New experimental mode : compress blocks using data from previous blocks (option -BD) (limitation : -hc mode only)

Changed deprecated function names to "static", in order to avoid duplicate definition. Thanks Maciej Adamczyk for reporting.
Changed a few command line options
Prettify help text


git-svn-id: https://lz4.googlecode.com/svn/trunk@96 650e7d94-2a16-8b24-b05c-7c0b3f6821cd
This commit is contained in:
yann.collet.73@gmail.com 2013-05-27 13:37:48 +00:00
parent e185b252f0
commit cd3bcd0043
8 changed files with 415 additions and 121 deletions

View File

@ -32,7 +32,7 @@
// Unix Large Files support (>4GB)
#if (defined(__sun__) && (!defined(__LP64__))) // Sun Solaris 32-bits requires specific definitions
# define _LARGEFILE_SOURCE
# define FILE_OFFSET_BITS=64
# define _FILE_OFFSET_BITS 64
#elif ! defined(__LP64__) // No point defining Large file for 64 bit
# define _LARGEFILE64_SOURCE
#endif

View File

@ -99,7 +99,7 @@ int FUZ_SecurityTest()
char* input;
int i, r;
printf("Starting overflow tests (issue 52)...");
printf("Overflow test (issue 52)...");
input = (char*) malloc (20<<20);
output = (char*) malloc (20<<20);
input[0] = 0x0F;

15
lz4.h
View File

@ -135,12 +135,23 @@ LZ4_decompress_safe_partial() :
*/
int LZ4_decompress_safe_withPrefix64k (const char* source, char* dest, int inputSize, int maxOutputSize);
int LZ4_decompress_fast_withPrefix64k (const char* source, char* dest, int outputSize);
/*
*_withPrefix64k() :
These decoding functions work the same as their "normal name" versions,
but will potentially use up to 64KB of data in front of 'char* dest'.
These functions are used for decoding inter-dependant blocks.
*/
//****************************
// Obsolete Functions
//****************************
inline int LZ4_uncompress (const char* source, char* dest, int outputSize) { return LZ4_decompress_fast(source, dest, outputSize); }
inline int LZ4_uncompress_unknownOutputSize (const char* source, char* dest, int isize, int maxOutputSize) { return LZ4_decompress_safe(source, dest, isize, maxOutputSize); }
static inline int LZ4_uncompress (const char* source, char* dest, int outputSize) { return LZ4_decompress_fast(source, dest, outputSize); }
static inline int LZ4_uncompress_unknownOutputSize (const char* source, char* dest, int isize, int maxOutputSize) { return LZ4_decompress_safe(source, dest, isize, maxOutputSize); }
/*
These functions are deprecated and should no longer be used.

321
lz4c.c
View File

@ -86,24 +86,31 @@
#define EXTENSION ".lz4"
#define WELCOME_MESSAGE "*** %s %s, by %s (%s) ***\n", COMPRESSOR_NAME, COMPRESSOR_VERSION, AUTHOR, COMPILED
#define CACHELINE 64
#define MAGICNUMBER_SIZE 4
#define LZ4S_MAGICNUMBER 0x184D2204
#define LZ4S_BLOCKSIZEID_DEFAULT 7
#define LZ4S_CHECKSUM_SEED 0
#define KB *(1U<<10)
#define MB *(1U<<20)
#define GB *(1U<<30)
#define LZ4S_SKIPPABLE0 0x184D2A50
#define LZ4S_SKIPPABLEMASK 0xFFFFFFF0
#define LEGACY_MAGICNUMBER 0x184C2102
#define LEGACY_BLOCKSIZE (8<<20) // 8 MB
#define _1BIT 1
#define _1BIT 0x01
#define _2BITS 0x03
#define _3BITS 0x07
#define _4BITS 0x0F
#define _8BITS 0xFF
#define MAGICNUMBER_SIZE 4
#define LZ4S_MAGICNUMBER 0x184D2204
#define LZ4S_SKIPPABLE0 0x184D2A50
#define LZ4S_SKIPPABLEMASK 0xFFFFFFF0
#define LEGACY_MAGICNUMBER 0x184C2102
#define CACHELINE 64
#define LEGACY_BLOCKSIZE (8 MB)
#define MIN_STREAM_BUFSIZE (1 MB + 64 KB)
#define LZ4S_BLOCKSIZEID_DEFAULT 7
#define LZ4S_CHECKSUM_SEED 0
#define LZ4S_EOS 0
#define LZ4S_MAXHEADERSIZE (4+2+8+4+1)
//**************************************
// Architecture Macros
//**************************************
@ -139,6 +146,7 @@ static int overwrite = 0;
static int blockSizeId = LZ4S_BLOCKSIZEID_DEFAULT;
static int blockChecksum = 0;
static int streamChecksum = 1;
static int blockIndependence = 1;
//**************************************
// Exceptions
@ -177,17 +185,18 @@ int usage_advanced()
DISPLAY( "\nAdvanced options :\n");
DISPLAY( " -t : test compressed file \n");
DISPLAY( " -B# : Block size [4-7](default : 7)\n");
DISPLAY( " -x : enable block checksum (default:disabled)\n");
DISPLAY( " -nx : disable stream checksum (default:enabled)\n");
DISPLAY( " -BD : Block dependency (improve compression ratio)\n");
DISPLAY( " -BX : enable block checksum (default:disabled)\n");
DISPLAY( " -Sx : disable stream checksum (default:enabled)\n");
DISPLAY( " -b# : benchmark files, using # [0-1] compression level\n");
DISPLAY( " -i# : iteration loops [1-9](default : 3), benchmark mode only\n");
DISPLAY( "input : can be 'stdin' (pipe) or a filename\n");
DISPLAY( "output : can be 'stdout'(pipe) or a filename or 'null'\n");
DISPLAY( " example : lz4c -hc stdin compressedfile.lz4");
DISPLAY( " example 1 : lz4c -hc stdin compressedfile.lz4\n");
DISPLAY( " example 2 : lz4c -hcyB4D filename \n");
return 0;
}
int badusage(char* exename)
{
DISPLAY("Wrong parameters\n");
@ -196,8 +205,9 @@ int badusage(char* exename)
}
static int LZ4S_GetBlocksize_FromBlockId (int id) { return (1 << (8 + (2 * id))); }
static int LZ4S_GetBlockSize_FromBlockId (int id) { return (1 << (8 + (2 * id))); }
static unsigned int LZ4S_GetCheckBits_FromXXH (unsigned int xxh) { return (xxh >> 8) & _8BITS; }
static int LZ4S_isSkippableMagicNumber(unsigned int magic) { return (magic & LZ4S_SKIPPABLEMASK) == LZ4S_SKIPPABLE0; }
int get_fileHandle(char* input_filename, char* output_filename, FILE** pfinput, FILE** pfoutput)
@ -271,7 +281,7 @@ int legacy_compress_file(char* input_filename, char* output_filename, int compre
{
case 0 : compressionFunction = LZ4_compress; break;
case 1 : compressionFunction = LZ4_compressHC; break;
default : compressionFunction = LZ4_compress;
default: compressionFunction = LZ4_compress;
}
start = clock();
get_fileHandle(input_filename, output_filename, &finput, &foutput);
@ -326,6 +336,157 @@ int legacy_compress_file(char* input_filename, char* output_filename, int compre
}
int compress_file_blockDependency(char* input_filename, char* output_filename, int compressionlevel)
{
void* (*initFunction) (const char*);
int (*compressionFunction)(void*, const char*, char*, int, int);
char* (*translateFunction) (void*);
int (*freeFunction) (void*);
void* ctx;
unsigned long long filesize = 0;
unsigned long long compressedfilesize = 0;
unsigned int checkbits;
char* in_buff, *in_start, *in_end;
char* out_buff;
FILE* finput;
FILE* foutput;
int errorcode;
int displayLevel = (compressionlevel>0);
clock_t start, end;
unsigned int blockSize, inputBufferSize;
size_t sizeCheck, header_size;
void* streamChecksumState=NULL;
// Init
start = clock();
switch (compressionlevel)
{
case 0 :
case 1 :
default:
initFunction = LZ4_createHC;
compressionFunction = LZ4_compressHC_limitedOutput_continue;
translateFunction = LZ4_slideInputBufferHC;
freeFunction = LZ4_freeHC;
}
errorcode = get_fileHandle(input_filename, output_filename, &finput, &foutput);
if (errorcode) return errorcode;
blockSize = LZ4S_GetBlockSize_FromBlockId (blockSizeId);
// Allocate Memory
inputBufferSize = blockSize + 64 KB;
if (inputBufferSize < MIN_STREAM_BUFSIZE) inputBufferSize = MIN_STREAM_BUFSIZE;
in_buff = (char*)malloc(inputBufferSize);
out_buff = (char*)malloc(blockSize+CACHELINE);
if (!in_buff || !out_buff) EXM_THROW(31, "Allocation error : not enough memory");
in_start = in_buff; in_end = in_buff + inputBufferSize;
if (streamChecksum) streamChecksumState = XXH32_init(LZ4S_CHECKSUM_SEED);
ctx = initFunction(in_buff);
// Write Archive Header
*(unsigned int*)out_buff = LITTLE_ENDIAN_32(LZ4S_MAGICNUMBER); // Magic Number, in Little Endian convention
*(out_buff+4) = (1 & _2BITS) << 6 ; // Version('01')
*(out_buff+4) |= (blockIndependence & _1BIT) << 5;
*(out_buff+4) |= (blockChecksum & _1BIT) << 4;
*(out_buff+4) |= (streamChecksum & _1BIT) << 2;
*(out_buff+5) = (blockSizeId & _3BITS) << 4;
checkbits = XXH32((out_buff+4), 2, LZ4S_CHECKSUM_SEED);
checkbits = LZ4S_GetCheckBits_FromXXH(checkbits);
*(out_buff+6) = (unsigned char) checkbits;
header_size = 7;
sizeCheck = fwrite(out_buff, 1, header_size, foutput);
if (sizeCheck!=header_size) EXM_THROW(32, "Write error : cannot write header");
compressedfilesize += header_size;
// Main Loop
while (1)
{
unsigned int outSize;
unsigned int inSize;
// Read Block
if ((in_start+blockSize) > in_end) in_start = translateFunction(ctx);
inSize = (unsigned int) fread(in_start, (size_t)1, (size_t)blockSize, finput);
if( inSize<=0 ) break; // No more input : end of compression
filesize += inSize;
if (displayLevel) DISPLAY("Read : %i MB \r", (int)(filesize>>20));
if (streamChecksum) XXH32_update(streamChecksumState, in_start, inSize);
// Compress Block
outSize = compressionFunction(ctx, in_start, out_buff+4, inSize, inSize-1);
if (outSize > 0) compressedfilesize += outSize+4; else compressedfilesize += inSize+4;
if (blockChecksum) compressedfilesize+=4;
if (displayLevel) DISPLAY("Read : %i MB ==> %.2f%%\r", (int)(filesize>>20), (double)compressedfilesize/filesize*100);
// Write Block
if (outSize > 0)
{
unsigned int checksum;
int sizeToWrite;
* (unsigned int*) out_buff = LITTLE_ENDIAN_32(outSize);
if (blockChecksum)
{
checksum = XXH32(out_buff+4, outSize, LZ4S_CHECKSUM_SEED);
* (unsigned int*) (out_buff+4+outSize) = LITTLE_ENDIAN_32(checksum);
}
sizeToWrite = 4 + outSize + (4*blockChecksum);
sizeCheck = fwrite(out_buff, 1, sizeToWrite, foutput);
if (sizeCheck!=(size_t)(sizeToWrite)) EXM_THROW(33, "Write error : cannot write compressed block");
}
else // Copy Original
{
unsigned int checksum;
* (unsigned int*) out_buff = LITTLE_ENDIAN_32(inSize|0x80000000); // Add Uncompressed flag
sizeCheck = fwrite(out_buff, 1, 4, foutput);
if (sizeCheck!=(size_t)(4)) EXM_THROW(34, "Write error : cannot write block header");
sizeCheck = fwrite(in_start, 1, inSize, foutput);
if (sizeCheck!=(size_t)(inSize)) EXM_THROW(35, "Write error : cannot write block");
if (blockChecksum)
{
checksum = XXH32(in_start, inSize, LZ4S_CHECKSUM_SEED);
* (unsigned int*) out_buff = LITTLE_ENDIAN_32(checksum);
sizeCheck = fwrite(out_buff, 1, 4, foutput);
if (sizeCheck!=(size_t)(4)) EXM_THROW(36, "Write error : cannot write block checksum");
}
}
in_start += inSize;
}
// End of Stream mark
* (unsigned int*) out_buff = LZ4S_EOS;
sizeCheck = fwrite(out_buff, 1, 4, foutput);
if (sizeCheck!=(size_t)(4)) EXM_THROW(37, "Write error : cannot write end of stream");
compressedfilesize += 4;
if (streamChecksum)
{
unsigned int checksum = XXH32_digest(streamChecksumState);
* (unsigned int*) out_buff = LITTLE_ENDIAN_32(checksum);
sizeCheck = fwrite(out_buff, 1, 4, foutput);
if (sizeCheck!=(size_t)(4)) EXM_THROW(37, "Write error : cannot write stream checksum");
compressedfilesize += 4;
}
// Status
end = clock();
DISPLAY( "Compressed %llu bytes into %llu bytes ==> %.2f%%\n",
(unsigned long long) filesize, (unsigned long long) compressedfilesize, (double)compressedfilesize/filesize*100);
{
double seconds = (double)(end - start)/CLOCKS_PER_SEC;
DISPLAY( "Done in %.2f s ==> %.2f MB/s\n", seconds, (double)filesize / seconds / 1024 / 1024);
}
// Close & Free
freeFunction(ctx);
free(in_buff);
free(out_buff);
fclose(finput);
fclose(foutput);
return 0;
}
int compress_file(char* input_filename, char* output_filename, int compressionlevel)
{
int (*compressionFunction)(const char*, char*, int, int);
@ -343,6 +504,8 @@ int compress_file(char* input_filename, char* output_filename, int compressionle
size_t sizeCheck, header_size;
void* streamChecksumState=NULL;
// Branch out
if (blockIndependence==0) return compress_file_blockDependency(input_filename, output_filename, compressionlevel);
// Init
start = clock();
@ -350,11 +513,11 @@ int compress_file(char* input_filename, char* output_filename, int compressionle
{
case 0 : compressionFunction = LZ4_compress_limitedOutput; break;
case 1 : compressionFunction = LZ4_compressHC_limitedOutput; break;
default : compressionFunction = LZ4_compress_limitedOutput;
default: compressionFunction = LZ4_compress_limitedOutput;
}
errorcode = get_fileHandle(input_filename, output_filename, &finput, &foutput);
if (errorcode) return errorcode;
blockSize = LZ4S_GetBlocksize_FromBlockId (blockSizeId);
blockSize = LZ4S_GetBlockSize_FromBlockId (blockSizeId);
// Allocate Memory
in_buff = (char*)malloc(blockSize);
@ -364,10 +527,11 @@ int compress_file(char* input_filename, char* output_filename, int compressionle
// Write Archive Header
*(unsigned int*)out_buff = LITTLE_ENDIAN_32(LZ4S_MAGICNUMBER); // Magic Number, in Little Endian convention
*(out_buff+4) = 0x60; // Version("01"), Block independence
*(out_buff+4) |= blockChecksum << 4;
*(out_buff+4) |= streamChecksum << 2;
*(out_buff+5) = (char)(blockSizeId<<4); // Block Size
*(out_buff+4) = (1 & _2BITS) << 6 ; // Version('01')
*(out_buff+4) |= (blockIndependence & _1BIT) << 5;
*(out_buff+4) |= (blockChecksum & _1BIT) << 4;
*(out_buff+4) |= (streamChecksum & _1BIT) << 2;
*(out_buff+5) = (blockSizeId & _3BITS) <<4;
checkbits = XXH32((out_buff+4), 2, LZ4S_CHECKSUM_SEED);
checkbits = LZ4S_GetCheckBits_FromXXH(checkbits);
*(out_buff+6) = (unsigned char) checkbits;
@ -382,7 +546,7 @@ int compress_file(char* input_filename, char* output_filename, int compressionle
unsigned int outSize;
// Read Block
unsigned int inSize = (unsigned int) fread(in_buff, (size_t)1, (size_t)blockSize, finput);
if( inSize<=0 ) break;
if( inSize<=0 ) break; // No more input : end of compression
filesize += inSize;
if (displayLevel) DISPLAY("Read : %i MB \r", (int)(filesize>>20));
if (streamChecksum) XXH32_update(streamChecksumState, in_buff, inSize);
@ -428,7 +592,7 @@ int compress_file(char* input_filename, char* output_filename, int compressionle
}
// End of Stream mark
* (unsigned int*) out_buff = 0;
* (unsigned int*) out_buff = LZ4S_EOS;
sizeCheck = fwrite(out_buff, 1, 4, foutput);
if (sizeCheck!=(size_t)(4)) EXM_THROW(37, "Write error : cannot write end of stream");
compressedfilesize += 4;
@ -499,7 +663,7 @@ unsigned long long decodeLegacyStream(FILE* finput, FILE* foutput)
// Write Block
sizeCheck = fwrite(out_buff, 1, sinkint, foutput);
if (sizeCheck != (size_t)sinkint) EXM_THROW(53, "Write error : cannot write decoded block\n");
if (sizeCheck != (size_t)sinkint) EXM_THROW(53, "Write error : cannot write decoded block into output\n");
}
// Free
@ -514,21 +678,22 @@ unsigned long long decodeLZ4S(FILE* finput, FILE* foutput)
{
unsigned long long filesize = 0;
char* in_buff;
char* out_buff;
unsigned char descriptor[3];
char* out_buff, *out_start, *out_end;
unsigned char descriptor[LZ4S_MAXHEADERSIZE];
size_t nbReadBytes;
int decodedBytes;
int decodedBytes=0;
unsigned int maxBlockSize;
size_t sizeCheck;
int blockChecksumFlag, streamChecksumFlag;
int blockChecksumFlag, streamChecksumFlag, blockIndependenceFlag;
void* streamChecksumState=NULL;
int (*decompressionFunction)(const char*, char*, int, int) = LZ4_decompress_safe;
unsigned int prefix64k = 0;
// Decode stream descriptor
nbReadBytes = fread(descriptor, 1, 3, finput);
if (nbReadBytes != 3) EXM_THROW(61, "Unreadable header");
{
int version = (descriptor[0] >> 6) & _2BITS;
int independance = (descriptor[0] >> 5) & _1BIT;
int streamSize = (descriptor[0] >> 3) & _1BIT;
int reserved1 = (descriptor[0] >> 1) & _1BIT;
int dictionary = (descriptor[0] >> 0) & _1BIT;
@ -539,18 +704,18 @@ unsigned long long decodeLZ4S(FILE* finput, FILE* foutput)
int checkBits = (descriptor[2] >> 0) & _8BITS;
int checkBits_xxh32;
blockIndependenceFlag=(descriptor[0] >> 5) & _1BIT;
blockChecksumFlag = (descriptor[0] >> 4) & _1BIT;
streamChecksumFlag= (descriptor[0] >> 2) & _1BIT;
if (version != 1) EXM_THROW(62, "Wrong version number");
if (independance != 1) EXM_THROW(63, "Does not support block inter-dependence");
if (streamSize == 1) EXM_THROW(64, "Does not support stream size");
if (reserved1 != 0) EXM_THROW(65, "Wrong value for reserved bits");
if (dictionary == 1) EXM_THROW(66, "Does not support dictionary");
if (reserved2 != 0) EXM_THROW(67, "Wrong value for reserved bits");
if (blockSizeId < 4) EXM_THROW(68, "Unsupported block size");
if (reserved3 != 0) EXM_THROW(67, "Wrong value for reserved bits");
maxBlockSize = LZ4S_GetBlocksize_FromBlockId(blockSizeId);
maxBlockSize = LZ4S_GetBlockSize_FromBlockId(blockSizeId);
// Checkbits verification
descriptor[1] &= 0xF0;
checkBits_xxh32 = XXH32(descriptor, 2, LZ4S_CHECKSUM_SEED);
@ -558,10 +723,22 @@ unsigned long long decodeLZ4S(FILE* finput, FILE* foutput)
if (checkBits != checkBits_xxh32) EXM_THROW(69, "Stream descriptor error detected");
}
if (!blockIndependenceFlag)
{
decompressionFunction = LZ4_decompress_safe_withPrefix64k;
prefix64k = 64 KB;
}
// Allocate Memory
{
unsigned int outbuffSize = prefix64k+maxBlockSize;
in_buff = (char*)malloc(maxBlockSize);
out_buff = (char*)malloc(maxBlockSize);
if (outbuffSize < MIN_STREAM_BUFSIZE) outbuffSize = MIN_STREAM_BUFSIZE;
out_buff = (char*)malloc(outbuffSize);
out_end = out_buff + outbuffSize;
out_start = out_buff + prefix64k;
if (!in_buff || !out_buff) EXM_THROW(70, "Allocation error : not enough memory");
}
if (streamChecksumFlag) streamChecksumState = XXH32_init(LZ4S_CHECKSUM_SEED);
// Main Loop
@ -572,7 +749,7 @@ unsigned long long decodeLZ4S(FILE* finput, FILE* foutput)
// Block Size
nbReadBytes = fread(&blockSize, 1, 4, finput);
if( nbReadBytes != 4 ) EXM_THROW(71, "Read error : cannot read next block size");
if (blockSize == 0) break; // End of Stream Mark : stream is completed
if (blockSize == LZ4S_EOS) break; // End of Stream Mark : stream is completed
blockSize = LITTLE_ENDIAN_32(blockSize); // Convert to little endian
uncompressedFlag = blockSize >> 31;
blockSize &= 0x7FFFFFFF;
@ -600,19 +777,42 @@ unsigned long long decodeLZ4S(FILE* finput, FILE* foutput)
if (sizeCheck != (size_t)blockSize) EXM_THROW(76, "Write error : cannot write data block");
filesize += blockSize;
if (streamChecksumFlag) XXH32_update(streamChecksumState, in_buff, blockSize);
if (!blockIndependenceFlag)
{
if (blockSize >= prefix64k)
{
memcpy(out_buff, in_buff + (blockSize - prefix64k), prefix64k); // Required for reference for next blocks
out_start = out_buff + prefix64k;
continue;
}
else
{
memcpy(out_start, in_buff, blockSize);
}
}
}
else
{
// Decode Block
decodedBytes = LZ4_decompress_safe(in_buff, out_buff, blockSize, maxBlockSize);
decodedBytes = decompressionFunction(in_buff, out_start, blockSize, maxBlockSize);
if (decodedBytes < 0) EXM_THROW(77, "Decoding Failed ! Corrupted input detected !");
filesize += decodedBytes;
if (streamChecksumFlag) XXH32_update(streamChecksumState, out_buff, decodedBytes);
if (streamChecksumFlag) XXH32_update(streamChecksumState, out_start, decodedBytes);
// Write Block
sizeCheck = fwrite(out_buff, 1, decodedBytes, foutput);
sizeCheck = fwrite(out_start, 1, decodedBytes, foutput);
if (sizeCheck != (size_t)decodedBytes) EXM_THROW(78, "Write error : cannot write decoded block\n");
}
if (!blockIndependenceFlag)
{
out_start += decodedBytes;
if (out_start + maxBlockSize > out_end)
{
memcpy(out_buff, out_start - prefix64k, prefix64k);
out_start = out_buff + prefix64k;
}
}
}
// Stream Checksum
@ -645,7 +845,7 @@ unsigned long long selectDecoder( FILE* finput, FILE* foutput)
if (nbReadBytes==0) return 0; // EOF
if (nbReadBytes != MAGICNUMBER_SIZE) EXM_THROW(41, "Unrecognized header : Magic Number unreadable");
magicNumber = LITTLE_ENDIAN_32(magicNumber); // Convert to Little Endian format
if ((magicNumber & LZ4S_SKIPPABLEMASK) == LZ4S_SKIPPABLE0) magicNumber = LZ4S_SKIPPABLE0; // fold skippable magic numbers
if (LZ4S_isSkippableMagicNumber(magicNumber)) magicNumber = LZ4S_SKIPPABLE0; // fold skippable magic numbers
switch(magicNumber)
{
@ -746,31 +946,45 @@ int main(int argc, char** argv)
case 'c': if ((argument[1] >='0') && (argument[1] <='1')) { cLevel=argument[1] - '0'; argument++; } break;
case 'h': if (argument[1]=='c') { cLevel=1; argument++; } break;
// Checksum management
case 'x': blockChecksum=1; break;
case 'n': if (argument[1]=='x') { streamChecksum=0; argument++; break; } else { badusage(exename); return 1; }
// Use Legacy format (hidden option)
case 'l': legacy_format=1; break;
// Decoding
case 'd': decode=1; break;
// Bench
case 'b': bench=1;
if ((argument[1] >='0') && (argument[1] <='1')) { cLevel=argument[1] - '0'; argument++; }
break;
// Test
case 't': decode=1; output_filename=nulmark; break;
// Modify Block Size
// Modify Block Properties
case 'B':
if ((argument[1] >='4') && (argument[1] <='7'))
while (argument[1]!=0)
switch(argument[1])
{
case '4':
case '5':
case '6':
case '7':
{
int B = argument[1] - '0';
int S = 1 << (8 + 2*B);
BMK_SetBlocksize(S);
blockSizeId = B;
argument++;
break;
}
case 'D': blockIndependence = 0, argument++; break;
case 'X': blockChecksum = 1, argument ++; break;
default : goto _exit_blockProperties;
}
_exit_blockProperties:
break;
// Modify Stream properties
case 'S': if (argument[1]=='x') { streamChecksum=0; argument++; break; } else { badusage(exename); return 1; }
// Bench
case 'b': bench=1;
if ((argument[1] >='0') && (argument[1] <='1')) { cLevel=argument[1] - '0'; argument++; }
break;
// Modify Nb Iterations (benchmark only)
@ -786,9 +1000,6 @@ int main(int argc, char** argv)
// Pause at the end (benchmark only) (hidden option)
case 'p': BMK_SetPause(); break;
// Test
case 't': decode=1; output_filename=nulmark; break;
// Overwrite
case 'y': overwrite=1; break;
@ -816,7 +1027,7 @@ int main(int argc, char** argv)
if (bench) return BMK_benchFile(argv+filenamesStart, argc-filenamesStart, cLevel);
// No output filename ==> build one automatically (for compression only)
// No output filename ==> build one automatically (when possible)
if (!output_filename)
{
if (!decode) // compression
@ -827,7 +1038,7 @@ int main(int argc, char** argv)
for (i=0;i<l;i++) output_filename[i] = input_filename[i];
for (i=l;i<l+4;i++) output_filename[i] = extension[i-l];
}
else
else // decompression (input file must respect format extension ".lz4")
{
int inl=0,outl;
while (input_filename[inl]!=0) inl++;

69
lz4hc.c
View File

@ -201,6 +201,9 @@ typedef struct _U64_S { U64 v; } _PACKED U64_S;
#define MINLENGTH (MFLIMIT+1)
#define OPTIMAL_ML (int)((ML_MASK-1)+MINMATCH)
#define KB *(1U<<10)
#define MB *(1U<<20)
#define GB *(1U<<30)
//**************************************
// Architecture-specific macros
@ -219,8 +222,10 @@ typedef struct _U64_S { U64 v; } _PACKED U64_S;
# define LZ4_COPYPACKET(s,d) LZ4_COPYSTEP(s,d); LZ4_COPYSTEP(s,d);
# define UARCH U32
# define AARCH A32
# define HTYPE const BYTE*
# define INITBASE(b,s) const int b = 0
//# define HTYPE const BYTE*
//# define INITBASE(b,s) const int b = 0
# define HTYPE U32
# define INITBASE(b,s) const BYTE* const b = s
#endif
#if defined(LZ4_BIG_ENDIAN)
@ -237,7 +242,9 @@ typedef struct _U64_S { U64 v; } _PACKED U64_S;
//************************************************************
typedef struct
{
const BYTE* inputBuffer;
const BYTE* base;
const BYTE* end;
HTYPE hashTable[HASHTABLESIZE];
U16 chainTable[MAXD];
const BYTE* nextToUpdate;
@ -325,35 +332,35 @@ inline static int LZ4_NbCommonBytes (register U32 val)
#endif
inline static int LZ4HC_Init (LZ4HC_Data_Structure* hc4, const BYTE* base)
static inline int LZ4_InitHC (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->nextToUpdate = base + 1;
hc4->base = base;
hc4->inputBuffer = base;
hc4->end = base;
return 1;
}
inline static void* LZ4HC_create (const BYTE* base)
extern inline void* LZ4_createHC (const char* slidingInputBuffer)
{
void* hc4 = ALLOCATOR(sizeof(LZ4HC_Data_Structure));
LZ4HC_Init ((LZ4HC_Data_Structure*)hc4, base);
LZ4_InitHC ((LZ4HC_Data_Structure*)hc4, (const BYTE*)slidingInputBuffer);
return hc4;
}
inline static int LZ4HC_free (void** LZ4HC_Data)
extern inline int LZ4_freeHC (void* LZ4HC_Data)
{
FREEMEM(*LZ4HC_Data);
*LZ4HC_Data = NULL;
return (1);
FREEMEM(LZ4HC_Data);
return (0);
}
// Update chains up to ip (excluded)
forceinline static void LZ4HC_Insert (LZ4HC_Data_Structure* hc4, const BYTE* ip)
static forceinline void LZ4HC_Insert (LZ4HC_Data_Structure* hc4, const BYTE* ip)
{
U16* chainTable = hc4->chainTable;
HTYPE* HashTable = hc4->hashTable;
@ -361,7 +368,7 @@ forceinline static void LZ4HC_Insert (LZ4HC_Data_Structure* hc4, const BYTE* ip)
while(hc4->nextToUpdate < ip)
{
const BYTE* p = hc4->nextToUpdate;
const BYTE* const p = hc4->nextToUpdate;
size_t delta = (p) - HASH_POINTER(p);
if (delta>MAX_DISTANCE) delta = MAX_DISTANCE;
DELTANEXT(p) = (U16)delta;
@ -371,7 +378,27 @@ forceinline static void LZ4HC_Insert (LZ4HC_Data_Structure* hc4, const BYTE* ip)
}
forceinline static size_t LZ4HC_CommonLength (const BYTE* p1, const BYTE* p2, const BYTE* const matchlimit)
char* LZ4_slideInputBufferHC(void* LZ4HC_Data)
{
LZ4HC_Data_Structure* hc4 = (LZ4HC_Data_Structure*)LZ4HC_Data;
U32 distance = (U32)(hc4->end - hc4->inputBuffer) - 64 KB;
distance = (distance >> 16) << 16; // Must be a multiple of 64 KB
LZ4HC_Insert(hc4, hc4->end - MINMATCH);
memcpy((void*)(hc4->end - 64 KB - distance), (const void*)(hc4->end - 64 KB), 64 KB);
hc4->nextToUpdate -= distance;
hc4->base -= distance;
if ((U32)(hc4->inputBuffer - hc4->base) > 1 GB + 64 KB) // Avoid overflow
{
int i;
hc4->base += 1 GB;
for (i=0; i<HASHTABLESIZE; i++) hc4->hashTable[i] -= 1 GB;
}
hc4->end -= distance;
return (char*)(hc4->end);
}
static forceinline size_t LZ4HC_CommonLength (const BYTE* p1, const BYTE* p2, const BYTE* const matchlimit)
{
const BYTE* p1t = p1;
@ -389,7 +416,7 @@ forceinline static size_t LZ4HC_CommonLength (const BYTE* p1, const BYTE* p2, co
}
forceinline static int LZ4HC_InsertAndFindBestMatch (LZ4HC_Data_Structure* hc4, const BYTE* ip, const BYTE* const matchlimit, const BYTE** matchpos)
static forceinline 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;
@ -397,7 +424,7 @@ forceinline static int LZ4HC_InsertAndFindBestMatch (LZ4HC_Data_Structure* hc4,
INITBASE(base,hc4->base);
int nbAttempts=MAX_NB_ATTEMPTS;
size_t repl=0, ml=0;
U16 delta;
U16 delta=0; // useless assignment, to remove an uninitialization warning
// HC4 match finder
LZ4HC_Insert(hc4, ip);
@ -406,7 +433,7 @@ forceinline static int LZ4HC_InsertAndFindBestMatch (LZ4HC_Data_Structure* hc4,
#define REPEAT_OPTIMIZATION
#ifdef REPEAT_OPTIMIZATION
// Detect repetitive sequences of length <= 4
if (ref >= ip-4) // potential repetition
if ((U32)(ip-ref) <= 4) // potential repetition
{
if (A32(ref) == A32(ip)) // confirmed
{
@ -418,7 +445,7 @@ forceinline static int LZ4HC_InsertAndFindBestMatch (LZ4HC_Data_Structure* hc4,
}
#endif
while ((ref >= ip-MAX_DISTANCE) && (nbAttempts))
while (((U32)(ip-ref) <= MAX_DISTANCE) && (nbAttempts))
{
nbAttempts--;
if (*(ref+ml) == *(ip+ml))
@ -457,7 +484,7 @@ forceinline static int LZ4HC_InsertAndFindBestMatch (LZ4HC_Data_Structure* hc4,
}
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)
static forceinline 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;
@ -470,7 +497,7 @@ forceinline static int LZ4HC_InsertAndGetWiderMatch (LZ4HC_Data_Structure* hc4,
LZ4HC_Insert(hc4, ip);
ref = HASH_POINTER(ip);
while ((ref >= ip-MAX_DISTANCE) && (nbAttempts))
while (((U32)(ip-ref) <= MAX_DISTANCE) && (nbAttempts))
{
nbAttempts--;
if (*(startLimit + longest) == *(ref - delta + longest))
@ -500,7 +527,7 @@ _endCount:
const BYTE* ipt = ip + MINMATCH + LZ4HC_CommonLength(ip+MINMATCH, ref+MINMATCH, matchlimit);
#endif
while ((startt>startLimit) && (reft > hc4->base) && (startt[-1] == reft[-1])) {startt--; reft--;}
while ((startt>startLimit) && (reft > hc4->inputBuffer) && (startt[-1] == reft[-1])) {startt--; reft--;}
if ((ipt-startt) > longest)
{

48
lz4hc.h
View File

@ -40,32 +40,70 @@ extern "C" {
int LZ4_compressHC (const char* source, char* dest, int inputSize);
/*
LZ4_compressHC :
return : the number of bytes in compressed buffer dest
or 0 if compression fails.
note : destination buffer must be already allocated.
To avoid any problem, size it to handle worst cases situations (input data not compressible)
Worst case size evaluation is provided by function LZ4_compressBound() (see "lz4.h")
*/
int LZ4_compressHC_limitedOutput (const char* source, char* dest, int inputSize, int maxOutputSize);
/*
LZ4_compress_limitedOutput() :
Compress 'inputSize' 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.
This function never writes outside of provided output buffer.
inputSize : Max supported value is ~1.9GB
inputSize : Max supported value is 1 GB
maxOutputSize : is maximum allowed size into the destination buffer (which must be already allocated)
return : the number of output bytes written in buffer 'dest'
or 0 if the compression fails
or 0 if compression fails.
*/
/* Note :
Decompression functions are provided within regular LZ4 source code (see "lz4.h") (BSD license)
Decompression functions are provided within LZ4 source code (see "lz4.h") (BSD license)
*/
/* Advanced Functions
*/
void* LZ4_createHC (const char* slidingInputBuffer);
int LZ4_compressHC_continue (void* LZ4HC_Data, const char* source, char* dest, int inputSize);
int LZ4_compressHC_limitedOutput_continue (void* LZ4HC_Data, const char* source, char* dest, int inputSize, int maxOutputSize);
char* LZ4_slideInputBufferHC (void* LZ4HC_Data);
int LZ4_freeHC (void* LZ4HC_Data);
/*
These functions allow the compression of dependent blocks, where each block benefits from prior 64 KB within preceding blocks.
In order to achieve this, it is necessary to start creating the LZ4HC Data Structure, thanks to the function :
void* LZ4_createHC (const char* slidingInputBuffer);
The result of the function is the (void*) pointer on the LZ4HC Data Structure.
This pointer will be needed in all other functions.
If the pointer returned is NULL, then the allocation has failed, and compression must be aborted.
The only parameter 'const char* slidingInputBuffer' must, obviously, point at the beginning of input buffer.
The input buffer must be already allocated, and size at least 192KB.
'slidingInputBuffer' will also be the 'const char* source' of the first block.
All blocks are expected to lay next to each other within the input buffer, starting from 'slidingInputBuffer'.
To compress each block, use either LZ4_compressHC_continue() or LZ4_compressHC_limitedOutput_continue().
Their behavior are identical to LZ4_compressHC() or LZ4_compressHC_limitedOutput(),
but require the LZ4HC Data Structure as their first argument, and check that each block starts right after the previous one.
If next block does not begin immediately after the previous one, the compression will fail (return 0).
When it's no longer possible to lay the next block after the previous one (not enough space left into input buffer), a call to :
char* LZ4_slideInputBufferHC(void* LZ4HC_Data);
must be performed. It will typically copy the latest 64KB of input at the beginning of input buffer.
Note that, for this function to work properly, minimum size of an input buffer must be 192KB.
==> The memory position where the next input data block must start is provided as the result of the function.
Compression can then resume, using LZ4_compressHC_continue() or LZ4_compressHC_limitedOutput_continue(), as usual.
When compression is completed, a call to LZ4_freeHC() will release the memory used by the LZ4HC Data Structure.
*/

View File

@ -107,8 +107,8 @@ forceinline static int ENCODE_SEQUENCE_NAME (
}
int COMBINED_NAME(FUNCTION_NAME,ctx) (
LZ4HC_Data_Structure* ctx,
int COMBINED_NAME(FUNCTION_NAME,_continue) (
void* ctxvoid,
const char* source,
char* dest,
int inputSize
@ -117,6 +117,7 @@ int COMBINED_NAME(FUNCTION_NAME,ctx) (
#endif
)
{
LZ4HC_Data_Structure* ctx = (LZ4HC_Data_Structure*) ctxvoid;
const BYTE* ip = (const BYTE*) source;
const BYTE* anchor = ip;
const BYTE* const iend = ip + inputSize;
@ -137,6 +138,10 @@ int COMBINED_NAME(FUNCTION_NAME,ctx) (
const BYTE* start0;
const BYTE* ref0;
// Ensure blocks follow each other
if (ip != ctx->end) return 0;
ctx->end += inputSize;
ip++;
// Main Loop
@ -311,13 +316,15 @@ int FUNCTION_NAME (const char* source,
#endif
)
{
void* ctx = LZ4HC_create((const BYTE*)source);
void* ctx = LZ4_createHC(source);
int result;
if (ctx==NULL) return 0;
#ifdef LIMITED_OUTPUT
int result = COMBINED_NAME(FUNCTION_NAME,ctx) (ctx, source, dest, inputSize, maxOutputSize);
result = COMBINED_NAME(FUNCTION_NAME,_continue) (ctx, source, dest, inputSize, maxOutputSize);
#else
int result = COMBINED_NAME(FUNCTION_NAME,ctx) (ctx, source, dest, inputSize);
result = COMBINED_NAME(FUNCTION_NAME,_continue) (ctx, source, dest, inputSize);
#endif
LZ4HC_free (&ctx);
LZ4_freeHC(ctx);
return result;
}