From 296e2ab6b2066f140e67d9ba3ff1c1b2cb71e44a Mon Sep 17 00:00:00 2001 From: Moinak Ghosh Date: Wed, 25 Jul 2012 21:07:36 +0530 Subject: [PATCH] Add support for LZ4 compression including multi-pass LZ4. Add missing Read_Adjusted() declaration, was causing a crash with 2GB chunks. Fix minor cut-paste issues in comments. --- Makefile | 24 +- lz4/lz4.c | 819 ++++++++++++++++++++++++++++++++++++++++++++++++ lz4/lz4.h | 142 +++++++++ lz4/lz4hc.c | 683 ++++++++++++++++++++++++++++++++++++++++ lz4/lz4hc.h | 82 +++++ lz4_compress.c | 141 +++++++++ lzfx/lzfx.c | 3 - lzfx/lzfx.h | 3 - lzfx_compress.c | 2 +- main.c | 8 + pcompress.h | 7 + utils.h | 2 + 12 files changed, 1902 insertions(+), 14 deletions(-) create mode 100644 lz4/lz4.c create mode 100644 lz4/lz4.h create mode 100644 lz4/lz4hc.c create mode 100644 lz4/lz4hc.h create mode 100644 lz4_compress.c diff --git a/Makefile b/Makefile index 1e5d2b9..513c066 100644 --- a/Makefile +++ b/Makefile @@ -23,7 +23,8 @@ PROG= pcompress MAINSRCS = main.c utils.c allocator.c zlib_compress.c bzip2_compress.c \ - lzma_compress.c ppmd_compress.c adaptive_compress.c lzfx_compress.c + lzma_compress.c ppmd_compress.c adaptive_compress.c lzfx_compress.c \ + lz4_compress.c MAINHDRS = allocator.h pcompress.h utils.h MAINOBJS = $(MAINSRCS:.c=.o) @@ -44,6 +45,10 @@ LZFXSRCS = lzfx/lzfx.c LZFXHDRS = lzfx/lzfx.h LZFXOBJS = $(LZFXSRCS:.c=.o) +LZ4SRCS = lz4/lz4.c lz4/lz4hc.c +LZ4HDRS = lz4/lz4.h lz4/lz4hc.h +LZ4OBJS = $(LZ4SRCS:.c=.o) + PPMDSRCS = lzma/Ppmd8.c lzma/Ppmd8Enc.c lzma/Ppmd8Dec.c PPMDHDRS = lzma/Ppmd.h lzma/Ppmd8.h PPMDOBJS = $(PPMDSRCS:.c=.o) @@ -52,10 +57,10 @@ CRCSRCS = lzma/crc64_fast.c lzma/crc64_table.c CRCHDRS = lzma/crc64_table_le.h lzma/crc64_table_be.h lzma/crc_macros.h CRCOBJS = $(CRCSRCS:.c=.o) -BAKFILES = *~ lzma/*~ lzfx/*~ rabin/*~ bsdiff/*~ +BAKFILES = *~ lzma/*~ lzfx/*~ lz4/*~ rabin/*~ bsdiff/*~ RM = rm -f -CPPFLAGS = -I. -I./lzma -I./lzfx -I./rabin -I./bsdiff -D_7ZIP_ST -DNODEFAULT_PROPS \ +CPPFLAGS = -I. -I./lzma -I./lzfx -I./lz4 -I./rabin -I./bsdiff -D_7ZIP_ST -DNODEFAULT_PROPS \ -DFILE_OFFSET_BITS=64 -D_REENTRANT -D__USE_SSE_INTRIN__ -D_LZMA_PROB32 VEC_FLAGS = -ftree-vectorize LOOP_OPTFLAGS = $(VEC_FLAGS) -floop-interchange -floop-block @@ -97,16 +102,21 @@ $(BSDIFFOBJS): $(BSDIFFSRCS) $(BSDIFFHDRS) $(COMPILE) $(VEC_FLAGS) $(CPPFLAGS) $(@:.o=.c) -o $@ $(LZFXOBJS): $(LZFXSRCS) $(LZFXHDRS) - $(COMPILE) $(CPPFLAGS) $(@:.o=.c) -o $@ + $(COMPILE) $(VEC_FLAGS) $(CPPFLAGS) $(@:.o=.c) -o $@ + +$(LZ4OBJS): $(LZ4SRCS) $(LZ4HDRS) + $(COMPILE) $(VEC_FLAGS) $(CPPFLAGS) $(@:.o=.c) -o $@ $(MAINOBJS): $(MAINSRCS) $(MAINHDRS) $(COMPILE) $(LOOP_OPTFLAGS) $(CPPFLAGS) $(@:.o=.c) -o $@ -$(PROG): $(MAINOBJS) $(LZMAOBJS) $(PPMDOBJS) $(LZFXOBJS) $(CRCOBJS) $(RABINOBJS) $(BSDIFFOBJS) +$(PROG): $(MAINOBJS) $(LZMAOBJS) $(PPMDOBJS) $(LZFXOBJS) $(LZ4OBJS) \ +$(CRCOBJS) $(RABINOBJS) $(BSDIFFOBJS) $(LINK) -o $@ $(MAINOBJS) $(LZMAOBJS) $(PPMDOBJS) $(CRCOBJS) \ - $(LZFXOBJS) $(RABINOBJS) $(BSDIFFOBJS) $(LDLIBS) + $(LZFXOBJS) $(LZ4OBJS) $(RABINOBJS) $(BSDIFFOBJS) \ + $(LDLIBS) clean: - $(RM) $(PROG) $(MAINOBJS) $(LZMAOBJS) $(PPMDOBJS) $(CRCOBJS) $(LZFXOBJS) \ + $(RM) $(PROG) $(MAINOBJS) $(LZMAOBJS) $(PPMDOBJS) $(CRCOBJS) $(LZFXOBJS) $(LZ4OBJS) \ $(RABINOBJS) $(BSDIFFOBJS) $(BAKFILES) diff --git a/lz4/lz4.c b/lz4/lz4.c new file mode 100644 index 0000000..06e2829 --- /dev/null +++ b/lz4/lz4.c @@ -0,0 +1,819 @@ +/* + LZ4 - Fast LZ compression algorithm + Copyright (C) 2011-2012, Yann Collet. + BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions are + met: + + * Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above + copyright notice, this list of conditions and the following disclaimer + in the documentation and/or other materials provided with the + distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT + OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT + LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE + OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + + You can contact the author at : + - LZ4 homepage : http://fastcompression.blogspot.com/p/lz4.html + - LZ4 source repository : http://code.google.com/p/lz4/ +*/ + +//************************************** +// 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 : +// 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 + +// 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 option is useless on Little_Endian CPU (such as x86) +//#define BIG_ENDIAN_NATIVE_BUT_INCOMPATIBLE 1 + + + +//************************************** +// CPU Feature Detection +//************************************** +// 32 or 64 bits ? +#if (defined(__x86_64__) || defined(__x86_64) || defined(__amd64__) || defined(__amd64) || defined(__ppc64__) || defined(_WIN64) || defined(__LP64__) || defined(_LP64) ) // Detects 64 bits mode +# define LZ4_ARCH64 1 +#else +# define LZ4_ARCH64 0 +#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__))) ) +# 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 defined(__ARM_FEATURE_UNALIGNED) +# 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 + + +//************************************** +// Compiler Options +//************************************** +#if __STDC_VERSION__ >= 199901L // C99 +/* "restrict" is a known keyword */ +#else +# define restrict // Disable restrict +#endif + +#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 +# 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 +#endif + +#ifdef _MSC_VER +# define lz4_bswap16(x) _byteswap_ushort(x) +#else +# define lz4_bswap16(x) ((unsigned short int) ((((x) >> 8) & 0xffu) | (((x) & 0xffu) << 8))) +#endif + +#if (GCC_VERSION >= 302) || (__INTEL_COMPILER >= 800) || defined(__clang__) +# define expect(expr,value) (__builtin_expect ((expr),(value)) ) +#else +# define expect(expr,value) (expr) +#endif + +#define likely(expr) expect((expr) != 0, 1) +#define unlikely(expr) expect((expr) != 0, 0) + + +//************************************** +// Includes +//************************************** +#include // for malloc +#include // for memset +#include "lz4.h" + + +//************************************** +// Basic Types +//************************************** +#if defined(_MSC_VER) // Visual Studio does not support 'stdint' natively +# define BYTE unsigned __int8 +# define U16 unsigned __int16 +# define U32 unsigned __int32 +# define S32 __int32 +# define U64 unsigned __int64 +#else +# include +# define BYTE uint8_t +# define U16 uint16_t +# define U32 uint32_t +# define S32 int32_t +# define U64 uint64_t +#endif + +#ifndef LZ4_FORCE_UNALIGNED_ACCESS +# pragma pack(push, 1) +#endif + +typedef struct _U16_S { U16 v; } U16_S; +typedef struct _U32_S { U32 v; } U32_S; +typedef struct _U64_S { U64 v; } U64_S; + +#ifndef LZ4_FORCE_UNALIGNED_ACCESS +# pragma pack(pop) +#endif + +#define A64(x) (((U64_S *)(x))->v) +#define A32(x) (((U32_S *)(x))->v) +#define A16(x) (((U16_S *)(x))->v) + + +//************************************** +// Constants +//************************************** +#define MINMATCH 4 + +#define HASH_LOG COMPRESSIONLEVEL +#define HASHTABLESIZE (1 << HASH_LOG) +#define HASH_MASK (HASHTABLESIZE - 1) + +#define SKIPSTRENGTH (NOTCOMPRESSIBLE_CONFIRMATION>2?NOTCOMPRESSIBLE_CONFIRMATION: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 +#define LASTLITERALS 5 +#define MFLIMIT (COPYLENGTH+MINMATCH) +#define MINLENGTH (MFLIMIT+1) + +#define MAXD_LOG 16 +#define MAX_DISTANCE ((1 << MAXD_LOG) - 1) + +#define ML_BITS 4 +#define ML_MASK ((1U<> ((MINMATCH*8)-HASH_LOG)) +#define LZ4_HASH_VALUE(p) LZ4_HASH_FUNCTION(A32(p)) +#define LZ4_WILDCOPY(s,d,e) do { LZ4_COPYPACKET(s,d) } while (d>3); + #elif defined(__GNUC__) && (GCC_VERSION >= 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; + #endif +#else + #if defined(_MSC_VER) && !defined(LZ4_FORCE_SW_BITCOUNT) + unsigned long r = 0; + _BitScanForward64( &r, val ); + return (int)(r>>3); + #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]; + #endif +#endif +} + +#else + +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; + _BitScanReverse( &r, val ); + return (int)(r>>3); + #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; + #endif +#else + #if defined(_MSC_VER) && !defined(LZ4_FORCE_SW_BITCOUNT) + unsigned long r = 0; + _BitScanForward( &r, val ); + return (int)(r>>3); + #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]; + #endif +#endif +} + +#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) +{ +#if HEAPMODE + struct refTables *srt = (struct refTables *) (*ctx); + HTYPE* HashTable; +#else + 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; +#define matchlimit (iend - LASTLITERALS) + + BYTE* op = (BYTE*) dest; + + int len, length; + const int skipStrength = SKIPSTRENGTH; + U32 forwardH; + + + // Init + if (isizehashTable); + memset((void*)HashTable, 0, sizeof(srt->hashTable)); +#else + (void) ctx; +#endif + + + // First Byte + HashTable[LZ4_HASH_VALUE(ip)] = ip - base; + ip++; forwardH = LZ4_HASH_VALUE(ip); + + // 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<=(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; } + + // 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; } + + // 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<> ((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) +{ +#if HEAPMODE + struct refTables *srt = (struct refTables *) (*ctx); + U16* HashTable; +#else + 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; +#define matchlimit (iend - LASTLITERALS) + + BYTE* op = (BYTE*) dest; + + int len, length; + const int skipStrength = SKIPSTRENGTH; + U32 forwardH; + + + // Init + if (isizehashTable); + memset((void*)HashTable, 0, sizeof(srt->hashTable)); +#else + (void) ctx; +#endif + + + // First Byte + ip++; forwardH = LZ4_HASH64K_VALUE(ip); + + // 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<=(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; } + + // Fill table + HashTable[LZ4_HASH64K_VALUE(ip-2)] = 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; } + + // 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<>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-refoend-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 (ipoend-COPYLENGTH) + { + if (cpy > oend) goto _output_error; // Error : request to write outside of destination buffer + LZ4_SECURECOPY(ref, op, (oend-COPYLENGTH)); + while(op= 199901L // C99 + /* "restrict" is a known keyword */ +#else +#define restrict // Disable restrict +#endif + +#ifdef _MSC_VER +#define inline __forceinline // Visual is not C99, but supports some kind of inline +#endif + +#ifdef _MSC_VER // Visual Studio +#define bswap16(x) _byteswap_ushort(x) +#else +#define bswap16(x) ((unsigned short int) ((((x) >> 8) & 0xffu) | (((x) & 0xffu) << 8))) +#endif + + +//************************************** +// Includes +//************************************** +#include // calloc, free +#include // memset, memcpy +#include "lz4hc.h" + +#define ALLOCATOR(s) calloc(1,s) +#define FREEMEM free +#define MEM_INIT memset + + +//************************************** +// Basic Types +//************************************** +#if defined(_MSC_VER) // Visual Studio does not support 'stdint' natively +#define BYTE unsigned __int8 +#define U16 unsigned __int16 +#define U32 unsigned __int32 +#define S32 __int32 +#define U64 unsigned __int64 +#else +#include +#define BYTE uint8_t +#define U16 uint16_t +#define U32 uint32_t +#define S32 int32_t +#define U64 uint64_t +#endif + +#ifndef LZ4_FORCE_UNALIGNED_ACCESS +#pragma pack(push, 1) +#endif + +typedef struct _U16_S { U16 v; } U16_S; +typedef struct _U32_S { U32 v; } U32_S; +typedef struct _U64_S { U64 v; } U64_S; + +#ifndef LZ4_FORCE_UNALIGNED_ACCESS +#pragma pack(pop) +#endif + +#define A64(x) (((U64_S *)(x))->v) +#define A32(x) (((U32_S *)(x))->v) +#define A16(x) (((U16_S *)(x))->v) + + +//************************************** +// Constants +//************************************** +#define MINMATCH 4 + +#define DICTIONARY_LOGSIZE 16 +#define MAXD (1<> ((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; } + + +//************************************** +// Private functions +//************************************** +#if LZ4_ARCH64 + +inline static int LZ4_NbCommonBytes (register U64 val) +{ +#if defined(LZ4_BIG_ENDIAN) + #if defined(_MSC_VER) && !defined(LZ4_FORCE_SW_BITCOUNT) + unsigned long r = 0; + _BitScanReverse64( &r, val ); + return (int)(r>>3); + #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; + #endif +#else + #if defined(_MSC_VER) && !defined(LZ4_FORCE_SW_BITCOUNT) + unsigned long r = 0; + _BitScanForward64( &r, val ); + return (int)(r>>3); + #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]; + #endif +#endif +} + +#else + +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; + _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; + #endif +#else + #if defined(_MSC_VER) && !defined(LZ4_FORCE_SW_BITCOUNT) + unsigned long r = 0; + _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]; + #endif +#endif +} + +#endif + + +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; +} + + +inline static void* LZ4HC_Create (const BYTE* base) +{ + void* hc4 = ALLOCATOR(sizeof(LZ4HC_Data_Structure)); + + LZ4HC_Init (hc4, base); + return hc4; +} + + +inline static int LZ4HC_Free (void** LZ4HC_Data) +{ + FREEMEM(*LZ4HC_Data); + *LZ4HC_Data = NULL; + return (1); +} + + +inline 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++; + } +} + + +inline 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; + 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; +} + + +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) +{ + 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 (iptstartLimit) && (reft > hc4->base) && (startt[-1] == reft[-1])) {startt--; reft--;} + + if ((ipt-startt) > longest) + { + longest = ipt-startt; + *matchpos = reft; + *startpos = startt; + } + } + ref = GETNEXT(ref); + } + + return longest; +} + + +inline static int LZ4_encodeSequence(const BYTE** ip, BYTE** op, const BYTE** anchor, int ml, const BYTE* ref) +{ + 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)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; + + return 0; +} + + +//**************************** +// Compression CODE +//**************************** + +int LZ4_compressHCCtx(LZ4HC_Data_Structure* ctx, + 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; + +_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< +#include +#include +#include +#include +#include +#include +#include +#include +#include + +struct lz4_params { + int level; +}; + +void +lz4_stats(int show) +{ +} + +int +lz4_init(void **data, int *level, ssize_t chunksize) +{ + struct lz4_params *lzdat; + int lev; + + if (chunksize > INT_MAX) { + fprintf(stderr, "Chunk size too big for LZ4.\n"); + return (1); + } + lzdat = slab_alloc(NULL, sizeof (struct lz4_params)); + + lev = *level; + if (lev > 3) lev = 3; + lzdat->level = lev; + *data = lzdat; + + if (*level > 9) *level = 9; + return (0); +} + +int +lz4_deinit(void **data) +{ + struct lz4_params *lzdat = (struct lz4_params *)(*data); + + if (lzdat) { + slab_free(NULL, lzdat); + } + *data = NULL; + return (0); +} + +int +lz4_compress(void *src, size_t srclen, void *dst, size_t *dstlen, + int level, uchar_t chdr, void *data) +{ + int rv; + struct lz4_params *lzdat = (struct lz4_params *)data; + int _srclen = srclen; + uchar_t *dst2; + + if (lzdat->level == 1) { + rv = LZ4_compress(src, dst, _srclen); + + } else if (lzdat->level == 2) { + rv = LZ4_compress(src, dst, _srclen); + if (rv == 0) { + return (-1); + } + dst2 = slab_alloc(NULL, rv + sizeof (int)); + *((int *)dst2) = htonl(rv); + rv = LZ4_compressHC(dst, dst2 + sizeof (int), rv); + if (rv != 0) { + rv += sizeof (int); + memcpy(dst, dst2, rv); + } + slab_free(NULL, dst2); + } else { + rv = LZ4_compressHC(src, dst, _srclen); + } + if (rv == 0) { + return (-1); + } + *dstlen = rv; + + return (0); +} + +int +lz4_decompress(void *src, size_t srclen, void *dst, size_t *dstlen, + int level, uchar_t chdr, void *data) +{ + int rv; + struct lz4_params *lzdat = (struct lz4_params *)data; + int _dstlen = *dstlen; + uchar_t *dst2; + + if (lzdat->level == 1 || lzdat->level == 3) { + rv = LZ4_uncompress(src, dst, _dstlen); + + } else if (lzdat->level == 2) { + int sz1; + + sz1 = ntohl(*((int *)src)); + rv = LZ4_uncompress(src + sizeof (int), dst, sz1); + if (rv == 0) { + return (-1); + } + memcpy(src, dst, sz1); + rv = LZ4_uncompress(src, dst, _dstlen); + } + if (rv == 0) { + return (-1); + } + *dstlen = rv; + return (0); +} diff --git a/lzfx/lzfx.c b/lzfx/lzfx.c index e0f5a6d..5b036b4 100644 --- a/lzfx/lzfx.c +++ b/lzfx/lzfx.c @@ -50,9 +50,6 @@ * * moinakg@belenix.org, http://moinakg.wordpress.com/ * - * This is a somewhat modified bsdiff implementation. It has been modified - * to do buffer to buffer diffing instead of file to file and also use - * a custom RLE encoding rather than Bzip2 on the diff output. */ #include "lzfx.h" diff --git a/lzfx/lzfx.h b/lzfx/lzfx.h index 4858ec2..d2b7b07 100644 --- a/lzfx/lzfx.h +++ b/lzfx/lzfx.h @@ -49,9 +49,6 @@ * * moinakg@belenix.org, http://moinakg.wordpress.com/ * - * This is a somewhat modified bsdiff implementation. It has been modified - * to do buffer to buffer diffing instead of file to file and also use - * a custom RLE encoding rather than Bzip2 on the diff output. */ diff --git a/lzfx_compress.c b/lzfx_compress.c index 38517a8..9ddaf18 100644 --- a/lzfx_compress.c +++ b/lzfx_compress.c @@ -72,7 +72,7 @@ lz_fx_deinit(void **data) return (0); } -void +static void lz_fx_err(int err) { switch (err) { diff --git a/main.c b/main.c index 776ce4a..9d562cf 100644 --- a/main.c +++ b/main.c @@ -1199,6 +1199,14 @@ init_algo(const char *algo, int bail) _stats_func = lz_fx_stats; rv = 0; + } else if (memcmp(algorithm, "lz4", 3) == 0) { + _compress_func = lz4_compress; + _decompress_func = lz4_decompress; + _init_func = lz4_init; + _deinit_func = lz4_deinit; + _stats_func = lz4_stats; + rv = 0; + /* adapt2 and adapt ordering of the checks matter here. */ } else if (memcmp(algorithm, "adapt2", 6) == 0) { _compress_func = adapt_compress; diff --git a/pcompress.h b/pcompress.h index 4f6e114..1eaa7f2 100644 --- a/pcompress.h +++ b/pcompress.h @@ -67,6 +67,8 @@ extern int ppmd_compress(void *src, size_t srclen, void *dst, size_t *dstlen, int level, uchar_t chdr, void *data); extern int lz_fx_compress(void *src, size_t srclen, void *dst, size_t *dstlen, int level, uchar_t chdr, void *data); +extern int lz4_compress(void *src, size_t srclen, void *dst, + size_t *dstlen, int level, uchar_t chdr, void *data); extern int zlib_decompress(void *src, size_t srclen, void *dst, size_t *dstlen, int level, uchar_t chdr, void *data); @@ -80,6 +82,8 @@ extern int ppmd_decompress(void *src, size_t srclen, void *dst, size_t *dstlen, int level, uchar_t chdr, void *data); extern int lz_fx_decompress(void *src, size_t srclen, void *dst, size_t *dstlen, int level, uchar_t chdr, void *data); +extern int lz4_decompress(void *src, size_t srclen, void *dst, + size_t *dstlen, int level, uchar_t chdr, void *data); extern int adapt_init(void **data, int *level, ssize_t chunksize); extern int adapt2_init(void **data, int *level, ssize_t chunksize); @@ -88,11 +92,13 @@ extern int ppmd_init(void **data, int *level, ssize_t chunksize); extern int bzip2_init(void **data, int *level, ssize_t chunksize); extern int zlib_init(void **data, int *level, ssize_t chunksize); extern int lz_fx_init(void **data, int *level, ssize_t chunksize); +extern int lz4_init(void **data, int *level, ssize_t chunksize); extern int adapt_deinit(void **data); extern int lzma_deinit(void **data); extern int ppmd_deinit(void **data); extern int lz_fx_deinit(void **data); +extern int lz4_deinit(void **data); extern void adapt_stats(int show); extern void ppmd_stats(int show); @@ -100,6 +106,7 @@ extern void lzma_stats(int show); extern void bzip2_stats(int show); extern void zlib_stats(int show); extern void lz_fx_stats(int show); +extern void lz4_stats(int show); /* * Per-thread data structure for compression and decompression threads. diff --git a/utils.h b/utils.h index bbf7e58..3f3f11f 100644 --- a/utils.h +++ b/utils.h @@ -103,6 +103,8 @@ extern const char *get_execname(const char *); extern int parse_numeric(ssize_t *val, const char *str); extern char *bytes_to_size(uint64_t bytes); extern ssize_t Read(int fd, void *buf, size_t count); +extern ssize_t Read_Adjusted(int fd, uchar_t *buf, size_t count, + ssize_t *rabin_count, void *ctx); extern ssize_t Write(int fd, const void *buf, size_t count); /* Pointer type for compress and decompress functions. */