSTRENGTH; - U32 forwardH; + int len, length; + const int skipStrength = SKIPSTRENGTH; + U32 forwardH; - // Init - if (isizehashTable); - memset((void*)HashTable, 0, sizeof(srt->hashTable)); + if (*ctx == NULL) + { + srt = (struct refTables *) malloc ( sizeof(struct refTables) ); + *ctx = (void*) srt; + } + HashTable = (HTYPE*)(srt->hashTable); + memset((void*)HashTable, 0, sizeof(srt->hashTable)); #else - (void) ctx; + (void) ctx; #endif - // First Byte - HashTable[LZ4_HASH_VALUE(ip)] = ip - base; - ip++; forwardH = LZ4_HASH_VALUE(ip); + // First Byte + HashTable[LZ4_HASH_VALUE(ip)] = ip - base; + ip++; forwardH = LZ4_HASH_VALUE(ip); - // Main Loop + // Main Loop for ( ; ; ) - { - int findMatchAttempts = (1U << skipStrength) + 3; - const BYTE* forwardIp = ip; - const BYTE* ref; - BYTE* token; - - // Find a match - do { - U32 h = forwardH; - int step = findMatchAttempts++ >> skipStrength; - ip = forwardIp; - forwardIp = ip + step; - - if unlikely(forwardIp > mflimit) { goto _last_literals; } - - forwardH = LZ4_HASH_VALUE(forwardIp); - ref = base + HashTable[h]; - HashTable[h] = ip - base; - - } while ((ref < ip - MAX_DISTANCE) || (A32(ref) != A32(ip))); - - // Catch up - while ((ip>anchor) && (ref>(BYTE*)source) && unlikely(ip[-1]==ref[-1])) { ip--; ref--; } - - // Encode Literal length - length = ip - anchor; - token = op++; - if (length>=(int)RUN_MASK) { *token=(RUN_MASK< 254 ; len-=255) *op++ = 255; *op++ = (BYTE)len; } - else *token = (length<> skipStrength; + ip = forwardIp; + forwardIp = ip + step; + + if unlikely(forwardIp > mflimit) { goto _last_literals; } + + forwardH = LZ4_HASH_VALUE(forwardIp); + ref = base + HashTable[h]; + HashTable[h] = ip - base; + + } while ((ref < ip - MAX_DISTANCE) || (A32(ref) != A32(ip))); + + // Catch up + while ((ip>anchor) && (ref>(BYTE*)source) && unlikely(ip[-1]==ref[-1])) { ip--; ref--; } + + // Encode Literal length + length = (int)(ip - anchor); + token = op++; + if unlikely(op + length + (2 + 1 + LASTLITERALS) + (length>>8) > oend) return 0; // Check output limit +#ifdef _MSC_VER + if (length>=(int)RUN_MASK) + { + int len = length-RUN_MASK; + *token=(RUN_MASK<254) + { + do { *op++ = 255; len -= 255; } while (len>254); + *op++ = (BYTE)len; + memcpy(op, anchor, length); + op += length; + goto _next_match; + } + else + *op++ = (BYTE)len; + } + else *token = (length<=(int)RUN_MASK) { *token=(RUN_MASK< 254 ; len-=255) *op++ = 255; *op++ = (BYTE)len; } + else *token = (length<=(int)ML_MASK) { *token+=ML_MASK; len-=ML_MASK; for(; len > 509 ; len-=510) { *op++ = 255; *op++ = 255; } if (len > 254) { len-=255; *op++ = 255; } *op++ = (BYTE)len; } - else *token += len; + // Encode MatchLength + len = (int)(ip - anchor); + if unlikely(op + (1 + LASTLITERALS) + (len>>8) > oend) return 0; // Check output limit + if (len>=(int)ML_MASK) { *token+=ML_MASK; len-=ML_MASK; for(; len > 509 ; len-=510) { *op++ = 255; *op++ = 255; } if (len > 254) { len-=255; *op++ = 255; } *op++ = (BYTE)len; } + else *token += len; - // Test end of chunk - if (ip > mflimit) { anchor = ip; break; } + // Test end of chunk + if (ip > mflimit) { anchor = ip; break; } - // Fill table - HashTable[LZ4_HASH_VALUE(ip-2)] = ip - 2 - base; + // Fill table + HashTable[LZ4_HASH_VALUE(ip-2)] = ip - 2 - base; - // Test next position - ref = base + HashTable[LZ4_HASH_VALUE(ip)]; - HashTable[LZ4_HASH_VALUE(ip)] = ip - base; - if ((ref > ip - (MAX_DISTANCE + 1)) && (A32(ref) == A32(ip))) { token = op++; *token=0; goto _next_match; } + // Test next position + ref = base + HashTable[LZ4_HASH_VALUE(ip)]; + HashTable[LZ4_HASH_VALUE(ip)] = ip - base; + if ((ref > ip - (MAX_DISTANCE + 1)) && (A32(ref) == A32(ip))) { token = op++; *token=0; goto _next_match; } - // Prepare next loop - anchor = ip++; - forwardH = LZ4_HASH_VALUE(ip); - } + // Prepare next loop + anchor = ip++; + forwardH = LZ4_HASH_VALUE(ip); + } _last_literals: - // Encode Last Literals - { - int lastRun = iend - anchor; - if ((LZ4_COMPRESSMIN>0) && (((op - (BYTE*)dest) + lastRun + 1 + ((lastRun-15)/255)) > isize - LZ4_COMPRESSMIN)) return 0; - if (lastRun>=(int)RUN_MASK) { *op++=(RUN_MASK< 254 ; lastRun-=255) *op++ = 255; *op++ = (BYTE) lastRun; } - else *op++ = (lastRun< (U32)maxOutputSize) return 0; + if (lastRun>=(int)RUN_MASK) { *op++=(RUN_MASK< 254 ; lastRun-=255) *op++ = 255; *op++ = (BYTE) lastRun; } + else *op++ = (lastRun<> ((MINMATCH*8)-HASHLOG64K)) #define LZ4_HASH64K_VALUE(p) LZ4_HASH64K_FUNCTION(A32(p)) -int LZ4_compress64kCtx(void** ctx, - const char* source, - char* dest, - int isize) +static inline int LZ4_compress64kCtx(void** ctx, + const char* source, + char* dest, + int isize, + int maxOutputSize) { #if HEAPMODE - struct refTables *srt = (struct refTables *) (*ctx); - U16* HashTable; + struct refTables *srt = (struct refTables *) (*ctx); + U16* HashTable; #else - U16 HashTable[HASH64KTABLESIZE] = {0}; + U16 HashTable[HASH64KTABLESIZE] = {0}; #endif - const BYTE* ip = (BYTE*) source; - const BYTE* anchor = ip; - const BYTE* const base = ip; - const BYTE* const iend = ip + isize; - const BYTE* const mflimit = iend - MFLIMIT; + const BYTE* ip = (BYTE*) source; + const BYTE* anchor = ip; + const BYTE* const base = ip; + const BYTE* const iend = ip + isize; + const BYTE* const mflimit = iend - MFLIMIT; #define matchlimit (iend - LASTLITERALS) - BYTE* op = (BYTE*) dest; + BYTE* op = (BYTE*) dest; + BYTE* const oend = op + maxOutputSize; - int len, length; - const int skipStrength = SKIPSTRENGTH; - U32 forwardH; + int len, length; + const int skipStrength = SKIPSTRENGTH; + U32 forwardH; - // Init - if (isizehashTable); - memset((void*)HashTable, 0, sizeof(srt->hashTable)); + if (*ctx == NULL) + { + srt = (struct refTables *) malloc ( sizeof(struct refTables) ); + *ctx = (void*) srt; + } + HashTable = (U16*)(srt->hashTable); + memset((void*)HashTable, 0, sizeof(srt->hashTable)); #else - (void) ctx; + (void) ctx; #endif - // First Byte - ip++; forwardH = LZ4_HASH64K_VALUE(ip); + // First Byte + ip++; forwardH = LZ4_HASH64K_VALUE(ip); - // Main Loop + // Main Loop for ( ; ; ) - { - int findMatchAttempts = (1U << skipStrength) + 3; - const BYTE* forwardIp = ip; - const BYTE* ref; - BYTE* token; - - // Find a match - do { - U32 h = forwardH; - int step = findMatchAttempts++ >> skipStrength; - ip = forwardIp; - forwardIp = ip + step; - - if (forwardIp > mflimit) { goto _last_literals; } - - forwardH = LZ4_HASH64K_VALUE(forwardIp); - ref = base + HashTable[h]; - HashTable[h] = ip - base; - - } while (A32(ref) != A32(ip)); - - // Catch up - while ((ip>anchor) && (ref>(BYTE*)source) && (ip[-1]==ref[-1])) { ip--; ref--; } - - // Encode Literal length - length = ip - anchor; - token = op++; - if (length>=(int)RUN_MASK) { *token=(RUN_MASK< 254 ; len-=255) *op++ = 255; *op++ = (BYTE)len; } - else *token = (length<> skipStrength; + ip = forwardIp; + forwardIp = ip + step; + + if (forwardIp > mflimit) { goto _last_literals; } + + forwardH = LZ4_HASH64K_VALUE(forwardIp); + ref = base + HashTable[h]; + HashTable[h] = (U16)(ip - base); + + } while (A32(ref) != A32(ip)); + + // Catch up + while ((ip>anchor) && (ref>(BYTE*)source) && (ip[-1]==ref[-1])) { ip--; ref--; } + + // Encode Literal length + length = (int)(ip - anchor); + token = op++; + if unlikely(op + length + (2 + 1 + LASTLITERALS) + (length>>8) > oend) return 0; // Check output limit +#ifdef _MSC_VER + if (length>=(int)RUN_MASK) + { + int len = length-RUN_MASK; + *token=(RUN_MASK<254) + { + do { *op++ = 255; len -= 255; } while (len>254); + *op++ = (BYTE)len; + memcpy(op, anchor, length); + op += length; + goto _next_match; + } + else + *op++ = (BYTE)len; + } + else *token = (length<=(int)RUN_MASK) { *token=(RUN_MASK< 254 ; len-=255) *op++ = 255; *op++ = (BYTE)len; } + else *token = (length<=(int)ML_MASK) { *token+=ML_MASK; len-=ML_MASK; for(; len > 509 ; len-=510) { *op++ = 255; *op++ = 255; } if (len > 254) { len-=255; *op++ = 255; } *op++ = (BYTE)len; } - else *token += len; + // Encode MatchLength + len = (int)(ip - anchor); + if unlikely(op + (1 + LASTLITERALS) + (len>>8) > oend) return 0; // Check output limit + if (len>=(int)ML_MASK) { *token+=ML_MASK; len-=ML_MASK; for(; len > 509 ; len-=510) { *op++ = 255; *op++ = 255; } if (len > 254) { len-=255; *op++ = 255; } *op++ = (BYTE)len; } + else *token += len; - // Test end of chunk - if (ip > mflimit) { anchor = ip; break; } + // Test end of chunk + if (ip > mflimit) { anchor = ip; break; } - // Fill table - HashTable[LZ4_HASH64K_VALUE(ip-2)] = ip - 2 - base; + // Fill table + HashTable[LZ4_HASH64K_VALUE(ip-2)] = (U16)(ip - 2 - base); - // Test next position - ref = base + HashTable[LZ4_HASH64K_VALUE(ip)]; - HashTable[LZ4_HASH64K_VALUE(ip)] = ip - base; - if (A32(ref) == A32(ip)) { token = op++; *token=0; goto _next_match; } + // Test next position + ref = base + HashTable[LZ4_HASH64K_VALUE(ip)]; + HashTable[LZ4_HASH64K_VALUE(ip)] = (U16)(ip - base); + if (A32(ref) == A32(ip)) { token = op++; *token=0; goto _next_match; } - // Prepare next loop - anchor = ip++; - forwardH = LZ4_HASH64K_VALUE(ip); - } + // Prepare next loop + anchor = ip++; + forwardH = LZ4_HASH64K_VALUE(ip); + } _last_literals: - // Encode Last Literals - { - int lastRun = iend - anchor; - if ((LZ4_COMPRESSMIN>0) && (((op - (BYTE*)dest) + lastRun + 1 + ((lastRun-15)/255)) > isize - LZ4_COMPRESSMIN)) return 0; - if (lastRun>=(int)RUN_MASK) { *op++=(RUN_MASK< 254 ; lastRun-=255) *op++ = 255; *op++ = (BYTE) lastRun; } - else *op++ = (lastRun< oend) return 0; + if (lastRun>=(int)RUN_MASK) { *op++=(RUN_MASK< 254 ; lastRun-=255) *op++ = 255; *op++ = (BYTE) lastRun; } + else *op++ = (lastRun<>ML_BITS)) == RUN_MASK) { for (;(len=*ip++)==255;length+=255){} length += len; } - - // copy literals - cpy = op+length; - if unlikely(cpy>oend-COPYLENGTH) - { - if (cpy > oend) goto _output_error; // Error : request to write beyond destination buffer - memcpy(op, ip, length); - ip += length; - break; // Necessarily EOF - } - LZ4_WILDCOPY(ip, op, cpy); ip -= (op-cpy); op = cpy; - - // get offset - LZ4_READ_LITTLEENDIAN_16(ref,cpy,ip); ip+=2; - if (ref < (BYTE* const)dest) goto _output_error; // Error : offset create reference outside destination buffer - - // get matchlength - if ((length=(token&ML_MASK)) == ML_MASK) { for (;*ip==255;length+=255) {ip++;} length += *ip++; } - - // copy repeated sequence - if unlikely(op-ref>ML_BITS)) == RUN_MASK) { size_t len; for (;(len=*ip++)==255;length+=255){} length += len; } + + // copy literals + cpy = op+length; + if unlikely(cpy>oend-COPYLENGTH) + { + if (cpy != oend) goto _output_error; // Error : not enough place for another match (min 4) + 5 literals + memcpy(op, ip, length); + ip += length; + break; // EOF + } + LZ4_WILDCOPY(ip, op, cpy); ip -= (op-cpy); op = cpy; + + // get offset + LZ4_READ_LITTLEENDIAN_16(ref,cpy,ip); ip+=2; + if unlikely(ref < (BYTE* const)dest) goto _output_error; // Error : offset create reference outside destination buffer + + // get matchlength + if ((length=(token&ML_MASK)) == ML_MASK) { for (;*ip==255;length+=255) {ip++;} length += *ip++; } + + // copy repeated sequence + if unlikely((op-ref)oend-COPYLENGTH) - { - if (cpy > oend) goto _output_error; // Error : request to write beyond destination buffer - LZ4_SECURECOPY(ref, op, (oend-COPYLENGTH)); - while(opoend-COPYLENGTH) + { + if (cpy > oend) goto _output_error; // Error : request to write beyond destination buffer + LZ4_SECURECOPY(ref, op, (oend-COPYLENGTH)); + while(op>ML_BITS)) == RUN_MASK) { int s=255; while ((ipoend-COPYLENGTH) || (ip+length>iend-COPYLENGTH)) - { - if (cpy > oend) goto _output_error; // Error : request to write beyond destination buffer - if (ip+length > iend) goto _output_error; // Error : request to read beyond source buffer - memcpy(op, ip, length); - op += length; - ip += length; - if (ip>ML_BITS)) == RUN_MASK) { int s=255; while ((ipoend-COPYLENGTH) || (ip+length>iend-COPYLENGTH)) + { + if (cpy > oend) goto _output_error; // Error : writes beyond output buffer + if (ip+length != iend) goto _output_error; // Error : LZ4 format requires to consume all input at this stage + memcpy(op, ip, length); + op += length; + break; // Necessarily EOF, due to parsing restrictions + } + LZ4_WILDCOPY(ip, op, cpy); ip -= (op-cpy); op = cpy; + + // get offset + LZ4_READ_LITTLEENDIAN_16(ref,cpy,ip); ip+=2; + if (ref < (BYTE* const)dest) goto _output_error; // Error : offset creates reference outside of destination buffer + + // get matchlength + if ((length=(token&ML_MASK)) == ML_MASK) { while (ipoend-COPYLENGTH) - { - if (cpy > oend) goto _output_error; // Error : request to write outside of destination buffer - LZ4_SECURECOPY(ref, op, (oend-COPYLENGTH)); - while(opoend-COPYLENGTH) + { + if (cpy > oend) goto _output_error; // Error : request to write outside of destination buffer + LZ4_SECURECOPY(ref, op, (oend-COPYLENGTH)); + while(op +# if (__BYTE_ORDER == __BIG_ENDIAN) +# define LZ4_BIG_ENDIAN 1 +# endif +#elif (defined(__BIG_ENDIAN__) || defined(__BIG_ENDIAN) || defined(_BIG_ENDIAN)) && !(defined(__LITTLE_ENDIAN__) || defined(__LITTLE_ENDIAN) || defined(_LITTLE_ENDIAN)) +# define LZ4_BIG_ENDIAN 1 +#elif defined(__sparc) || defined(__sparc__) \ + || defined(__ppc__) || defined(_POWER) || defined(__powerpc__) || defined(_ARCH_PPC) || defined(__PPC__) || defined(__PPC) || defined(PPC) || defined(__powerpc__) || defined(__powerpc) || defined(powerpc) \ + || defined(__hpux) || defined(__hppa) \ + || defined(_MIPSEB) || defined(__s390__) +# define LZ4_BIG_ENDIAN 1 #else // Little Endian assumed. PDP Endian and other very rare endian format are unsupported. #endif // Unaligned memory access is automatically enabled for "common" CPU, such as x86. // For others CPU, the compiler will be more cautious, and insert extra code to ensure aligned access is respected -// If you know your target CPU supports unaligned memory access, you may want to force this option manually to improve performance +// If you know your target CPU supports unaligned memory access, you want to force this option manually to improve performance #if defined(__ARM_FEATURE_UNALIGNED) -#define LZ4_FORCE_UNALIGNED_ACCESS 1 +# define LZ4_FORCE_UNALIGNED_ACCESS 1 +#endif + +// Define this parameter if your target system or compiler does not support hardware bit count +#if defined(_MSC_VER) && defined(_WIN32_WCE) // Visual Studio for Windows CE does not support Hardware bit count +# define LZ4_FORCE_SW_BITCOUNT #endif @@ -63,17 +79,32 @@ #if __STDC_VERSION__ >= 199901L // C99 /* "restrict" is a known keyword */ #else -#define restrict // Disable restrict +# define restrict // Disable restrict #endif #ifdef _MSC_VER -#define inline __forceinline // Visual is not C99, but supports some kind of inline +# define inline __inline // Visual is not C99, but supports some kind of inline +# define forceinline __forceinline +# include // For Visual 2005 +# if LZ4_ARCH64 // 64-bit +# pragma intrinsic(_BitScanForward64) // For Visual 2005 +# pragma intrinsic(_BitScanReverse64) // For Visual 2005 +# else +# pragma intrinsic(_BitScanForward) // For Visual 2005 +# pragma intrinsic(_BitScanReverse) // For Visual 2005 +# endif +#else +# ifdef __GNUC__ +# define forceinline inline __attribute__((always_inline)) +# else +# define forceinline inline +# endif #endif #ifdef _MSC_VER // Visual Studio -#define bswap16(x) _byteswap_ushort(x) +#define lz4_bswap16(x) _byteswap_ushort(x) #else -#define bswap16(x) ((unsigned short int) ((((x) >> 8) & 0xffu) | (((x) & 0xffu) << 8))) +#define lz4_bswap16(x) ((unsigned short int) ((((x) >> 8) & 0xffu) | (((x) & 0xffu) << 8))) #endif @@ -174,8 +205,8 @@ typedef struct _U64_S { U64 v; } U64_S; #endif #if defined(LZ4_BIG_ENDIAN) -#define LZ4_READ_LITTLEENDIAN_16(d,s,p) { U16 v = A16(p); v = bswap16(v); d = (s) - v; } -#define LZ4_WRITE_LITTLEENDIAN_16(p,i) { U16 v = (U16)(i); v = bswap16(v); A16(p) = v; p+=2; } +#define LZ4_READ_LITTLEENDIAN_16(d,s,p) { U16 v = A16(p); v = lz4_bswap16(v); d = (s) - v; } +#define LZ4_WRITE_LITTLEENDIAN_16(p,i) { U16 v = (U16)(i); v = lz4_bswap16(v); A16(p) = v; p+=2; } #else // Little Endian #define LZ4_READ_LITTLEENDIAN_16(d,s,p) { d = (s) - A16(p); } #define LZ4_WRITE_LITTLEENDIAN_16(p,v) { A16(p) = v; p+=2; } @@ -187,24 +218,23 @@ typedef struct _U64_S { U64 v; } U64_S; //************************************************************ typedef struct { - const BYTE* base; - HTYPE hashTable[HASHTABLESIZE]; - U16 chainTable[MAXD]; - const BYTE* nextToUpdate; + const BYTE* base; + HTYPE hashTable[HASHTABLESIZE]; + U16 chainTable[MAXD]; + const BYTE* nextToUpdate; } LZ4HC_Data_Structure; //************************************** // Macros //************************************** -#define LZ4_WILDCOPY(s,d,e) do { LZ4_COPYPACKET(s,d) } while (d> ((MINMATCH*8)-HASH_LOG)) -#define HASH_VALUE(p) HASH_FUNCTION(*(U32*)(p)) -#define HASH_POINTER(p) (HashTable[HASH_VALUE(p)] + base) -#define DELTANEXT(p) chainTable[(size_t)(p) & MAXD_MASK] -#define GETNEXT(p) ((p) - (size_t)DELTANEXT(p)) -#define ADD_HASH(p) { size_t delta = (p) - HASH_POINTER(p); if (delta>MAX_DISTANCE) delta = MAX_DISTANCE; DELTANEXT(p) = (U16)delta; HashTable[HASH_VALUE(p)] = (p) - base; } +#define LZ4_WILDCOPY(s,d,e) do { LZ4_COPYPACKET(s,d) } while (d> ((MINMATCH*8)-HASH_LOG)) +#define HASH_VALUE(p) HASH_FUNCTION(A32(p)) +#define HASH_POINTER(p) (HashTable[HASH_VALUE(p)] + base) +#define DELTANEXT(p) chainTable[(size_t)(p) & MAXD_MASK] +#define GETNEXT(p) ((p) - (size_t)DELTANEXT(p)) //************************************** @@ -222,11 +252,11 @@ inline static int LZ4_NbCommonBytes (register U64 val) #elif defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT) return (__builtin_clzll(val) >> 3); #else - int r; - if (!(val>>32)) { r=4; } else { r=0; val>>=32; } - if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; } - r += (!val); - return r; + int r; + if (!(val>>32)) { r=4; } else { r=0; val>>=32; } + if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; } + r += (!val); + return r; #endif #else #if defined(_MSC_VER) && !defined(LZ4_FORCE_SW_BITCOUNT) @@ -236,8 +266,8 @@ inline static int LZ4_NbCommonBytes (register U64 val) #elif defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT) return (__builtin_ctzll(val) >> 3); #else - static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2, 0, 3, 1, 3, 1, 4, 2, 7, 0, 2, 3, 6, 1, 5, 3, 5, 1, 3, 4, 4, 2, 5, 6, 7, 7, 0, 1, 2, 3, 3, 4, 6, 2, 6, 5, 5, 3, 4, 5, 6, 7, 1, 2, 4, 6, 4, 4, 5, 7, 2, 6, 5, 7, 6, 7, 7 }; - return DeBruijnBytePos[((U64)((val & -val) * 0x0218A392CDABBD3F)) >> 58]; + static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2, 0, 3, 1, 3, 1, 4, 2, 7, 0, 2, 3, 6, 1, 5, 3, 5, 1, 3, 4, 4, 2, 5, 6, 7, 7, 0, 1, 2, 3, 3, 4, 6, 2, 6, 5, 5, 3, 4, 5, 6, 7, 1, 2, 4, 6, 4, 4, 5, 7, 2, 6, 5, 7, 6, 7, 7 }; + return DeBruijnBytePos[((U64)((val & -val) * 0x0218A392CDABBD3F)) >> 58]; #endif #endif } @@ -248,27 +278,27 @@ inline static int LZ4_NbCommonBytes (register U32 val) { #if defined(LZ4_BIG_ENDIAN) #if defined(_MSC_VER) && !defined(LZ4_FORCE_SW_BITCOUNT) - unsigned long r = 0; + unsigned long r; _BitScanReverse( &r, val ); return (int)(r>>3); #elif defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT) return (__builtin_clz(val) >> 3); #else - int r; - if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; } - r += (!val); - return r; + int r; + if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; } + r += (!val); + return r; #endif #else #if defined(_MSC_VER) && !defined(LZ4_FORCE_SW_BITCOUNT) - unsigned long r = 0; + unsigned long r; _BitScanForward( &r, val ); return (int)(r>>3); #elif defined(__GNUC__) && ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) && !defined(LZ4_FORCE_SW_BITCOUNT) return (__builtin_ctz(val) >> 3); #else - static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0, 3, 2, 2, 1, 3, 2, 0, 1, 3, 3, 1, 2, 2, 2, 2, 0, 3, 1, 2, 0, 1, 0, 1, 1 }; - return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27]; + static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0, 3, 2, 2, 1, 3, 2, 0, 1, 3, 3, 1, 2, 2, 2, 2, 0, 3, 1, 2, 0, 1, 0, 1, 1 }; + return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27]; #endif #endif } @@ -278,166 +308,211 @@ inline static int LZ4_NbCommonBytes (register U32 val) inline static int LZ4HC_Init (LZ4HC_Data_Structure* hc4, const BYTE* base) { - MEM_INIT((void*)hc4->hashTable, 0, sizeof(hc4->hashTable)); - MEM_INIT(hc4->chainTable, 0xFF, sizeof(hc4->chainTable)); - hc4->nextToUpdate = base + LZ4_ARCH64; - hc4->base = base; - return 1; + MEM_INIT((void*)hc4->hashTable, 0, sizeof(hc4->hashTable)); + MEM_INIT(hc4->chainTable, 0xFF, sizeof(hc4->chainTable)); + hc4->nextToUpdate = base + LZ4_ARCH64; + hc4->base = base; + return 1; } inline static void* LZ4HC_Create (const BYTE* base) { - void* hc4 = ALLOCATOR(sizeof(LZ4HC_Data_Structure)); + void* hc4 = ALLOCATOR(sizeof(LZ4HC_Data_Structure)); - LZ4HC_Init (hc4, base); - return hc4; + LZ4HC_Init ((LZ4HC_Data_Structure*)hc4, base); + return hc4; } inline static int LZ4HC_Free (void** LZ4HC_Data) { - FREEMEM(*LZ4HC_Data); - *LZ4HC_Data = NULL; - return (1); + FREEMEM(*LZ4HC_Data); + *LZ4HC_Data = NULL; + return (1); } -inline static void LZ4HC_Insert (LZ4HC_Data_Structure* hc4, const BYTE* ip) +// Update chains up to ip (excluded) +forceinline static void LZ4HC_Insert (LZ4HC_Data_Structure* hc4, const BYTE* ip) { - U16* chainTable = hc4->chainTable; - HTYPE* HashTable = hc4->hashTable; - INITBASE(base,hc4->base); - - while(hc4->nextToUpdate < ip) - { - ADD_HASH(hc4->nextToUpdate); - hc4->nextToUpdate++; - } + U16* chainTable = hc4->chainTable; + HTYPE* HashTable = hc4->hashTable; + INITBASE(base,hc4->base); + + while(hc4->nextToUpdate < ip) + { + const BYTE* p = hc4->nextToUpdate; + size_t delta = (p) - HASH_POINTER(p); + if (delta>MAX_DISTANCE) delta = MAX_DISTANCE; + DELTANEXT(p) = (U16)delta; + HashTable[HASH_VALUE(p)] = (p) - base; + hc4->nextToUpdate++; + } } -inline static int LZ4HC_InsertAndFindBestMatch (LZ4HC_Data_Structure* hc4, const BYTE* ip, const BYTE* const matchlimit, const BYTE** matchpos) +forceinline static size_t LZ4HC_CommonLength (const BYTE* p1, const BYTE* p2, const BYTE* const matchlimit) { - U16* const chainTable = hc4->chainTable; - HTYPE* const HashTable = hc4->hashTable; - const BYTE* ref; - INITBASE(base,hc4->base); - int nbAttempts=MAX_NB_ATTEMPTS; - int ml=0; - - // HC4 match finder - LZ4HC_Insert(hc4, ip); - ref = HASH_POINTER(ip); - while ((ref > (ip-MAX_DISTANCE)) && (nbAttempts)) - { - nbAttempts--; - if (*(ref+ml) == *(ip+ml)) - if (*(U32*)ref == *(U32*)ip) - { - const BYTE* reft = ref+MINMATCH; - const BYTE* ipt = ip+MINMATCH; - - while (ipt ml) { ml = ipt-ip; *matchpos = ref; } - } - ref = GETNEXT(ref); - } - return ml; +forceinline static int LZ4HC_InsertAndFindBestMatch (LZ4HC_Data_Structure* hc4, const BYTE* ip, const BYTE* const matchlimit, const BYTE** matchpos) +{ + U16* const chainTable = hc4->chainTable; + HTYPE* const HashTable = hc4->hashTable; + const BYTE* ref; + INITBASE(base,hc4->base); + int nbAttempts=MAX_NB_ATTEMPTS; + size_t ml=0; + + // HC4 match finder + LZ4HC_Insert(hc4, ip); + ref = HASH_POINTER(ip); + +#if 1 + if (ref >= ip-4) // potential repetition + { + if (A32(ref) == A32(ip)) // confirmed + { + const U16 delta = (U16)(ip-ref); + const BYTE* ptr = ip; + const BYTE* end; + ml = LZ4HC_CommonLength(ip+MINMATCH, ref+MINMATCH, matchlimit) + MINMATCH; + end = ip + ml - (MINMATCH-1); + while(ptr < end-delta) + { + DELTANEXT(ptr) = delta; // Pre-Load + ptr++; + } + do + { + DELTANEXT(ptr) = delta; + HashTable[HASH_VALUE(ptr)] = (ptr) - base; // Head of chain + ptr++; + } while(ptr < end); + hc4->nextToUpdate = end; + *matchpos = ref; + } + ref = GETNEXT(ref); + } +#endif + + while ((ref >= (ip-MAX_DISTANCE)) && (nbAttempts)) + { + nbAttempts--; + if (*(ref+ml) == *(ip+ml)) + if (A32(ref) == A32(ip)) + { + size_t mlt = LZ4HC_CommonLength(ip+MINMATCH, ref+MINMATCH, matchlimit) + MINMATCH; + if (mlt > ml) { ml = mlt; *matchpos = ref; } + } + ref = GETNEXT(ref); + } + + return (int)ml; } -inline static int LZ4HC_InsertAndGetWiderMatch (LZ4HC_Data_Structure* hc4, const BYTE* ip, const BYTE* startLimit, const BYTE* matchlimit, int longest, const BYTE** matchpos, const BYTE** startpos) +forceinline static int LZ4HC_InsertAndGetWiderMatch (LZ4HC_Data_Structure* hc4, const BYTE* ip, const BYTE* startLimit, const BYTE* matchlimit, int longest, const BYTE** matchpos, const BYTE** startpos) { - U16* const chainTable = hc4->chainTable; - HTYPE* const HashTable = hc4->hashTable; - INITBASE(base,hc4->base); - const BYTE* ref; - int nbAttempts = MAX_NB_ATTEMPTS; - int delta = ip-startLimit; - - // First Match - LZ4HC_Insert(hc4, ip); - ref = HASH_POINTER(ip); - - while ((ref > ip-MAX_DISTANCE) && (ref >= hc4->base) && (nbAttempts)) - { - nbAttempts--; - if (*(startLimit + longest) == *(ref - delta + longest)) - if (*(U32*)ref == *(U32*)ip) - { - const BYTE* reft = ref+MINMATCH; - const BYTE* ipt = ip+MINMATCH; - const BYTE* startt = ip; - - while (iptchainTable; + HTYPE* const HashTable = hc4->hashTable; + INITBASE(base,hc4->base); + const BYTE* ref; + int nbAttempts = MAX_NB_ATTEMPTS; + int delta = (int)(ip-startLimit); + + // First Match + LZ4HC_Insert(hc4, ip); + ref = HASH_POINTER(ip); + + while ((ref >= ip-MAX_DISTANCE) && (ref >= hc4->base) && (nbAttempts)) + { + nbAttempts--; + if (*(startLimit + longest) == *(ref - delta + longest)) + if (A32(ref) == A32(ip)) + { +#if 1 + const BYTE* reft = ref+MINMATCH; + const BYTE* ipt = ip+MINMATCH; + const BYTE* startt = ip; + + while (iptstartLimit) && (reft > hc4->base) && (startt[-1] == reft[-1])) {startt--; reft--;} + while ((startt>startLimit) && (reft > hc4->base) && (startt[-1] == reft[-1])) {startt--; reft--;} - if ((ipt-startt) > longest) - { - longest = ipt-startt; - *matchpos = reft; - *startpos = startt; - } - } - ref = GETNEXT(ref); - } + if ((ipt-startt) > longest) + { + longest = (int)(ipt-startt); + *matchpos = reft; + *startpos = startt; + } + } + ref = GETNEXT(ref); + } - return longest; + return longest; } -inline static int LZ4_encodeSequence(const BYTE** ip, BYTE** op, const BYTE** anchor, int ml, const BYTE* ref) +forceinline static int LZ4_encodeSequence(const BYTE** ip, BYTE** op, const BYTE** anchor, int ml, const BYTE* ref) { - int length, len; - BYTE* token; + int length, len; + BYTE* token; - // Encode Literal length - length = *ip - *anchor; - token = (*op)++; - if (length>=(int)RUN_MASK) { *token=(RUN_MASK< 254 ; len-=255) *(*op)++ = 255; *(*op)++ = (BYTE)len; } - else *token = (length<=(int)RUN_MASK) { *token=(RUN_MASK< 254 ; len-=255) *(*op)++ = 255; *(*op)++ = (BYTE)len; } + else *token = (length<=(int)ML_MASK) { *token+=ML_MASK; len-=ML_MASK; for(; len > 509 ; len-=510) { *(*op)++ = 255; *(*op)++ = 255; } if (len > 254) { len-=255; *(*op)++ = 255; } *(*op)++ = (BYTE)len; } - else *token += len; + // Encode MatchLength + len = (int)(ml-MINMATCH); + if (len>=(int)ML_MASK) { *token+=ML_MASK; len-=ML_MASK; for(; len > 509 ; len-=510) { *(*op)++ = 255; *(*op)++ = 255; } if (len > 254) { len-=255; *(*op)++ = 255; } *(*op)++ = (BYTE)len; } + else *token += len; - // Prepare next loop - *ip += ml; - *anchor = *ip; + // Prepare next loop + *ip += ml; + *anchor = *ip; - return 0; + return 0; } @@ -446,218 +521,218 @@ inline static int LZ4_encodeSequence(const BYTE** ip, BYTE** op, const BYTE** an //**************************** int LZ4_compressHCCtx(LZ4HC_Data_Structure* ctx, - const char* source, - char* dest, - int isize) + const char* source, + char* dest, + int isize) { - const BYTE* ip = (const BYTE*) source; - const BYTE* anchor = ip; - const BYTE* const iend = ip + isize; - const BYTE* const mflimit = iend - MFLIMIT; - const BYTE* const matchlimit = (iend - LASTLITERALS); - - BYTE* op = (BYTE*) dest; - - int ml, ml2, ml3, ml0; - const BYTE* ref=NULL; - const BYTE* start2=NULL; - const BYTE* ref2=NULL; - const BYTE* start3=NULL; - const BYTE* ref3=NULL; - const BYTE* start0; - const BYTE* ref0; - - ip++; - - // Main Loop - while (ip < mflimit) - { - ml = LZ4HC_InsertAndFindBestMatch (ctx, ip, matchlimit, (&ref)); - if (!ml) { ip++; continue; } - - // saved, in case we would skip too much - start0 = ip; - ref0 = ref; - ml0 = ml; + const BYTE* ip = (const BYTE*) source; + const BYTE* anchor = ip; + const BYTE* const iend = ip + isize; + const BYTE* const mflimit = iend - MFLIMIT; + const BYTE* const matchlimit = (iend - LASTLITERALS); + + BYTE* op = (BYTE*) dest; + + int ml, ml2, ml3, ml0; + const BYTE* ref=NULL; + const BYTE* start2=NULL; + const BYTE* ref2=NULL; + const BYTE* start3=NULL; + const BYTE* ref3=NULL; + const BYTE* start0; + const BYTE* ref0; + + ip++; + + // Main Loop + while (ip < mflimit) + { + ml = LZ4HC_InsertAndFindBestMatch (ctx, ip, matchlimit, (&ref)); + if (!ml) { ip++; continue; } + + // saved, in case we would skip too much + start0 = ip; + ref0 = ref; + ml0 = ml; _Search2: - if (ip+ml < mflimit) - ml2 = LZ4HC_InsertAndGetWiderMatch(ctx, ip + ml - 2, ip + 1, matchlimit, ml, &ref2, &start2); - else ml2=ml; - - if (ml2 == ml) // No better match - { - LZ4_encodeSequence(&ip, &op, &anchor, ml, ref); - continue; - } - - if (start0 < ip) - { - if (start2 < ip + ml0) // empirical - { - ip = start0; - ref = ref0; - ml = ml0; - } - } - - // Here, start0==ip - if ((start2 - ip) < 3) // First Match too small : removed - { - ml = ml2; - ip = start2; - ref =ref2; - goto _Search2; - } + if (ip+ml < mflimit) + ml2 = LZ4HC_InsertAndGetWiderMatch(ctx, ip + ml - 2, ip + 1, matchlimit, ml, &ref2, &start2); + else ml2=ml; + + if (ml2 == ml) // No better match + { + LZ4_encodeSequence(&ip, &op, &anchor, ml, ref); + continue; + } + + if (start0 < ip) + { + if (start2 < ip + ml0) // empirical + { + ip = start0; + ref = ref0; + ml = ml0; + } + } + + // Here, start0==ip + if ((start2 - ip) < 3) // First Match too small : removed + { + ml = ml2; + ip = start2; + ref =ref2; + goto _Search2; + } _Search3: - // Currently we have : - // ml2 > ml1, and - // ip1+3 <= ip2 (usually < ip1+ml1) - if ((start2 - ip) < OPTIMAL_ML) - { - int correction; - int new_ml = ml; - if (new_ml > OPTIMAL_ML) new_ml = OPTIMAL_ML; - if (ip+new_ml > start2 + ml2 - MINMATCH) new_ml = start2 - ip + ml2 - MINMATCH; - correction = new_ml - (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) - - if (start2 + ml2 < mflimit) - ml3 = LZ4HC_InsertAndGetWiderMatch(ctx, start2 + ml2 - 3, start2, matchlimit, ml2, &ref3, &start3); - else ml3=ml2; - - if (ml3 == ml2) // No better match : 2 sequences to encode - { - // ip & ref are known; Now for ml - if (start2 < ip+ml) - { - if ((start2 - ip) < OPTIMAL_ML) - { - int correction; - if (ml > OPTIMAL_ML) ml = OPTIMAL_ML; - if (ip+ml > start2 + ml2 - MINMATCH) ml = start2 - ip + ml2 - MINMATCH; - correction = ml - (start2 - ip); - if (correction > 0) - { - start2 += correction; - ref2 += correction; - ml2 -= correction; - } - } - else - { - ml = start2 - ip; - } - } - // Now, encode 2 sequences - LZ4_encodeSequence(&ip, &op, &anchor, ml, ref); - ip = start2; - LZ4_encodeSequence(&ip, &op, &anchor, ml2, ref2); - 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 = (ip+ml) - start2; - start2 += correction; - ref2 += correction; - ml2 -= correction; - if (ml2 < MINMATCH) - { - start2 = start3; - ref2 = ref3; - ml2 = ml3; - } - } - - LZ4_encodeSequence(&ip, &op, &anchor, ml, ref); - ip = start3; - ref = 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)ML_MASK) - { - int correction; - if (ml > OPTIMAL_ML) ml = OPTIMAL_ML; - if (ip + ml > start2 + ml2 - MINMATCH) ml = start2 - ip + ml2 - MINMATCH; - correction = ml - (start2 - ip); - if (correction > 0) - { - start2 += correction; - ref2 += correction; - ml2 -= correction; - } - } - else - { - ml = start2 - ip; - } - } - LZ4_encodeSequence(&ip, &op, &anchor, ml, ref); - - ip = start2; - ref = ref2; - ml = ml2; - - start2 = start3; - ref2 = ref3; - ml2 = ml3; - - goto _Search3; - - } - - // Encode Last Literals - { - int lastRun = iend - anchor; - if (lastRun>=(int)RUN_MASK) { *op++=(RUN_MASK< 254 ; lastRun-=255) *op++ = 255; *op++ = (BYTE) lastRun; } - else *op++ = (lastRun< ml1, and + // ip1+3 <= ip2 (usually < ip1+ml1) + if ((start2 - ip) < OPTIMAL_ML) + { + int correction; + int new_ml = ml; + if (new_ml > OPTIMAL_ML) new_ml = OPTIMAL_ML; + if (ip+new_ml > start2 + ml2 - MINMATCH) new_ml = (int)(start2 - ip) + ml2 - MINMATCH; + correction = new_ml - (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) + + if (start2 + ml2 < mflimit) + ml3 = LZ4HC_InsertAndGetWiderMatch(ctx, start2 + ml2 - 3, start2, matchlimit, ml2, &ref3, &start3); + else ml3=ml2; + + if (ml3 == ml2) // No better match : 2 sequences to encode + { + // ip & ref are known; Now for ml + if (start2 < ip+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; + correction = ml - (int)(start2 - ip); + if (correction > 0) + { + start2 += correction; + ref2 += correction; + ml2 -= correction; + } + } + else + { + ml = (int)(start2 - ip); + } + } + // Now, encode 2 sequences + LZ4_encodeSequence(&ip, &op, &anchor, ml, ref); + ip = start2; + LZ4_encodeSequence(&ip, &op, &anchor, ml2, ref2); + 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; + } + } + + LZ4_encodeSequence(&ip, &op, &anchor, ml, ref); + ip = start3; + ref = 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)ML_MASK) + { + int correction; + if (ml > OPTIMAL_ML) ml = OPTIMAL_ML; + if (ip + ml > start2 + ml2 - MINMATCH) ml = (int)(start2 - ip) + ml2 - MINMATCH; + correction = ml - (int)(start2 - ip); + if (correction > 0) + { + start2 += correction; + ref2 += correction; + ml2 -= correction; + } + } + else + { + ml = (int)(start2 - ip); + } + } + LZ4_encodeSequence(&ip, &op, &anchor, ml, ref); + + ip = start2; + ref = ref2; + ml = ml2; + + start2 = start3; + ref2 = ref3; + ml2 = ml3; + + goto _Search3; + + } + + // Encode Last Literals + { + int lastRun = (int)(iend - anchor); + if (lastRun>=(int)RUN_MASK) { *op++=(RUN_MASK< 254 ; lastRun-=255) *op++ = 255; *op++ = (BYTE) lastRun; } + else *op++ = (lastRun<