From: Stella Lau Date: Wed, 12 Jul 2017 01:13:26 +0000 (-0700) Subject: Update rolling hash X-Git-Tag: v1.3.1^2~12^2~7^2~43 X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=583dda17a811a9f4aab42ed4178d296ca5551447;p=thirdparty%2Fzstd.git Update rolling hash --- diff --git a/contrib/long_distance_matching/ldm.c b/contrib/long_distance_matching/ldm.c index 1dedf5c37..ca4f0f2cf 100644 --- a/contrib/long_distance_matching/ldm.c +++ b/contrib/long_distance_matching/ldm.c @@ -9,7 +9,7 @@ #define HASH_EVERY 1 -#define LDM_MEMORY_USAGE 16 +#define LDM_MEMORY_USAGE 22 #define LDM_HASHLOG (LDM_MEMORY_USAGE-2) #define LDM_HASHTABLESIZE (1 << (LDM_MEMORY_USAGE)) #define LDM_HASHTABLESIZE_U32 ((LDM_HASHTABLESIZE) >> 2) @@ -17,10 +17,12 @@ #define LDM_OFFSET_SIZE 4 -#define WINDOW_SIZE (1 << 20) +#define WINDOW_SIZE (1 << 23) #define MAX_WINDOW_SIZE 31 #define HASH_SIZE 4 -#define LDM_HASH_LENGTH 4 +#define LDM_HASH_LENGTH 100 + +// Should be multiple of four #define MINMATCH 4 #define ML_BITS 4 @@ -28,7 +30,9 @@ #define RUN_BITS (8-ML_BITS) #define RUN_MASK ((1U<totalLiteralLength) / (double)stats->numMatches); printf("Average offset length: %.1f\n", ((double)stats->totalOffset) / (double)stats->numMatches); + printf("Num collisions, num hash inserts, %% collisions: %u, %u, %.3f\n", + stats->numCollisions, stats->numHashInserts, + stats->numHashInserts == 0 ? + 1.0 : (100.0 * (double)stats->numCollisions) / + (double)stats->numHashInserts); printf("=====================\n"); } @@ -95,17 +108,43 @@ typedef struct LDM_CCtx { const BYTE *nextIp; hash_t nextHash; /* Hash corresponding to nextIp */ + // Members for rolling hash. + U32 lastSum; + U32 nextSum; + unsigned step; + + // DEBUG + const BYTE *DEBUG_setNextHash; } LDM_CCtx; +static int LDM_isValidMatch(const BYTE *p, const BYTE *match) { + U16 lengthLeft = MINMATCH; + const BYTE *curP = p; + const BYTE *curMatch = match; + + for (; lengthLeft >= 8; lengthLeft -= 8) { + if (LDM_read64(curP) != LDM_read64(curMatch)) { + return 0; + } + curP += 8; + curMatch += 8; + } + if (lengthLeft > 0) { + return LDM_read32(curP) == LDM_read32(curMatch); + } + return 1; +} + + + #ifdef LDM_ROLLING_HASH /** * Convert a sum computed from LDM_getRollingHash to a hash value in the range * of the hash table. */ static hash_t LDM_sumToHash(U32 sum) { - return sum % (LDM_HASHTABLESIZE >> 2); -// return sum & (LDM_HASHTABLESIZE - 1); + return sum & (LDM_HASH_SIZE_U32 - 1); } static U32 LDM_getRollingHash(const char *data, U32 len) { @@ -126,18 +165,102 @@ static U32 LDM_getRollingHash(const char *data, U32 len) { return (s1 & 0xffff) + (s2 << 16); } -static hash_t LDM_hashPosition(const void * const p) { - return LDM_sumToHash(LDM_getRollingHash((const char *)p, LDM_HASH_LENGTH)); -} - typedef struct LDM_sumStruct { U16 s1, s2; } LDM_sumStruct; +static U32 LDM_updateRollingHash(U32 sum, U32 len, + schar toRemove, schar toAdd) { + U32 s1 = (sum & 0xffff) - toRemove + toAdd; + U32 s2 = (sum >> 16) - (toRemove * len) + s1; + + return (s1 & 0xffff) + (s2 << 16); +} + + +/* +static hash_t LDM_hashPosition(const void * const p) { + return LDM_sumToHash(LDM_getRollingHash((const char *)p, LDM_HASH_LENGTH)); +} +*/ + +/* static void LDM_getRollingHashParts(U32 sum, LDM_sumStruct *sumStruct) { sumStruct->s1 = sum & 0xffff; sumStruct->s2 = sum >> 16; } +*/ + +static void LDM_setNextHash(LDM_CCtx *cctx) { + U32 check; + +#ifdef RUN_CHECKS + if ((cctx->nextIp - cctx->ibase != 1) && + (cctx->nextIp - cctx->DEBUG_setNextHash != 1)) { + printf("CHECK debug fail: %zu %zu\n", cctx->nextIp - cctx->ibase, + cctx->DEBUG_setNextHash - cctx->ibase); + } + + cctx->DEBUG_setNextHash = cctx->nextIp; +#endif + + cctx->nextSum = LDM_getRollingHash((const char *)cctx->nextIp, LDM_HASH_LENGTH); + /* + check = LDM_updateRollingHash( + cctx->lastSum, LDM_HASH_LENGTH, + (schar)((cctx->lastPosHashed)[0]), + (schar)((cctx->lastPosHashed)[LDM_HASH_LENGTH])); + */ + +#ifdef RUN_CHECKS + if (check != cctx->nextSum) { + printf("CHECK: setNextHash failed %u %u\n", check, cctx->nextSum); +// printf("INFO: %u %u %u\n", LDM_read32(cctx->nextIp), + } else { +// printf("CHECK: setNextHash passed\n"); + } +#endif + cctx->nextHash = LDM_sumToHash(cctx->nextSum); + +#ifdef RUN_CHECKS + if ((cctx->nextIp - cctx->lastPosHashed) != 1) { + printf("setNextHash: nextIp != lastPosHashed + 1. %zu %zu %zu\n", + cctx->nextIp - cctx->ibase, cctx->lastPosHashed - cctx->ibase, + cctx->ip - cctx->ibase); + } +#endif + +} + +static void LDM_putHashOfCurrentPositionFromHash( + LDM_CCtx *cctx, hash_t hash, U32 sum) { + /* + if (((cctx->ip - cctx->ibase) & HASH_EVERY) != HASH_EVERY) { + return; + } + */ +#ifdef COMPUTE_STATS + if (cctx->stats.numHashInserts < LDM_HASHTABLESIZE_U32) { + offset_t offset = (cctx->hashTable)[hash].offset; + cctx->stats.numHashInserts++; + if (offset == 0 && !LDM_isValidMatch(cctx->ip, offset + cctx->ibase)) { + cctx->stats.numCollisions++; + } + } +#endif + (cctx->hashTable)[hash] = (LDM_hashEntry){ (hash_t)(cctx->ip - cctx->ibase) }; + cctx->lastPosHashed = cctx->ip; + cctx->lastHash = hash; + cctx->lastSum = sum; +} + +static void LDM_putHashOfCurrentPosition(LDM_CCtx *cctx) { + U32 sum = LDM_getRollingHash((const char *)cctx->ip, LDM_HASH_LENGTH); + hash_t hash = LDM_sumToHash(sum); +// hash_t hash = LDM_hashPosition(cctx->ip); + LDM_putHashOfCurrentPositionFromHash(cctx, hash, sum); +// printf("Offset %zu\n", cctx->ip - cctx->ibase); +} #else static hash_t LDM_hash(U32 sequence) { @@ -147,26 +270,30 @@ static hash_t LDM_hash(U32 sequence) { static hash_t LDM_hashPosition(const void * const p) { return LDM_hash(LDM_read32(p)); } -#endif - -/* -static hash_t LDM_hash5(U64 sequence) { - static const U64 prime5bytes = 889523592379ULL; - static const U64 prime8bytes = 11400714785074694791ULL; - const U32 hashLog = LDM_HASHLOG; - if (LDM_isLittleEndian()) - return (((sequence << 24) * prime5bytes) >> (64 - hashLog)); - else - return (((sequence >> 24) * prime8bytes) >> (64 - hashLog)); -} -*/ static void LDM_putHashOfCurrentPositionFromHash( LDM_CCtx *cctx, hash_t hash) { + /* if (((cctx->ip - cctx->ibase) & HASH_EVERY) != HASH_EVERY) { return; } + */ +#ifdef COMPUTE_STATS + if (cctx->stats.numHashInserts < LDM_HASHTABLESIZE_U32) { + offset_t offset = (cctx->hashTable)[hash].offset; + cctx->stats.numHashInserts++; + if (offset == 0 && !LDM_isValidMatch(cctx->ip, offset + cctx->ibase)) { + cctx->stats.numCollisions++; + } + } +#endif + (cctx->hashTable)[hash] = (LDM_hashEntry){ (hash_t)(cctx->ip - cctx->ibase) }; +#ifdef RUN_CHECKS + if (cctx->ip - cctx->lastPosHashed != 1) { + printf("putHashError\n"); + } +#endif cctx->lastPosHashed = cctx->ip; cctx->lastHash = hash; } @@ -176,23 +303,33 @@ static void LDM_putHashOfCurrentPosition(LDM_CCtx *cctx) { LDM_putHashOfCurrentPositionFromHash(cctx, hash); } -static const BYTE *LDM_get_position_on_hash( +#endif + +/* +static hash_t LDM_hash5(U64 sequence) { + static const U64 prime5bytes = 889523592379ULL; + static const U64 prime8bytes = 11400714785074694791ULL; + const U32 hashLog = LDM_HASHLOG; + if (LDM_isLittleEndian()) + return (((sequence << 24) * prime5bytes) >> (64 - hashLog)); + else + return (((sequence >> 24) * prime8bytes) >> (64 - hashLog)); +} +*/ + + +static const BYTE *LDM_getPositionOnHash( hash_t h, void *tableBase, const BYTE *srcBase) { const LDM_hashEntry * const hashTable = (LDM_hashEntry *)tableBase; return hashTable[h].offset + srcBase; } -static BYTE LDM_read_byte(const void *memPtr) { - BYTE val; - memcpy(&val, memPtr, 1); - return val; -} static unsigned LDM_count(const BYTE *pIn, const BYTE *pMatch, const BYTE *pInLimit) { const BYTE * const pStart = pIn; while (pIn < pInLimit - 1) { - BYTE const diff = LDM_read_byte(pMatch) ^ LDM_read_byte(pIn); + BYTE const diff = LDM_readByte(pMatch) ^ LDM_readByte(pIn); if (!diff) { pIn++; pMatch++; @@ -220,7 +357,11 @@ static void LDM_initializeCCtx(LDM_CCtx *cctx, cctx->ip = cctx->ibase; cctx->iend = cctx->ibase + srcSize; +#ifdef LDM_ROLLING_HASH + cctx->ihashLimit = cctx->iend - LDM_HASH_LENGTH; +#else cctx->ihashLimit = cctx->iend - HASH_SIZE; +#endif cctx->imatchLimit = cctx->iend - MINMATCH; cctx->obase = (BYTE *)dst; @@ -232,11 +373,46 @@ static void LDM_initializeCCtx(LDM_CCtx *cctx, memset(cctx->hashTable, 0, sizeof(cctx->hashTable)); cctx->lastPosHashed = NULL; - cctx->nextIp = NULL; cctx->step = 1; + cctx->nextIp = cctx->ip + cctx->step; + + cctx->DEBUG_setNextHash = 0; } +#ifdef LDM_ROLLING_HASH +static int LDM_findBestMatch(LDM_CCtx *cctx, const BYTE **match) { + cctx->nextIp = cctx->ip + cctx->step; + + do { + hash_t h; + U32 sum; +// printf("Call A\n"); + LDM_setNextHash(cctx); +// printf("End call a\n"); + h = cctx->nextHash; + sum = cctx->nextSum; + cctx->ip = cctx->nextIp; + cctx->nextIp += cctx->step; + + if (cctx->ip > cctx->imatchLimit) { + return 1; + } + + *match = LDM_getPositionOnHash(h, cctx->hashTable, cctx->ibase); + +// // Compute cctx->nextSum and cctx->nextHash from cctx->nextIp. +// LDM_setNextHash(cctx); + LDM_putHashOfCurrentPositionFromHash(cctx, h, sum); + +// printf("%u %u\n", cctx->lastHash, cctx->nextHash); + } while (cctx->ip - *match > WINDOW_SIZE || + !LDM_isValidMatch(cctx->ip, *match)); +// LDM_read64(*match) != LDM_read64(cctx->ip)); + LDM_setNextHash(cctx); + return 0; +} +#else static int LDM_findBestMatch(LDM_CCtx *cctx, const BYTE **match) { cctx->nextIp = cctx->ip; @@ -245,33 +421,131 @@ static int LDM_findBestMatch(LDM_CCtx *cctx, const BYTE **match) { cctx->ip = cctx->nextIp; cctx->nextIp += cctx->step; - if (cctx->nextIp > cctx->imatchLimit) { + if (cctx->ip > cctx->imatchLimit) { return 1; } - *match = LDM_get_position_on_hash(h, cctx->hashTable, cctx->ibase); + *match = LDM_getPositionOnHash(h, cctx->hashTable, cctx->ibase); cctx->nextHash = LDM_hashPosition(cctx->nextIp); LDM_putHashOfCurrentPositionFromHash(cctx, h); + } while (cctx->ip - *match > WINDOW_SIZE || - LDM_read64(*match) != LDM_read64(cctx->ip)); + !LDM_isValidMatch(cctx->ip, *match)); return 0; } +#endif + +/** + * Write current block (literals, literal length, match offset, + * match length). + * + * Update input pointer, inserting hashes into hash table along the + * way. + */ +static void LDM_outputBlock(LDM_CCtx *cctx, const BYTE *match) { + unsigned const literalLength = (unsigned)(cctx->ip - cctx->anchor); + unsigned const offset = cctx->ip - match; + unsigned const matchLength = LDM_count( + cctx->ip + MINMATCH, match + MINMATCH, cctx->ihashLimit); + BYTE *token = cctx->op++; + + cctx->stats.totalLiteralLength += literalLength; + cctx->stats.totalOffset += offset; + cctx->stats.totalMatchLength += matchLength + MINMATCH; + + /* Encode the literal length. */ + if (literalLength >= RUN_MASK) { + int len = (int)literalLength - RUN_MASK; + *token = (RUN_MASK << ML_BITS); + for (; len >= 255; len -= 255) { + *(cctx->op)++ = 255; + } + *(cctx->op)++ = (BYTE)len; + } else { + *token = (BYTE)(literalLength << ML_BITS); + } + + /* Encode the literals. */ + memcpy(cctx->op, cctx->anchor, literalLength); + cctx->op += literalLength; + + /* Encode the offset. */ + LDM_write32(cctx->op, offset); + cctx->op += LDM_OFFSET_SIZE; + + /* Encode match length */ + if (matchLength >= ML_MASK) { + unsigned matchLengthRemaining = matchLength; + *token += ML_MASK; + matchLengthRemaining -= ML_MASK; + LDM_write32(cctx->op, 0xFFFFFFFF); + while (matchLengthRemaining >= 4*0xFF) { + cctx->op += 4; + LDM_write32(cctx->op, 0xffffffff); + matchLengthRemaining -= 4*0xFF; + } + cctx->op += matchLengthRemaining / 255; + *(cctx->op)++ = (BYTE)(matchLengthRemaining % 255); + } else { + *token += (BYTE)(matchLength); + } + +// LDM_setNextHash(cctx); +// cctx->ip = cctx->lastPosHashed + 1; +// cctx->nextIp = cctx->ip + cctx->step; +// printf("HERE: %zu %zu %zu\n", cctx->ip - cctx->ibase, +// cctx->lastPosHashed - cctx->ibase, cctx->nextIp - cctx->ibase); + + cctx->nextIp = cctx->ip + cctx->step; + + while (cctx->ip < cctx->anchor + MINMATCH + matchLength + literalLength) { +// printf("Loop\n"); + if (cctx->ip > cctx->lastPosHashed) { + LDM_putHashOfCurrentPosition(cctx); +#ifdef LDM_ROLLING_HASH + LDM_setNextHash(cctx); +#endif + } + /* + printf("Call b %zu %zu %zu\n", + cctx->lastPosHashed - cctx->ibase, + cctx->nextIp - cctx->ibase, + cctx->ip - cctx->ibase); + */ +// printf("end call b\n"); + cctx->ip++; + cctx->nextIp++; + } + +// printf("There: %zu %zu\n", cctx->ip - cctx->ibase, cctx->lastPosHashed - cctx->ibase); +} + // TODO: srcSize and maxDstSize is unused size_t LDM_compress(const void *src, size_t srcSize, void *dst, size_t maxDstSize) { LDM_CCtx cctx; + U32 tmp_hash; LDM_initializeCCtx(&cctx, src, srcSize, dst, maxDstSize); /* Hash the first position and put it into the hash table. */ LDM_putHashOfCurrentPosition(&cctx); +#ifdef LDM_ROLLING_HASH +// LDM_setNextHash(&cctx); +// tmp_hash = LDM_updateRollingHash(cctx.lastSum, LDM_HASH_LENGTH, +// cctx.ip[0], cctx.ip[LDM_HASH_LENGTH]); +// printf("Update test: %u %u\n", tmp_hash, cctx.nextSum); +// cctx.ip++; +#else cctx.ip++; cctx.nextHash = LDM_hashPosition(cctx.ip); +#endif // TODO: loop condition is not accurate. while (1) { const BYTE *match; +// printf("Start of loop\n"); /** * Find a match. @@ -282,6 +556,7 @@ size_t LDM_compress(const void *src, size_t srcSize, if (LDM_findBestMatch(&cctx, &match) != 0) { goto _last_literals; } +// printf("End of match finding\n"); cctx.stats.numMatches++; @@ -290,6 +565,7 @@ size_t LDM_compress(const void *src, size_t srcSize, */ while (cctx.ip > cctx.anchor && match > cctx.ibase && cctx.ip[-1] == match[-1]) { +// printf("Catch up\n"); cctx.ip--; match--; } @@ -298,67 +574,24 @@ size_t LDM_compress(const void *src, size_t srcSize, * Write current block (literals, literal length, match offset, match * length) and update pointers and hashes. */ - { - unsigned const literalLength = (unsigned)(cctx.ip - cctx.anchor); - unsigned const offset = cctx.ip - match; - unsigned const matchLength = LDM_count( - cctx.ip + MINMATCH, match + MINMATCH, cctx.ihashLimit); - BYTE *token = cctx.op++; - - cctx.stats.totalLiteralLength += literalLength; - cctx.stats.totalOffset += offset; - cctx.stats.totalMatchLength += matchLength + MINMATCH; - - /* Encode the literal length. */ - if (literalLength >= RUN_MASK) { - int len = (int)literalLength - RUN_MASK; - *token = (RUN_MASK << ML_BITS); - for (; len >= 255; len -= 255) { - *(cctx.op)++ = 255; - } - *(cctx.op)++ = (BYTE)len; - } else { - *token = (BYTE)(literalLength << ML_BITS); - } - - /* Encode the literals. */ - memcpy(cctx.op, cctx.anchor, literalLength); - cctx.op += literalLength; - - /* Encode the offset. */ - LDM_write32(cctx.op, offset); - cctx.op += LDM_OFFSET_SIZE; - - /* Encode match length */ - if (matchLength >= ML_MASK) { - unsigned matchLengthRemaining = matchLength; - *token += ML_MASK; - matchLengthRemaining -= ML_MASK; - LDM_write32(cctx.op, 0xFFFFFFFF); - while (matchLengthRemaining >= 4*0xFF) { - cctx.op += 4; - LDM_write32(cctx.op, 0xffffffff); - matchLengthRemaining -= 4*0xFF; - } - cctx.op += matchLengthRemaining / 255; - *(cctx.op)++ = (BYTE)(matchLengthRemaining % 255); - } else { - *token += (BYTE)(matchLength); - } - - /* Update input pointer, inserting hashes into hash table along the - * way. - */ - while (cctx.ip < cctx.anchor + MINMATCH + matchLength + literalLength) { - LDM_putHashOfCurrentPosition(&cctx); - cctx.ip++; - } - } + LDM_outputBlock(&cctx, match); +// printf("End of loop\n"); // Set start of next block to current input pointer. cctx.anchor = cctx.ip; LDM_putHashOfCurrentPosition(&cctx); - cctx.nextHash = LDM_hashPosition(++cctx.ip); +#ifndef LDM_ROLLING_HASH + cctx.ip++; +#endif + + /* + LDM_putHashOfCurrentPosition(&cctx); + printf("Call c\n"); + LDM_setNextHash(&cctx); + printf("End call c\n"); + cctx.ip++; + cctx.nextIp++; + */ } _last_literals: /* Encode the last literals (no more matches). */ @@ -453,7 +686,7 @@ size_t LDM_decompress(const void *src, size_t compressSize, /* Copy match. */ cpy = dctx.op + length; - // Inefficient for now + // Inefficient for now. while (match < cpy - offset && dctx.op < dctx.oend) { *(dctx.op)++ = *match++; } @@ -461,4 +694,20 @@ size_t LDM_decompress(const void *src, size_t compressSize, return dctx.op - (BYTE *)dst; } +void LDM_test(const void *src, size_t srcSize, + void *dst, size_t maxDstSize) { +#ifdef LDM_ROLLING_HASH + const BYTE *ip = (const BYTE *)src + 1125; + U32 sum = LDM_getRollingHash((const char *)ip, LDM_HASH_LENGTH); + U32 sum2; + ++ip; + for (; ip < (const BYTE *)src + 1125 + 100; ip++) { + sum2 = LDM_updateRollingHash(sum, LDM_HASH_LENGTH, + ip[-1], ip[LDM_HASH_LENGTH - 1]); + sum = LDM_getRollingHash((const char *)ip, LDM_HASH_LENGTH); + printf("TEST HASH: %zu %u %u\n", ip - (const BYTE *)src, sum, sum2); + } +#endif +} + diff --git a/contrib/long_distance_matching/ldm.h b/contrib/long_distance_matching/ldm.h index 287d444dd..a34faac4f 100644 --- a/contrib/long_distance_matching/ldm.h +++ b/contrib/long_distance_matching/ldm.h @@ -16,4 +16,7 @@ size_t LDM_decompress(const void *src, size_t srcSize, void LDM_readHeader(const void *src, size_t *compressSize, size_t *decompressSize); +void LDM_test(const void *src, size_t srcSize, + void *dst, size_t maxDstSize); + #endif /* LDM_H */ diff --git a/contrib/long_distance_matching/main-ldm.c b/contrib/long_distance_matching/main-ldm.c index 724d735dd..f8ae54698 100644 --- a/contrib/long_distance_matching/main-ldm.c +++ b/contrib/long_distance_matching/main-ldm.c @@ -15,6 +15,7 @@ // #define BUF_SIZE 16*1024 // Block size #define DEBUG +//#define TEST //#define ZSTD @@ -74,6 +75,11 @@ static int compress(const char *fname, const char *oname) { return 1; } +#ifdef TEST + LDM_test(src, statbuf.st_size, + dst + LDM_HEADER_SIZE, statbuf.st_size); +#endif + #ifdef ZSTD compressSize = ZSTD_compress(dst, statbuf.st_size, src, statbuf.st_size, 1); @@ -144,11 +150,6 @@ static int decompress(const char *fname, const char *oname) { /* Read the header. */ LDM_readHeader(src, &compressSize, &decompressSize); -#ifdef DEBUG - printf("Size, compressSize, decompressSize: %zu %zu %zu\n", - (size_t)statbuf.st_size, compressSize, decompressSize); -#endif - /* Go to the location corresponding to the last byte. */ if (lseek(fdout, decompressSize - 1, SEEK_SET) == -1) { perror("lseek error"); @@ -256,7 +257,7 @@ int main(int argc, const char *argv[]) { return 1; } gettimeofday(&tv2, NULL); - printf("Total time = %f seconds\n", + printf("Total compress time = %f seconds\n", (double) (tv2.tv_usec - tv1.tv_usec) / 1000000 + (double) (tv2.tv_sec - tv1.tv_sec)); } @@ -270,7 +271,7 @@ int main(int argc, const char *argv[]) { return 1; } gettimeofday(&tv2, NULL); - printf("Total time = %f seconds\n", + printf("Total decompress time = %f seconds\n", (double) (tv2.tv_usec - tv1.tv_usec) / 1000000 + (double) (tv2.tv_sec - tv1.tv_sec)); } diff --git a/contrib/long_distance_matching/util.c b/contrib/long_distance_matching/util.c index 9ea4ca1e5..70fcbc2ce 100644 --- a/contrib/long_distance_matching/util.c +++ b/contrib/long_distance_matching/util.c @@ -61,4 +61,9 @@ void LDM_copy8(void *dst, const void *src) { memcpy(dst, src, 8); } +BYTE LDM_readByte(const void *memPtr) { + BYTE val; + memcpy(&val, memPtr, 1); + return val; +} diff --git a/contrib/long_distance_matching/util.h b/contrib/long_distance_matching/util.h index 90726412e..d1c3c999b 100644 --- a/contrib/long_distance_matching/util.h +++ b/contrib/long_distance_matching/util.h @@ -19,5 +19,7 @@ uint64_t LDM_read64(const void *ptr); void LDM_copy8(void *dst, const void *src); +uint8_t LDM_readByte(const void *ptr); + #endif /* LDM_UTIL_H */ diff --git a/contrib/long_distance_matching/versions/v0.1/ldm.c b/contrib/long_distance_matching/versions/v0.1/ldm.c new file mode 100644 index 000000000..266425f8a --- /dev/null +++ b/contrib/long_distance_matching/versions/v0.1/ldm.c @@ -0,0 +1,394 @@ +#include +#include +#include +#include + +#include "ldm.h" + +#define LDM_MEMORY_USAGE 14 +#define LDM_HASHLOG (LDM_MEMORY_USAGE-2) +#define LDM_HASHTABLESIZE (1 << (LDM_MEMORY_USAGE)) +#define LDM_HASHTABLESIZE_U32 ((LDM_HASHTABLESIZE) >> 2) +#define LDM_HASH_SIZE_U32 (1 << (LDM_HASHLOG)) + +#define WINDOW_SIZE (1 << 20) +#define MAX_WINDOW_SIZE 31 +#define HASH_SIZE 4 +#define MINMATCH 4 + +#define ML_BITS 4 +#define ML_MASK ((1U<>8); + } +} + +static U32 LDM_read32(const void *ptr) { + return *(const U32 *)ptr; +} + +static U64 LDM_read64(const void *ptr) { + return *(const U64 *)ptr; +} + + +static void LDM_copy8(void *dst, const void *src) { + memcpy(dst, src, 8); +} + +static void LDM_wild_copy(void *dstPtr, const void *srcPtr, void *dstEnd) { + BYTE *d = (BYTE *)dstPtr; + const BYTE *s = (const BYTE *)srcPtr; + BYTE * const e = (BYTE *)dstEnd; + + do { + LDM_copy8(d, s); + d += 8; + s += 8; + } while (d < e); + +} + +struct hash_entry { + U64 offset; + tag t; +}; + +static U32 LDM_hash(U32 sequence) { + return ((sequence * 2654435761U) >> ((32)-LDM_HASHLOG)); +} + +static U32 LDM_hash5(U64 sequence) { + static const U64 prime5bytes = 889523592379ULL; + static const U64 prime8bytes = 11400714785074694791ULL; + const U32 hashLog = LDM_HASHLOG; + if (LDM_isLittleEndian()) + return (U32)(((sequence << 24) * prime5bytes) >> (64 - hashLog)); + else + return (U32)(((sequence >> 24) * prime8bytes) >> (64 - hashLog)); +} + +static U32 LDM_hash_position(const void * const p) { + return LDM_hash(LDM_read32(p)); +} + +static void LDM_put_position_on_hash(const BYTE *p, U32 h, void *tableBase, + const BYTE *srcBase) { + U32 *hashTable = (U32 *) tableBase; + hashTable[h] = (U32)(p - srcBase); +} + +static void LDM_put_position(const BYTE *p, void *tableBase, + const BYTE *srcBase) { + U32 const h = LDM_hash_position(p); + LDM_put_position_on_hash(p, h, tableBase, srcBase); +} + +static const BYTE *LDM_get_position_on_hash( + U32 h, void *tableBase, const BYTE *srcBase) { + const U32 * const hashTable = (U32*)tableBase; + return hashTable[h] + srcBase; +} + +static BYTE LDM_read_byte(const void *memPtr) { + BYTE val; + memcpy(&val, memPtr, 1); + return val; +} + +static unsigned LDM_count(const BYTE *pIn, const BYTE *pMatch, + const BYTE *pInLimit) { + const BYTE * const pStart = pIn; + while (pIn < pInLimit - 1) { + BYTE const diff = LDM_read_byte(pMatch) ^ LDM_read_byte(pIn); + if (!diff) { + pIn++; + pMatch++; + continue; + } + return (unsigned)(pIn - pStart); + } + return (unsigned)(pIn - pStart); +} + +void LDM_read_header(void const *source, size_t *compressed_size, + size_t *decompressed_size) { + const U32 *ip = (const U32 *)source; + *compressed_size = *ip++; + *decompressed_size = *ip; +} + +size_t LDM_compress(void const *source, void *dest, size_t source_size, + size_t max_dest_size) { + const BYTE * const istart = (const BYTE*)source; + const BYTE *ip = istart; + const BYTE * const iend = istart + source_size; + const BYTE *ilimit = iend - HASH_SIZE; + const BYTE * const matchlimit = iend - HASH_SIZE; + const BYTE * const mflimit = iend - MINMATCH; + BYTE *op = (BYTE*) dest; + U32 hashTable[LDM_HASHTABLESIZE_U32]; + memset(hashTable, 0, sizeof(hashTable)); + + const BYTE *anchor = (const BYTE *)source; +// struct LDM_cctx cctx; + size_t output_size = 0; + + U32 forwardH; + + /* Hash first byte: put into hash table */ + + LDM_put_position(ip, hashTable, istart); + ip++; + forwardH = LDM_hash_position(ip); + + //TODO Loop terminates before ip>=ilimit. + while (ip < ilimit) { + const BYTE *match; + BYTE *token; + + /* Find a match */ + { + const BYTE *forwardIp = ip; + unsigned step = 1; + + do { + U32 const h = forwardH; + ip = forwardIp; + forwardIp += step; + + if (forwardIp > mflimit) { + goto _last_literals; + } + + match = LDM_get_position_on_hash(h, hashTable, istart); + + forwardH = LDM_hash_position(forwardIp); + LDM_put_position_on_hash(ip, h, hashTable, istart); + } while (ip - match > WINDOW_SIZE || + LDM_read64(match) != LDM_read64(ip)); + } + + // TODO catchup + while (ip > anchor && match > istart && ip[-1] == match[-1]) { + ip--; + match--; + } + + /* Encode literals */ + { + unsigned const litLength = (unsigned)(ip - anchor); + token = op++; + +#ifdef LDM_DEBUG + printf("Cur position: %zu\n", anchor - istart); + printf("LitLength %zu. (Match offset). %zu\n", litLength, ip - match); +#endif + /* + fwrite(match, 4, 1, stdout); + printf("\n"); + */ + + if (litLength >= RUN_MASK) { + int len = (int)litLength - RUN_MASK; + *token = (RUN_MASK << ML_BITS); + for (; len >= 255; len -= 255) { + *op++ = 255; + } + *op++ = (BYTE)len; + } else { + *token = (BYTE)(litLength << ML_BITS); + } +#ifdef LDM_DEBUG + printf("Literals "); + fwrite(anchor, litLength, 1, stdout); + printf("\n"); +#endif + memcpy(op, anchor, litLength); + //LDM_wild_copy(op, anchor, op + litLength); + op += litLength; + } +_next_match: + /* Encode offset */ + { + LDM_write32(op, ip - match); + op += 4; + } + + /* Encode Match Length */ + { + unsigned matchCode; + matchCode = LDM_count(ip + MINMATCH, match + MINMATCH, + matchlimit); +#ifdef LDM_DEBUG + printf("Match length %zu\n", matchCode + MINMATCH); + fwrite(ip, MINMATCH + matchCode, 1, stdout); + printf("\n"); +#endif + ip += MINMATCH + matchCode; + if (matchCode >= ML_MASK) { + *token += ML_MASK; + matchCode -= ML_MASK; + LDM_write32(op, 0xFFFFFFFF); + while (matchCode >= 4*0xFF) { + op += 4; + LDM_write32(op, 0xffffffff); + matchCode -= 4*0xFF; + } + op += matchCode / 255; + *op++ = (BYTE)(matchCode % 255); + } else { + *token += (BYTE)(matchCode); + } +#ifdef LDM_DEBUG + printf("\n"); +#endif + } + + anchor = ip; + + LDM_put_position(ip, hashTable, istart); + forwardH = LDM_hash_position(++ip); + } +_last_literals: + /* Encode last literals */ + { + size_t const lastRun = (size_t)(iend - anchor); + if (lastRun >= RUN_MASK) { + size_t accumulator = lastRun - RUN_MASK; + *op++ = RUN_MASK << ML_BITS; + for(; accumulator >= 255; accumulator -= 255) { + *op++ = 255; + } + *op++ = (BYTE)accumulator; + } else { + *op++ = (BYTE)(lastRun << ML_BITS); + } + memcpy(op, anchor, lastRun); + op += lastRun; + } + return (op - (BYTE *)dest); +} + +size_t LDM_decompress(void const *source, void *dest, size_t compressed_size, + size_t max_decompressed_size) { + const BYTE *ip = (const BYTE *)source; + const BYTE * const iend = ip + compressed_size; + BYTE *op = (BYTE *)dest; + BYTE * const oend = op + max_decompressed_size; + BYTE *cpy; + + while (ip < iend) { + size_t length; + const BYTE *match; + size_t offset; + + /* get literal length */ + unsigned const token = *ip++; + if ((length=(token >> ML_BITS)) == RUN_MASK) { + unsigned s; + do { + s = *ip++; + length += s; + } while (s == 255); + } +#ifdef LDM_DEBUG + printf("Literal length: %zu\n", length); +#endif + + /* copy literals */ + cpy = op + length; +#ifdef LDM_DEBUG + printf("Literals "); + fwrite(ip, length, 1, stdout); + printf("\n"); +#endif + memcpy(op, ip, length); +// LDM_wild_copy(op, ip, cpy); + ip += length; + op = cpy; + + /* get offset */ + offset = LDM_read32(ip); + +#ifdef LDM_DEBUG + printf("Offset: %zu\n", offset); +#endif + ip += 4; + match = op - offset; + // LDM_write32(op, (U32)offset); + + /* get matchlength */ + length = token & ML_MASK; + if (length == ML_MASK) { + unsigned s; + do { + s = *ip++; + length += s; + } while (s == 255); + } + length += MINMATCH; +#ifdef LDM_DEBUG + printf("Match length: %zu\n", length); +#endif + /* copy match */ + cpy = op + length; + + // Inefficient for now + + while (match < cpy - offset && op < oend) { + *op++ = *match++; + } + } +// memcpy(dest, source, compressed_size); + return op - (BYTE *)dest; +} + + diff --git a/contrib/long_distance_matching/versions/v0.1/ldm.h b/contrib/long_distance_matching/versions/v0.1/ldm.h new file mode 100644 index 000000000..f4ca25a38 --- /dev/null +++ b/contrib/long_distance_matching/versions/v0.1/ldm.h @@ -0,0 +1,19 @@ +#ifndef LDM_H +#define LDM_H + +#include /* size_t */ + +#define LDM_COMPRESS_SIZE 4 +#define LDM_DECOMPRESS_SIZE 4 +#define LDM_HEADER_SIZE ((LDM_COMPRESS_SIZE)+(LDM_DECOMPRESS_SIZE)) + +size_t LDM_compress(void const *source, void *dest, size_t source_size, + size_t max_dest_size); + +size_t LDM_decompress(void const *source, void *dest, size_t compressed_size, + size_t max_decompressed_size); + +void LDM_read_header(void const *source, size_t *compressed_size, + size_t *decompressed_size); + +#endif /* LDM_H */ diff --git a/contrib/long_distance_matching/versions/v0.1/main-ldm.c b/contrib/long_distance_matching/versions/v0.1/main-ldm.c new file mode 100644 index 000000000..10869cce3 --- /dev/null +++ b/contrib/long_distance_matching/versions/v0.1/main-ldm.c @@ -0,0 +1,459 @@ +// TODO: file size must fit into a U32 + +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include "ldm.h" + +// #define BUF_SIZE 16*1024 // Block size +#define DEBUG + +//#define ZSTD + +#if 0 +static size_t compress_file(FILE *in, FILE *out, size_t *size_in, + size_t *size_out) { + char *src, *buf = NULL; + size_t r = 1; + size_t size, n, k, count_in = 0, count_out = 0, offset, frame_size = 0; + + src = malloc(BUF_SIZE); + if (!src) { + printf("Not enough memory\n"); + goto cleanup; + } + + size = BUF_SIZE + LDM_HEADER_SIZE; + buf = malloc(size); + if (!buf) { + printf("Not enough memory\n"); + goto cleanup; + } + + + for (;;) { + k = fread(src, 1, BUF_SIZE, in); + if (k == 0) + break; + count_in += k; + + n = LDM_compress(src, buf, k, BUF_SIZE); + + // n = k; + // offset += n; + offset = k; + count_out += k; + +// k = fwrite(src, 1, offset, out); + + k = fwrite(buf, 1, offset, out); + if (k < offset) { + if (ferror(out)) + printf("Write failed\n"); + else + printf("Short write\n"); + goto cleanup; + } + + } + *size_in = count_in; + *size_out = count_out; + r = 0; + cleanup: + free(src); + free(buf); + return r; +} + +static size_t decompress_file(FILE *in, FILE *out) { + void *src = malloc(BUF_SIZE); + void *dst = NULL; + size_t dst_capacity = BUF_SIZE; + size_t ret = 1; + size_t bytes_written = 0; + + if (!src) { + perror("decompress_file(src)"); + goto cleanup; + } + + while (ret != 0) { + /* Load more input */ + size_t src_size = fread(src, 1, BUF_SIZE, in); + void *src_ptr = src; + void *src_end = src_ptr + src_size; + if (src_size == 0 || ferror(in)) { + printf("(TODO): Decompress: not enough input or error reading file\n"); + //TODO + ret = 0; + goto cleanup; + } + + /* Allocate destination buffer if it hasn't been allocated already */ + if (!dst) { + dst = malloc(dst_capacity); + if (!dst) { + perror("decompress_file(dst)"); + goto cleanup; + } + } + + // TODO + + /* Decompress: + * Continue while there is more input to read. + */ + while (src_ptr != src_end && ret != 0) { + // size_t dst_size = src_size; + size_t dst_size = LDM_decompress(src, dst, src_size, dst_capacity); + size_t written = fwrite(dst, 1, dst_size, out); +// printf("Writing %zu bytes\n", dst_size); + bytes_written += dst_size; + if (written != dst_size) { + printf("Decompress: Failed to write to file\n"); + goto cleanup; + } + src_ptr += src_size; + src_size = src_end - src_ptr; + } + + /* Update input */ + + } + + printf("Wrote %zu bytes\n", bytes_written); + + cleanup: + free(src); + free(dst); + + return ret; +} +#endif + +static size_t compress(const char *fname, const char *oname) { + int fdin, fdout; + struct stat statbuf; + char *src, *dst; + + /* open the input file */ + if ((fdin = open(fname, O_RDONLY)) < 0) { + perror("Error in file opening"); + return 1; + } + + /* open the output file */ + if ((fdout = open(oname, O_RDWR | O_CREAT | O_TRUNC, (mode_t)0600)) < 0) { + perror("Can't create output file"); + return 1; + } + + /* find size of input file */ + if (fstat (fdin, &statbuf) < 0) { + perror("Fstat error"); + return 1; + } + size_t size_in = statbuf.st_size; + + /* go to the location corresponding to the last byte */ + if (lseek(fdout, size_in + LDM_HEADER_SIZE - 1, SEEK_SET) == -1) { + perror("lseek error"); + return 1; + } + + /* write a dummy byte at the last location */ + if (write(fdout, "", 1) != 1) { + perror("write error"); + return 1; + } + + /* mmap the input file */ + if ((src = mmap(0, statbuf.st_size, PROT_READ, MAP_SHARED, fdin, 0)) + == (caddr_t) - 1) { + perror("mmap error for input"); + return 1; + } + size_t out_size = statbuf.st_size + LDM_HEADER_SIZE; + + /* mmap the output file */ + if ((dst = mmap(0, out_size, PROT_READ | PROT_WRITE, + MAP_SHARED, fdout, 0)) == (caddr_t) - 1) { + perror("mmap error for output"); + return 1; + } + + #ifdef ZSTD + size_t size_out = ZSTD_compress(dst, statbuf.st_size, + src, statbuf.st_size, 1); + #else + size_t size_out = LDM_compress(src, dst + LDM_HEADER_SIZE, statbuf.st_size, + statbuf.st_size); + size_out += LDM_HEADER_SIZE; + + // TODO: should depend on LDM_DECOMPRESS_SIZE write32 + memcpy(dst, &size_out, 4); + memcpy(dst + 4, &(statbuf.st_size), 4); + printf("Compressed size: %zu\n", size_out); + printf("Decompressed size: %zu\n", statbuf.st_size); + #endif + ftruncate(fdout, size_out); + + printf("%25s : %6u -> %7u - %s (%.1f%%)\n", fname, + (unsigned)statbuf.st_size, (unsigned)size_out, oname, + (double)size_out / (statbuf.st_size) * 100); + + close(fdin); + close(fdout); + return 0; +} + +static size_t decompress(const char *fname, const char *oname) { + int fdin, fdout; + struct stat statbuf; + char *src, *dst; + + /* open the input file */ + if ((fdin = open(fname, O_RDONLY)) < 0) { + perror("Error in file opening"); + return 1; + } + + /* open the output file */ + if ((fdout = open(oname, O_RDWR | O_CREAT | O_TRUNC, (mode_t)0600)) < 0) { + perror("Can't create output file"); + return 1; + } + + /* find size of input file */ + if (fstat (fdin, &statbuf) < 0) { + perror("Fstat error"); + return 1; + } + + /* mmap the input file */ + if ((src = mmap(0, statbuf.st_size, PROT_READ, MAP_SHARED, fdin, 0)) + == (caddr_t) - 1) { + perror("mmap error for input"); + return 1; + } + + /* read header */ + size_t compressed_size, decompressed_size; + LDM_read_header(src, &compressed_size, &decompressed_size); + + printf("Size, compressed_size, decompressed_size: %zu %zu %zu\n", + statbuf.st_size, compressed_size, decompressed_size); + + /* go to the location corresponding to the last byte */ + if (lseek(fdout, decompressed_size - 1, SEEK_SET) == -1) { + perror("lseek error"); + return 1; + } + + /* write a dummy byte at the last location */ + if (write(fdout, "", 1) != 1) { + perror("write error"); + return 1; + } + + /* mmap the output file */ + if ((dst = mmap(0, decompressed_size, PROT_READ | PROT_WRITE, + MAP_SHARED, fdout, 0)) == (caddr_t) - 1) { + perror("mmap error for output"); + return 1; + } + + /* Copy input file to output file */ +// memcpy(dst, src, statbuf.st_size); + + #ifdef ZSTD + size_t size_out = ZSTD_decompress(dst, decomrpessed_size, + src + LDM_HEADER_SIZE, + statbuf.st_size - LDM_HEADER_SIZE); + #else + size_t size_out = LDM_decompress(src + LDM_HEADER_SIZE, dst, + statbuf.st_size - LDM_HEADER_SIZE, + decompressed_size); + printf("Ret size out: %zu\n", size_out); + #endif + ftruncate(fdout, size_out); + + close(fdin); + close(fdout); + return 0; +} + +static int compare(FILE *fp0, FILE *fp1) { + int result = 0; + while (result == 0) { + char b0[1024]; + char b1[1024]; + const size_t r0 = fread(b0, 1, sizeof(b0), fp0); + const size_t r1 = fread(b1, 1, sizeof(b1), fp1); + + result = (int)r0 - (int)r1; + + if (0 == r0 || 0 == r1) { + break; + } + if (0 == result) { + result = memcmp(b0, b1, r0); + } + } + return result; +} + +static void verify(const char *inpFilename, const char *decFilename) { + FILE *inpFp = fopen(inpFilename, "rb"); + FILE *decFp = fopen(decFilename, "rb"); + + printf("verify : %s <-> %s\n", inpFilename, decFilename); + const int cmp = compare(inpFp, decFp); + if(0 == cmp) { + printf("verify : OK\n"); + } else { + printf("verify : NG\n"); + } + + fclose(decFp); + fclose(inpFp); +} + +int main(int argc, const char *argv[]) { + const char * const exeName = argv[0]; + char inpFilename[256] = { 0 }; + char ldmFilename[256] = { 0 }; + char decFilename[256] = { 0 }; + + if (argc < 2) { + printf("Wrong arguments\n"); + printf("Usage:\n"); + printf("%s FILE\n", exeName); + return 1; + } + + snprintf(inpFilename, 256, "%s", argv[1]); + snprintf(ldmFilename, 256, "%s.ldm", argv[1]); + snprintf(decFilename, 256, "%s.ldm.dec", argv[1]); + + printf("inp = [%s]\n", inpFilename); + printf("ldm = [%s]\n", ldmFilename); + printf("dec = [%s]\n", decFilename); + + struct timeval tv1, tv2; + /* compress */ + { + gettimeofday(&tv1, NULL); + if (compress(inpFilename, ldmFilename)) { + printf("Compress error"); + return 1; + } + gettimeofday(&tv2, NULL); + printf("Total time = %f seconds\n", + (double) (tv2.tv_usec - tv1.tv_usec) / 1000000 + + (double) (tv2.tv_sec - tv1.tv_sec)); + } + + /* decompress */ + + gettimeofday(&tv1, NULL); + if (decompress(ldmFilename, decFilename)) { + printf("Decompress error"); + return 1; + } + gettimeofday(&tv2, NULL); + printf("Total time = %f seconds\n", + (double) (tv2.tv_usec - tv1.tv_usec) / 1000000 + + (double) (tv2.tv_sec - tv1.tv_sec)); + + /* verify */ + verify(inpFilename, decFilename); + return 0; +} + +#if 0 +int main2(int argc, char *argv[]) { + char inpFilename[256] = { 0 }; + char ldmFilename[256] = { 0 }; + char decFilename[256] = { 0 }; + + if (argc < 2) { + printf("Please specify input filename\n"); + return 0; + } + snprintf(inpFilename, 256, "%s", argv[1]); + snprintf(ldmFilename, 256, "%s.ldm", argv[1]); + snprintf(decFilename, 256, "%s.ldm.dec", argv[1]); + + printf("inp = [%s]\n", inpFilename); + printf("ldm = [%s]\n", ldmFilename); + printf("dec = [%s]\n", decFilename); + + /* compress */ + { + FILE *inpFp = fopen(inpFilename, "rb"); + FILE *outFp = fopen(ldmFilename, "wb"); + size_t sizeIn = 0; + size_t sizeOut = 0; + size_t ret; + printf("compress : %s -> %s\n", inpFilename, ldmFilename); + ret = compress_file(inpFp, outFp, &sizeIn, &sizeOut); + if (ret) { + printf("compress : failed with code %zu\n", ret); + return ret; + } + printf("%s: %zu → %zu bytes, %.1f%%\n", + inpFilename, sizeIn, sizeOut, + (double)sizeOut / sizeIn * 100); + printf("compress : done\n"); + + fclose(outFp); + fclose(inpFp); + } + + /* decompress */ + { + FILE *inpFp = fopen(ldmFilename, "rb"); + FILE *outFp = fopen(decFilename, "wb"); + size_t ret; + + printf("decompress : %s -> %s\n", ldmFilename, decFilename); + ret = decompress_file(inpFp, outFp); + if (ret) { + printf("decompress : failed with code %zu\n", ret); + return ret; + } + printf("decompress : done\n"); + + fclose(outFp); + fclose(inpFp); + } + + /* verify */ + { + FILE *inpFp = fopen(inpFilename, "rb"); + FILE *decFp = fopen(decFilename, "rb"); + + printf("verify : %s <-> %s\n", inpFilename, decFilename); + const int cmp = compare(inpFp, decFp); + if(0 == cmp) { + printf("verify : OK\n"); + } else { + printf("verify : NG\n"); + } + + fclose(decFp); + fclose(inpFp); + } + return 0; +} +#endif + diff --git a/contrib/long_distance_matching/versions/v0.2/Makefile b/contrib/long_distance_matching/versions/v0.2/Makefile new file mode 100644 index 000000000..4e04fd6a2 --- /dev/null +++ b/contrib/long_distance_matching/versions/v0.2/Makefile @@ -0,0 +1,32 @@ +# ################################################################ +# Copyright (c) 2016-present, Yann Collet, Facebook, Inc. +# All rights reserved. +# +# This source code is licensed under the BSD-style license found in the +# LICENSE file in the root directory of this source tree. An additional grant +# of patent rights can be found in the PATENTS file in the same directory. +# ################################################################ + +# This Makefile presumes libzstd is installed, using `sudo make install` + + +LDFLAGS += -lzstd + +.PHONY: default all clean + +default: all + +all: main-ldm + + +#main : ldm.c main.c +# $(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@ + +main-ldm : ldm.c main-ldm.c + $(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@ + +clean: + @rm -f core *.o tmp* result* *.ldm *.ldm.dec \ + main main-ldm + @echo Cleaning completed + diff --git a/contrib/long_distance_matching/versions/v0.2/ldm.c b/contrib/long_distance_matching/versions/v0.2/ldm.c new file mode 100644 index 000000000..9081d1362 --- /dev/null +++ b/contrib/long_distance_matching/versions/v0.2/ldm.c @@ -0,0 +1,436 @@ +#include +#include +#include +#include + +#include "ldm.h" + +#define HASH_EVERY 7 + +#define LDM_MEMORY_USAGE 14 +#define LDM_HASHLOG (LDM_MEMORY_USAGE-2) +#define LDM_HASHTABLESIZE (1 << (LDM_MEMORY_USAGE)) +#define LDM_HASHTABLESIZE_U32 ((LDM_HASHTABLESIZE) >> 2) +#define LDM_HASH_SIZE_U32 (1 << (LDM_HASHLOG)) + +#define WINDOW_SIZE (1 << 20) +#define MAX_WINDOW_SIZE 31 +#define HASH_SIZE 8 +#define MINMATCH 8 + +#define ML_BITS 4 +#define ML_MASK ((1U<>8); + } +} + +static U32 LDM_read32(const void *ptr) { + return *(const U32 *)ptr; +} + +static U64 LDM_read64(const void *ptr) { + return *(const U64 *)ptr; +} + +static void LDM_copy8(void *dst, const void *src) { + memcpy(dst, src, 8); +} + +typedef struct compress_stats { + U32 num_matches; + U32 total_match_length; + U32 total_literal_length; + U64 total_offset; +} compress_stats; + +static void LDM_printCompressStats(const compress_stats *stats) { + printf("=====================\n"); + printf("Compression statistics\n"); + printf("Total number of matches: %u\n", stats->num_matches); + printf("Average match length: %.1f\n", ((double)stats->total_match_length) / + (double)stats->num_matches); + printf("Average literal length: %.1f\n", + ((double)stats->total_literal_length) / (double)stats->num_matches); + printf("Average offset length: %.1f\n", + ((double)stats->total_offset) / (double)stats->num_matches); + printf("=====================\n"); +} + +// TODO: unused. +struct hash_entry { + U64 offset; + tag t; +}; + +static U32 LDM_hash(U32 sequence) { + return ((sequence * 2654435761U) >> ((32)-LDM_HASHLOG)); +} + +static U32 LDM_hash5(U64 sequence) { + static const U64 prime5bytes = 889523592379ULL; + static const U64 prime8bytes = 11400714785074694791ULL; + const U32 hashLog = LDM_HASHLOG; + if (LDM_isLittleEndian()) + return (U32)(((sequence << 24) * prime5bytes) >> (64 - hashLog)); + else + return (U32)(((sequence >> 24) * prime8bytes) >> (64 - hashLog)); +} + +static U32 LDM_hash_position(const void * const p) { + return LDM_hash(LDM_read32(p)); +} + +static void LDM_put_position_on_hash(const BYTE *p, U32 h, void *tableBase, + const BYTE *srcBase) { + if (((p - srcBase) & HASH_EVERY) != HASH_EVERY) { + return; + } + + U32 *hashTable = (U32 *) tableBase; + hashTable[h] = (U32)(p - srcBase); +} + +static void LDM_put_position(const BYTE *p, void *tableBase, + const BYTE *srcBase) { + if (((p - srcBase) & HASH_EVERY) != HASH_EVERY) { + return; + } + U32 const h = LDM_hash_position(p); + LDM_put_position_on_hash(p, h, tableBase, srcBase); +} + +static const BYTE *LDM_get_position_on_hash( + U32 h, void *tableBase, const BYTE *srcBase) { + const U32 * const hashTable = (U32*)tableBase; + return hashTable[h] + srcBase; +} + +static BYTE LDM_read_byte(const void *memPtr) { + BYTE val; + memcpy(&val, memPtr, 1); + return val; +} + +static unsigned LDM_count(const BYTE *pIn, const BYTE *pMatch, + const BYTE *pInLimit) { + const BYTE * const pStart = pIn; + while (pIn < pInLimit - 1) { + BYTE const diff = LDM_read_byte(pMatch) ^ LDM_read_byte(pIn); + if (!diff) { + pIn++; + pMatch++; + continue; + } + return (unsigned)(pIn - pStart); + } + return (unsigned)(pIn - pStart); +} + +void LDM_read_header(const void *src, size_t *compressSize, + size_t *decompressSize) { + const U32 *ip = (const U32 *)src; + *compressSize = *ip++; + *decompressSize = *ip; +} + +// TODO: maxDstSize is unused +size_t LDM_compress(const void *src, size_t srcSize, + void *dst, size_t maxDstSize) { + const BYTE * const istart = (const BYTE*)src; + const BYTE *ip = istart; + const BYTE * const iend = istart + srcSize; + const BYTE *ilimit = iend - HASH_SIZE; + const BYTE * const matchlimit = iend - HASH_SIZE; + const BYTE * const mflimit = iend - MINMATCH; + BYTE *op = (BYTE*) dst; + + compress_stats compressStats = { 0 }; + + U32 hashTable[LDM_HASHTABLESIZE_U32]; + memset(hashTable, 0, sizeof(hashTable)); + + const BYTE *anchor = (const BYTE *)src; +// struct LDM_cctx cctx; + size_t output_size = 0; + + U32 forwardH; + + /* Hash first byte: put into hash table */ + + LDM_put_position(ip, hashTable, istart); + const BYTE *lastHash = ip; + ip++; + forwardH = LDM_hash_position(ip); + + //TODO Loop terminates before ip>=ilimit. + while (ip < ilimit) { + const BYTE *match; + BYTE *token; + + /* Find a match */ + { + const BYTE *forwardIp = ip; + unsigned step = 1; + + do { + U32 const h = forwardH; + ip = forwardIp; + forwardIp += step; + + if (forwardIp > mflimit) { + goto _last_literals; + } + + match = LDM_get_position_on_hash(h, hashTable, istart); + + forwardH = LDM_hash_position(forwardIp); + LDM_put_position_on_hash(ip, h, hashTable, istart); + lastHash = ip; + } while (ip - match > WINDOW_SIZE || + LDM_read64(match) != LDM_read64(ip)); + } + compressStats.num_matches++; + + /* Catchup: look back to extend match from found match */ + while (ip > anchor && match > istart && ip[-1] == match[-1]) { + ip--; + match--; + } + + /* Encode literals */ + { + unsigned const litLength = (unsigned)(ip - anchor); + token = op++; + + compressStats.total_literal_length += litLength; + +#ifdef LDM_DEBUG + printf("Cur position: %zu\n", anchor - istart); + printf("LitLength %zu. (Match offset). %zu\n", litLength, ip - match); +#endif + + if (litLength >= RUN_MASK) { + int len = (int)litLength - RUN_MASK; + *token = (RUN_MASK << ML_BITS); + for (; len >= 255; len -= 255) { + *op++ = 255; + } + *op++ = (BYTE)len; + } else { + *token = (BYTE)(litLength << ML_BITS); + } +#ifdef LDM_DEBUG + printf("Literals "); + fwrite(anchor, litLength, 1, stdout); + printf("\n"); +#endif + memcpy(op, anchor, litLength); + op += litLength; + } +_next_match: + /* Encode offset */ + { + /* + LDM_writeLE16(op, ip-match); + op += 2; + */ + LDM_write32(op, ip - match); + op += 4; + compressStats.total_offset += (ip - match); + } + + /* Encode Match Length */ + { + unsigned matchCode; + matchCode = LDM_count(ip + MINMATCH, match + MINMATCH, + matchlimit); +#ifdef LDM_DEBUG + printf("Match length %zu\n", matchCode + MINMATCH); + fwrite(ip, MINMATCH + matchCode, 1, stdout); + printf("\n"); +#endif + compressStats.total_match_length += matchCode + MINMATCH; + unsigned ctr = 1; + ip++; + for (; ctr < MINMATCH + matchCode; ip++, ctr++) { + LDM_put_position(ip, hashTable, istart); + } +// ip += MINMATCH + matchCode; + if (matchCode >= ML_MASK) { + *token += ML_MASK; + matchCode -= ML_MASK; + LDM_write32(op, 0xFFFFFFFF); + while (matchCode >= 4*0xFF) { + op += 4; + LDM_write32(op, 0xffffffff); + matchCode -= 4*0xFF; + } + op += matchCode / 255; + *op++ = (BYTE)(matchCode % 255); + } else { + *token += (BYTE)(matchCode); + } +#ifdef LDM_DEBUG + printf("\n"); + +#endif + } + + anchor = ip; + + LDM_put_position(ip, hashTable, istart); + forwardH = LDM_hash_position(++ip); + lastHash = ip; + } +_last_literals: + /* Encode last literals */ + { + size_t const lastRun = (size_t)(iend - anchor); + if (lastRun >= RUN_MASK) { + size_t accumulator = lastRun - RUN_MASK; + *op++ = RUN_MASK << ML_BITS; + for(; accumulator >= 255; accumulator -= 255) { + *op++ = 255; + } + *op++ = (BYTE)accumulator; + } else { + *op++ = (BYTE)(lastRun << ML_BITS); + } + memcpy(op, anchor, lastRun); + op += lastRun; + } + LDM_printCompressStats(&compressStats); + return (op - (BYTE *)dst); +} + +typedef struct LDM_DCtx { + const BYTE * const ibase; /* Pointer to base of input */ + const BYTE *ip; /* Pointer to current input position */ + const BYTE *iend; /* End of source */ + BYTE *op; /* Pointer to output */ + const BYTE * const oend; /* Pointer to end of output */ + +} LDM_DCtx; + +size_t LDM_decompress(const void *src, size_t compressed_size, + void *dst, size_t max_decompressed_size) { + const BYTE *ip = (const BYTE *)src; + const BYTE * const iend = ip + compressed_size; + BYTE *op = (BYTE *)dst; + BYTE * const oend = op + max_decompressed_size; + BYTE *cpy; + + while (ip < iend) { + size_t length; + const BYTE *match; + size_t offset; + + /* get literal length */ + unsigned const token = *ip++; + if ((length=(token >> ML_BITS)) == RUN_MASK) { + unsigned s; + do { + s = *ip++; + length += s; + } while (s == 255); + } +#ifdef LDM_DEBUG + printf("Literal length: %zu\n", length); +#endif + + /* copy literals */ + cpy = op + length; +#ifdef LDM_DEBUG + printf("Literals "); + fwrite(ip, length, 1, stdout); + printf("\n"); +#endif + memcpy(op, ip, length); + ip += length; + op = cpy; + + /* get offset */ + /* + offset = LDM_readLE16(ip); + ip += 2; + */ + offset = LDM_read32(ip); + ip += 4; +#ifdef LDM_DEBUG + printf("Offset: %zu\n", offset); +#endif + match = op - offset; + // LDM_write32(op, (U32)offset); + + /* get matchlength */ + length = token & ML_MASK; + if (length == ML_MASK) { + unsigned s; + do { + s = *ip++; + length += s; + } while (s == 255); + } + length += MINMATCH; +#ifdef LDM_DEBUG + printf("Match length: %zu\n", length); +#endif + /* copy match */ + cpy = op + length; + + // Inefficient for now + while (match < cpy - offset && op < oend) { + *op++ = *match++; + } + } + return op - (BYTE *)dst; +} + + diff --git a/contrib/long_distance_matching/versions/v0.2/ldm.h b/contrib/long_distance_matching/versions/v0.2/ldm.h new file mode 100644 index 000000000..0ac7b2ece --- /dev/null +++ b/contrib/long_distance_matching/versions/v0.2/ldm.h @@ -0,0 +1,19 @@ +#ifndef LDM_H +#define LDM_H + +#include /* size_t */ + +#define LDM_COMPRESS_SIZE 4 +#define LDM_DECOMPRESS_SIZE 4 +#define LDM_HEADER_SIZE ((LDM_COMPRESS_SIZE)+(LDM_DECOMPRESS_SIZE)) + +size_t LDM_compress(const void *src, size_t srcSize, + void *dst, size_t maxDstSize); + +size_t LDM_decompress(const void *src, size_t srcSize, + void *dst, size_t maxDstSize); + +void LDM_read_header(const void *src, size_t *compressSize, + size_t *decompressSize); + +#endif /* LDM_H */ diff --git a/contrib/long_distance_matching/versions/v0.2/main-ldm.c b/contrib/long_distance_matching/versions/v0.2/main-ldm.c new file mode 100644 index 000000000..0017335b8 --- /dev/null +++ b/contrib/long_distance_matching/versions/v0.2/main-ldm.c @@ -0,0 +1,474 @@ +// TODO: file size must fit into a U32 + +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include "ldm.h" + +// #define BUF_SIZE 16*1024 // Block size +#define DEBUG + +//#define ZSTD + +/* Compress file given by fname and output to oname. + * Returns 0 if successful, error code otherwise. + */ +static int compress(const char *fname, const char *oname) { + int fdin, fdout; + struct stat statbuf; + char *src, *dst; + + /* Open the input file. */ + if ((fdin = open(fname, O_RDONLY)) < 0) { + perror("Error in file opening"); + return 1; + } + + /* Open the output file. */ + if ((fdout = open(oname, O_RDWR | O_CREAT | O_TRUNC, (mode_t)0600)) < 0) { + perror("Can't create output file"); + return 1; + } + + /* Find the size of the input file. */ + if (fstat (fdin, &statbuf) < 0) { + perror("Fstat error"); + return 1; + } + + size_t maxCompressSize = statbuf.st_size + LDM_HEADER_SIZE; + + /* Go to the location corresponding to the last byte. */ + /* TODO: fallocate? */ + if (lseek(fdout, maxCompressSize - 1, SEEK_SET) == -1) { + perror("lseek error"); + return 1; + } + + /* Write a dummy byte at the last location. */ + if (write(fdout, "", 1) != 1) { + perror("write error"); + return 1; + } + + /* mmap the input file. */ + if ((src = mmap(0, statbuf.st_size, PROT_READ, MAP_SHARED, fdin, 0)) + == (caddr_t) - 1) { + perror("mmap error for input"); + return 1; + } + + /* mmap the output file */ + if ((dst = mmap(0, maxCompressSize, PROT_READ | PROT_WRITE, + MAP_SHARED, fdout, 0)) == (caddr_t) - 1) { + perror("mmap error for output"); + return 1; + } + +#ifdef ZSTD + size_t compressSize = ZSTD_compress(dst, statbuf.st_size, + src, statbuf.st_size, 1); +#else + size_t compressSize = LDM_HEADER_SIZE + + LDM_compress(src, statbuf.st_size, + dst + LDM_HEADER_SIZE, statbuf.st_size); + + // Write compress and decompress size to header + // TODO: should depend on LDM_DECOMPRESS_SIZE write32 + memcpy(dst, &compressSize, 4); + memcpy(dst + 4, &(statbuf.st_size), 4); + +#ifdef DEBUG + printf("Compressed size: %zu\n", compressSize); + printf("Decompressed size: %zu\n", statbuf.st_size); +#endif +#endif + + // Truncate file to compressSize. + ftruncate(fdout, compressSize); + + printf("%25s : %6u -> %7u - %s (%.1f%%)\n", fname, + (unsigned)statbuf.st_size, (unsigned)compressSize, oname, + (double)compressSize / (statbuf.st_size) * 100); + + // Close files. + close(fdin); + close(fdout); + return 0; +} + +/* Decompress file compressed using LDM_compress. + * The input file should have the LDM_HEADER followed by payload. + * Returns 0 if succesful, and an error code otherwise. + */ +static int decompress(const char *fname, const char *oname) { + int fdin, fdout; + struct stat statbuf; + char *src, *dst; + + /* Open the input file. */ + if ((fdin = open(fname, O_RDONLY)) < 0) { + perror("Error in file opening"); + return 1; + } + + /* Open the output file. */ + if ((fdout = open(oname, O_RDWR | O_CREAT | O_TRUNC, (mode_t)0600)) < 0) { + perror("Can't create output file"); + return 1; + } + + /* Find the size of the input file. */ + if (fstat (fdin, &statbuf) < 0) { + perror("Fstat error"); + return 1; + } + + /* mmap the input file. */ + if ((src = mmap(0, statbuf.st_size, PROT_READ, MAP_SHARED, fdin, 0)) + == (caddr_t) - 1) { + perror("mmap error for input"); + return 1; + } + + /* Read the header. */ + size_t compressSize, decompressSize; + LDM_read_header(src, &compressSize, &decompressSize); + +#ifdef DEBUG + printf("Size, compressSize, decompressSize: %zu %zu %zu\n", + statbuf.st_size, compressSize, decompressSize); +#endif + + /* Go to the location corresponding to the last byte. */ + if (lseek(fdout, decompressSize - 1, SEEK_SET) == -1) { + perror("lseek error"); + return 1; + } + + /* write a dummy byte at the last location */ + if (write(fdout, "", 1) != 1) { + perror("write error"); + return 1; + } + + /* mmap the output file */ + if ((dst = mmap(0, decompressSize, PROT_READ | PROT_WRITE, + MAP_SHARED, fdout, 0)) == (caddr_t) - 1) { + perror("mmap error for output"); + return 1; + } + +#ifdef ZSTD + size_t outSize = ZSTD_decompress(dst, decomrpessed_size, + src + LDM_HEADER_SIZE, + statbuf.st_size - LDM_HEADER_SIZE); +#else + size_t outSize = LDM_decompress( + src + LDM_HEADER_SIZE, statbuf.st_size - LDM_HEADER_SIZE, + dst, decompressSize); + + printf("Ret size out: %zu\n", outSize); + #endif + ftruncate(fdout, outSize); + + close(fdin); + close(fdout); + return 0; +} + +/* Compare two files. + * Returns 0 iff they are the same. + */ +static int compare(FILE *fp0, FILE *fp1) { + int result = 0; + while (result == 0) { + char b0[1024]; + char b1[1024]; + const size_t r0 = fread(b0, 1, sizeof(b0), fp0); + const size_t r1 = fread(b1, 1, sizeof(b1), fp1); + + result = (int)r0 - (int)r1; + + if (0 == r0 || 0 == r1) break; + + if (0 == result) result = memcmp(b0, b1, r0); + } + return result; +} + +/* Verify the input file is the same as the decompressed file. */ +static void verify(const char *inpFilename, const char *decFilename) { + FILE *inpFp = fopen(inpFilename, "rb"); + FILE *decFp = fopen(decFilename, "rb"); + + printf("verify : %s <-> %s\n", inpFilename, decFilename); + const int cmp = compare(inpFp, decFp); + if(0 == cmp) { + printf("verify : OK\n"); + } else { + printf("verify : NG\n"); + } + + fclose(decFp); + fclose(inpFp); +} + +int main(int argc, const char *argv[]) { + const char * const exeName = argv[0]; + char inpFilename[256] = { 0 }; + char ldmFilename[256] = { 0 }; + char decFilename[256] = { 0 }; + + if (argc < 2) { + printf("Wrong arguments\n"); + printf("Usage:\n"); + printf("%s FILE\n", exeName); + return 1; + } + + snprintf(inpFilename, 256, "%s", argv[1]); + snprintf(ldmFilename, 256, "%s.ldm", argv[1]); + snprintf(decFilename, 256, "%s.ldm.dec", argv[1]); + + printf("inp = [%s]\n", inpFilename); + printf("ldm = [%s]\n", ldmFilename); + printf("dec = [%s]\n", decFilename); + + struct timeval tv1, tv2; + + /* Compress */ + + gettimeofday(&tv1, NULL); + if (compress(inpFilename, ldmFilename)) { + printf("Compress error"); + return 1; + } + gettimeofday(&tv2, NULL); + printf("Total time = %f seconds\n", + (double) (tv2.tv_usec - tv1.tv_usec) / 1000000 + + (double) (tv2.tv_sec - tv1.tv_sec)); + + /* Decompress */ + + gettimeofday(&tv1, NULL); + if (decompress(ldmFilename, decFilename)) { + printf("Decompress error"); + return 1; + } + gettimeofday(&tv2, NULL); + printf("Total time = %f seconds\n", + (double) (tv2.tv_usec - tv1.tv_usec) / 1000000 + + (double) (tv2.tv_sec - tv1.tv_sec)); + + /* verify */ + verify(inpFilename, decFilename); + return 0; +} + + +#if 0 +static size_t compress_file(FILE *in, FILE *out, size_t *size_in, + size_t *size_out) { + char *src, *buf = NULL; + size_t r = 1; + size_t size, n, k, count_in = 0, count_out = 0, offset, frame_size = 0; + + src = malloc(BUF_SIZE); + if (!src) { + printf("Not enough memory\n"); + goto cleanup; + } + + size = BUF_SIZE + LDM_HEADER_SIZE; + buf = malloc(size); + if (!buf) { + printf("Not enough memory\n"); + goto cleanup; + } + + + for (;;) { + k = fread(src, 1, BUF_SIZE, in); + if (k == 0) + break; + count_in += k; + + n = LDM_compress(src, buf, k, BUF_SIZE); + + // n = k; + // offset += n; + offset = k; + count_out += k; + +// k = fwrite(src, 1, offset, out); + + k = fwrite(buf, 1, offset, out); + if (k < offset) { + if (ferror(out)) + printf("Write failed\n"); + else + printf("Short write\n"); + goto cleanup; + } + + } + *size_in = count_in; + *size_out = count_out; + r = 0; + cleanup: + free(src); + free(buf); + return r; +} + +static size_t decompress_file(FILE *in, FILE *out) { + void *src = malloc(BUF_SIZE); + void *dst = NULL; + size_t dst_capacity = BUF_SIZE; + size_t ret = 1; + size_t bytes_written = 0; + + if (!src) { + perror("decompress_file(src)"); + goto cleanup; + } + + while (ret != 0) { + /* Load more input */ + size_t src_size = fread(src, 1, BUF_SIZE, in); + void *src_ptr = src; + void *src_end = src_ptr + src_size; + if (src_size == 0 || ferror(in)) { + printf("(TODO): Decompress: not enough input or error reading file\n"); + //TODO + ret = 0; + goto cleanup; + } + + /* Allocate destination buffer if it hasn't been allocated already */ + if (!dst) { + dst = malloc(dst_capacity); + if (!dst) { + perror("decompress_file(dst)"); + goto cleanup; + } + } + + // TODO + + /* Decompress: + * Continue while there is more input to read. + */ + while (src_ptr != src_end && ret != 0) { + // size_t dst_size = src_size; + size_t dst_size = LDM_decompress(src, dst, src_size, dst_capacity); + size_t written = fwrite(dst, 1, dst_size, out); +// printf("Writing %zu bytes\n", dst_size); + bytes_written += dst_size; + if (written != dst_size) { + printf("Decompress: Failed to write to file\n"); + goto cleanup; + } + src_ptr += src_size; + src_size = src_end - src_ptr; + } + + /* Update input */ + + } + + printf("Wrote %zu bytes\n", bytes_written); + + cleanup: + free(src); + free(dst); + + return ret; +} + +int main2(int argc, char *argv[]) { + char inpFilename[256] = { 0 }; + char ldmFilename[256] = { 0 }; + char decFilename[256] = { 0 }; + + if (argc < 2) { + printf("Please specify input filename\n"); + return 0; + } + snprintf(inpFilename, 256, "%s", argv[1]); + snprintf(ldmFilename, 256, "%s.ldm", argv[1]); + snprintf(decFilename, 256, "%s.ldm.dec", argv[1]); + + printf("inp = [%s]\n", inpFilename); + printf("ldm = [%s]\n", ldmFilename); + printf("dec = [%s]\n", decFilename); + + /* compress */ + { + FILE *inpFp = fopen(inpFilename, "rb"); + FILE *outFp = fopen(ldmFilename, "wb"); + size_t sizeIn = 0; + size_t sizeOut = 0; + size_t ret; + printf("compress : %s -> %s\n", inpFilename, ldmFilename); + ret = compress_file(inpFp, outFp, &sizeIn, &sizeOut); + if (ret) { + printf("compress : failed with code %zu\n", ret); + return ret; + } + printf("%s: %zu → %zu bytes, %.1f%%\n", + inpFilename, sizeIn, sizeOut, + (double)sizeOut / sizeIn * 100); + printf("compress : done\n"); + + fclose(outFp); + fclose(inpFp); + } + + /* decompress */ + { + FILE *inpFp = fopen(ldmFilename, "rb"); + FILE *outFp = fopen(decFilename, "wb"); + size_t ret; + + printf("decompress : %s -> %s\n", ldmFilename, decFilename); + ret = decompress_file(inpFp, outFp); + if (ret) { + printf("decompress : failed with code %zu\n", ret); + return ret; + } + printf("decompress : done\n"); + + fclose(outFp); + fclose(inpFp); + } + + /* verify */ + { + FILE *inpFp = fopen(inpFilename, "rb"); + FILE *decFp = fopen(decFilename, "rb"); + + printf("verify : %s <-> %s\n", inpFilename, decFilename); + const int cmp = compare(inpFp, decFp); + if(0 == cmp) { + printf("verify : OK\n"); + } else { + printf("verify : NG\n"); + } + + fclose(decFp); + fclose(inpFp); + } + return 0; +} +#endif + diff --git a/contrib/long_distance_matching/versions/v0.3/Makefile b/contrib/long_distance_matching/versions/v0.3/Makefile new file mode 100644 index 000000000..5ffd4eafe --- /dev/null +++ b/contrib/long_distance_matching/versions/v0.3/Makefile @@ -0,0 +1,40 @@ +# ################################################################ +# Copyright (c) 2016-present, Yann Collet, Facebook, Inc. +# All rights reserved. +# +# This source code is licensed under the BSD-style license found in the +# LICENSE file in the root directory of this source tree. An additional grant +# of patent rights can be found in the PATENTS file in the same directory. +# ################################################################ + +# This Makefile presumes libzstd is installed, using `sudo make install` + +CFLAGS ?= -O3 +DEBUGFLAGS = -Wall -Wextra -Wcast-qual -Wcast-align -Wshadow \ + -Wstrict-aliasing=1 -Wswitch-enum -Wdeclaration-after-statement \ + -Wstrict-prototypes -Wundef -Wpointer-arith -Wformat-security \ + -Wvla -Wformat=2 -Winit-self -Wfloat-equal -Wwrite-strings \ + -Wredundant-decls +CFLAGS += $(DEBUGFLAGS) $(MOREFLAGS) +FLAGS = $(CPPFLAGS) $(CFLAGS) + +LDFLAGS += -lzstd + +.PHONY: default all clean + +default: all + +all: main-ldm + + +#main : ldm.c main.c +# $(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@ + +main-ldm : util.c ldm.c main-ldm.c + $(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@ + +clean: + @rm -f core *.o tmp* result* *.ldm *.ldm.dec \ + main main-ldm + @echo Cleaning completed + diff --git a/contrib/long_distance_matching/versions/v0.3/ldm.c b/contrib/long_distance_matching/versions/v0.3/ldm.c new file mode 100644 index 000000000..1dedf5c37 --- /dev/null +++ b/contrib/long_distance_matching/versions/v0.3/ldm.c @@ -0,0 +1,464 @@ +#include +#include +#include +#include + + +#include "ldm.h" +#include "util.h" + +#define HASH_EVERY 1 + +#define LDM_MEMORY_USAGE 16 +#define LDM_HASHLOG (LDM_MEMORY_USAGE-2) +#define LDM_HASHTABLESIZE (1 << (LDM_MEMORY_USAGE)) +#define LDM_HASHTABLESIZE_U32 ((LDM_HASHTABLESIZE) >> 2) +#define LDM_HASH_SIZE_U32 (1 << (LDM_HASHLOG)) + +#define LDM_OFFSET_SIZE 4 + +#define WINDOW_SIZE (1 << 20) +#define MAX_WINDOW_SIZE 31 +#define HASH_SIZE 4 +#define LDM_HASH_LENGTH 4 +#define MINMATCH 4 + +#define ML_BITS 4 +#define ML_MASK ((1U<numMatches); + printf("Average match length: %.1f\n", ((double)stats->totalMatchLength) / + (double)stats->numMatches); + printf("Average literal length: %.1f\n", + ((double)stats->totalLiteralLength) / (double)stats->numMatches); + printf("Average offset length: %.1f\n", + ((double)stats->totalOffset) / (double)stats->numMatches); + printf("=====================\n"); +} + +typedef struct LDM_CCtx { + size_t isize; /* Input size */ + size_t maxOSize; /* Maximum output size */ + + const BYTE *ibase; /* Base of input */ + const BYTE *ip; /* Current input position */ + const BYTE *iend; /* End of input */ + + // Maximum input position such that hashing at the position does not exceed + // end of input. + const BYTE *ihashLimit; + + // Maximum input position such that finding a match of at least the minimum + // match length does not exceed end of input. + const BYTE *imatchLimit; + + const BYTE *obase; /* Base of output */ + BYTE *op; /* Output */ + + const BYTE *anchor; /* Anchor to start of current (match) block */ + + LDM_compressStats stats; /* Compression statistics */ + + LDM_hashEntry hashTable[LDM_HASHTABLESIZE_U32]; + + const BYTE *lastPosHashed; /* Last position hashed */ + hash_t lastHash; /* Hash corresponding to lastPosHashed */ + const BYTE *nextIp; + hash_t nextHash; /* Hash corresponding to nextIp */ + + unsigned step; +} LDM_CCtx; + +#ifdef LDM_ROLLING_HASH +/** + * Convert a sum computed from LDM_getRollingHash to a hash value in the range + * of the hash table. + */ +static hash_t LDM_sumToHash(U32 sum) { + return sum % (LDM_HASHTABLESIZE >> 2); +// return sum & (LDM_HASHTABLESIZE - 1); +} + +static U32 LDM_getRollingHash(const char *data, U32 len) { + U32 i; + U32 s1, s2; + const schar *buf = (const schar *)data; + + s1 = s2 = 0; + for (i = 0; i < (len - 4); i += 4) { + s2 += (4 * (s1 + buf[i])) + (3 * buf[i + 1]) + + (2 * buf[i + 2]) + (buf[i + 3]); + s1 += buf[i] + buf[i + 1] + buf[i + 2] + buf[i + 3]; + } + for(; i < len; i++) { + s1 += buf[i]; + s2 += s1; + } + return (s1 & 0xffff) + (s2 << 16); +} + +static hash_t LDM_hashPosition(const void * const p) { + return LDM_sumToHash(LDM_getRollingHash((const char *)p, LDM_HASH_LENGTH)); +} + +typedef struct LDM_sumStruct { + U16 s1, s2; +} LDM_sumStruct; + +static void LDM_getRollingHashParts(U32 sum, LDM_sumStruct *sumStruct) { + sumStruct->s1 = sum & 0xffff; + sumStruct->s2 = sum >> 16; +} + +#else +static hash_t LDM_hash(U32 sequence) { + return ((sequence * 2654435761U) >> ((32)-LDM_HASHLOG)); +} + +static hash_t LDM_hashPosition(const void * const p) { + return LDM_hash(LDM_read32(p)); +} +#endif + +/* +static hash_t LDM_hash5(U64 sequence) { + static const U64 prime5bytes = 889523592379ULL; + static const U64 prime8bytes = 11400714785074694791ULL; + const U32 hashLog = LDM_HASHLOG; + if (LDM_isLittleEndian()) + return (((sequence << 24) * prime5bytes) >> (64 - hashLog)); + else + return (((sequence >> 24) * prime8bytes) >> (64 - hashLog)); +} +*/ + +static void LDM_putHashOfCurrentPositionFromHash( + LDM_CCtx *cctx, hash_t hash) { + if (((cctx->ip - cctx->ibase) & HASH_EVERY) != HASH_EVERY) { + return; + } + (cctx->hashTable)[hash] = (LDM_hashEntry){ (hash_t)(cctx->ip - cctx->ibase) }; + cctx->lastPosHashed = cctx->ip; + cctx->lastHash = hash; +} + +static void LDM_putHashOfCurrentPosition(LDM_CCtx *cctx) { + hash_t hash = LDM_hashPosition(cctx->ip); + LDM_putHashOfCurrentPositionFromHash(cctx, hash); +} + +static const BYTE *LDM_get_position_on_hash( + hash_t h, void *tableBase, const BYTE *srcBase) { + const LDM_hashEntry * const hashTable = (LDM_hashEntry *)tableBase; + return hashTable[h].offset + srcBase; +} + +static BYTE LDM_read_byte(const void *memPtr) { + BYTE val; + memcpy(&val, memPtr, 1); + return val; +} + +static unsigned LDM_count(const BYTE *pIn, const BYTE *pMatch, + const BYTE *pInLimit) { + const BYTE * const pStart = pIn; + while (pIn < pInLimit - 1) { + BYTE const diff = LDM_read_byte(pMatch) ^ LDM_read_byte(pIn); + if (!diff) { + pIn++; + pMatch++; + continue; + } + return (unsigned)(pIn - pStart); + } + return (unsigned)(pIn - pStart); +} + +void LDM_readHeader(const void *src, size_t *compressSize, + size_t *decompressSize) { + const U32 *ip = (const U32 *)src; + *compressSize = *ip++; + *decompressSize = *ip; +} + +static void LDM_initializeCCtx(LDM_CCtx *cctx, + const void *src, size_t srcSize, + void *dst, size_t maxDstSize) { + cctx->isize = srcSize; + cctx->maxOSize = maxDstSize; + + cctx->ibase = (const BYTE *)src; + cctx->ip = cctx->ibase; + cctx->iend = cctx->ibase + srcSize; + + cctx->ihashLimit = cctx->iend - HASH_SIZE; + cctx->imatchLimit = cctx->iend - MINMATCH; + + cctx->obase = (BYTE *)dst; + cctx->op = (BYTE *)dst; + + cctx->anchor = cctx->ibase; + + memset(&(cctx->stats), 0, sizeof(cctx->stats)); + memset(cctx->hashTable, 0, sizeof(cctx->hashTable)); + + cctx->lastPosHashed = NULL; + cctx->nextIp = NULL; + + cctx->step = 1; +} + +static int LDM_findBestMatch(LDM_CCtx *cctx, const BYTE **match) { + cctx->nextIp = cctx->ip; + + do { + hash_t const h = cctx->nextHash; + cctx->ip = cctx->nextIp; + cctx->nextIp += cctx->step; + + if (cctx->nextIp > cctx->imatchLimit) { + return 1; + } + + *match = LDM_get_position_on_hash(h, cctx->hashTable, cctx->ibase); + + cctx->nextHash = LDM_hashPosition(cctx->nextIp); + LDM_putHashOfCurrentPositionFromHash(cctx, h); + } while (cctx->ip - *match > WINDOW_SIZE || + LDM_read64(*match) != LDM_read64(cctx->ip)); + return 0; +} + +// TODO: srcSize and maxDstSize is unused +size_t LDM_compress(const void *src, size_t srcSize, + void *dst, size_t maxDstSize) { + LDM_CCtx cctx; + LDM_initializeCCtx(&cctx, src, srcSize, dst, maxDstSize); + + /* Hash the first position and put it into the hash table. */ + LDM_putHashOfCurrentPosition(&cctx); + cctx.ip++; + cctx.nextHash = LDM_hashPosition(cctx.ip); + + // TODO: loop condition is not accurate. + while (1) { + const BYTE *match; + + /** + * Find a match. + * If no more matches can be found (i.e. the length of the remaining input + * is less than the minimum match length), then stop searching for matches + * and encode the final literals. + */ + if (LDM_findBestMatch(&cctx, &match) != 0) { + goto _last_literals; + } + + cctx.stats.numMatches++; + + /** + * Catch up: look back to extend the match backwards from the found match. + */ + while (cctx.ip > cctx.anchor && match > cctx.ibase && + cctx.ip[-1] == match[-1]) { + cctx.ip--; + match--; + } + + /** + * Write current block (literals, literal length, match offset, match + * length) and update pointers and hashes. + */ + { + unsigned const literalLength = (unsigned)(cctx.ip - cctx.anchor); + unsigned const offset = cctx.ip - match; + unsigned const matchLength = LDM_count( + cctx.ip + MINMATCH, match + MINMATCH, cctx.ihashLimit); + BYTE *token = cctx.op++; + + cctx.stats.totalLiteralLength += literalLength; + cctx.stats.totalOffset += offset; + cctx.stats.totalMatchLength += matchLength + MINMATCH; + + /* Encode the literal length. */ + if (literalLength >= RUN_MASK) { + int len = (int)literalLength - RUN_MASK; + *token = (RUN_MASK << ML_BITS); + for (; len >= 255; len -= 255) { + *(cctx.op)++ = 255; + } + *(cctx.op)++ = (BYTE)len; + } else { + *token = (BYTE)(literalLength << ML_BITS); + } + + /* Encode the literals. */ + memcpy(cctx.op, cctx.anchor, literalLength); + cctx.op += literalLength; + + /* Encode the offset. */ + LDM_write32(cctx.op, offset); + cctx.op += LDM_OFFSET_SIZE; + + /* Encode match length */ + if (matchLength >= ML_MASK) { + unsigned matchLengthRemaining = matchLength; + *token += ML_MASK; + matchLengthRemaining -= ML_MASK; + LDM_write32(cctx.op, 0xFFFFFFFF); + while (matchLengthRemaining >= 4*0xFF) { + cctx.op += 4; + LDM_write32(cctx.op, 0xffffffff); + matchLengthRemaining -= 4*0xFF; + } + cctx.op += matchLengthRemaining / 255; + *(cctx.op)++ = (BYTE)(matchLengthRemaining % 255); + } else { + *token += (BYTE)(matchLength); + } + + /* Update input pointer, inserting hashes into hash table along the + * way. + */ + while (cctx.ip < cctx.anchor + MINMATCH + matchLength + literalLength) { + LDM_putHashOfCurrentPosition(&cctx); + cctx.ip++; + } + } + + // Set start of next block to current input pointer. + cctx.anchor = cctx.ip; + LDM_putHashOfCurrentPosition(&cctx); + cctx.nextHash = LDM_hashPosition(++cctx.ip); + } +_last_literals: + /* Encode the last literals (no more matches). */ + { + size_t const lastRun = (size_t)(cctx.iend - cctx.anchor); + if (lastRun >= RUN_MASK) { + size_t accumulator = lastRun - RUN_MASK; + *(cctx.op)++ = RUN_MASK << ML_BITS; + for(; accumulator >= 255; accumulator -= 255) { + *(cctx.op)++ = 255; + } + *(cctx.op)++ = (BYTE)accumulator; + } else { + *(cctx.op)++ = (BYTE)(lastRun << ML_BITS); + } + memcpy(cctx.op, cctx.anchor, lastRun); + cctx.op += lastRun; + } + LDM_printCompressStats(&cctx.stats); + return (cctx.op - (const BYTE *)cctx.obase); +} + +typedef struct LDM_DCtx { + size_t compressSize; + size_t maxDecompressSize; + + const BYTE *ibase; /* Base of input */ + const BYTE *ip; /* Current input position */ + const BYTE *iend; /* End of source */ + + const BYTE *obase; /* Base of output */ + BYTE *op; /* Current output position */ + const BYTE *oend; /* End of output */ +} LDM_DCtx; + +static void LDM_initializeDCtx(LDM_DCtx *dctx, + const void *src, size_t compressSize, + void *dst, size_t maxDecompressSize) { + dctx->compressSize = compressSize; + dctx->maxDecompressSize = maxDecompressSize; + + dctx->ibase = src; + dctx->ip = (const BYTE *)src; + dctx->iend = dctx->ip + dctx->compressSize; + dctx->op = dst; + dctx->oend = dctx->op + dctx->maxDecompressSize; + +} + +size_t LDM_decompress(const void *src, size_t compressSize, + void *dst, size_t maxDecompressSize) { + LDM_DCtx dctx; + LDM_initializeDCtx(&dctx, src, compressSize, dst, maxDecompressSize); + + while (dctx.ip < dctx.iend) { + BYTE *cpy; + const BYTE *match; + size_t length, offset; + + /* Get the literal length. */ + unsigned const token = *(dctx.ip)++; + if ((length = (token >> ML_BITS)) == RUN_MASK) { + unsigned s; + do { + s = *(dctx.ip)++; + length += s; + } while (s == 255); + } + + /* Copy literals. */ + cpy = dctx.op + length; + memcpy(dctx.op, dctx.ip, length); + dctx.ip += length; + dctx.op = cpy; + + //TODO : dynamic offset size + offset = LDM_read32(dctx.ip); + dctx.ip += LDM_OFFSET_SIZE; + match = dctx.op - offset; + + /* Get the match length. */ + length = token & ML_MASK; + if (length == ML_MASK) { + unsigned s; + do { + s = *(dctx.ip)++; + length += s; + } while (s == 255); + } + length += MINMATCH; + + /* Copy match. */ + cpy = dctx.op + length; + + // Inefficient for now + while (match < cpy - offset && dctx.op < dctx.oend) { + *(dctx.op)++ = *match++; + } + } + return dctx.op - (BYTE *)dst; +} + + diff --git a/contrib/long_distance_matching/versions/v0.3/ldm.h b/contrib/long_distance_matching/versions/v0.3/ldm.h new file mode 100644 index 000000000..287d444dd --- /dev/null +++ b/contrib/long_distance_matching/versions/v0.3/ldm.h @@ -0,0 +1,19 @@ +#ifndef LDM_H +#define LDM_H + +#include /* size_t */ + +#define LDM_COMPRESS_SIZE 4 +#define LDM_DECOMPRESS_SIZE 4 +#define LDM_HEADER_SIZE ((LDM_COMPRESS_SIZE)+(LDM_DECOMPRESS_SIZE)) + +size_t LDM_compress(const void *src, size_t srcSize, + void *dst, size_t maxDstSize); + +size_t LDM_decompress(const void *src, size_t srcSize, + void *dst, size_t maxDstSize); + +void LDM_readHeader(const void *src, size_t *compressSize, + size_t *decompressSize); + +#endif /* LDM_H */ diff --git a/contrib/long_distance_matching/versions/v0.3/main-ldm.c b/contrib/long_distance_matching/versions/v0.3/main-ldm.c new file mode 100644 index 000000000..724d735dd --- /dev/null +++ b/contrib/long_distance_matching/versions/v0.3/main-ldm.c @@ -0,0 +1,479 @@ +// TODO: file size must fit into a U32 + +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include "ldm.h" + +// #define BUF_SIZE 16*1024 // Block size +#define DEBUG + +//#define ZSTD + +/* Compress file given by fname and output to oname. + * Returns 0 if successful, error code otherwise. + */ +static int compress(const char *fname, const char *oname) { + int fdin, fdout; + struct stat statbuf; + char *src, *dst; + size_t maxCompressSize, compressSize; + + /* Open the input file. */ + if ((fdin = open(fname, O_RDONLY)) < 0) { + perror("Error in file opening"); + return 1; + } + + /* Open the output file. */ + if ((fdout = open(oname, O_RDWR | O_CREAT | O_TRUNC, (mode_t)0600)) < 0) { + perror("Can't create output file"); + return 1; + } + + /* Find the size of the input file. */ + if (fstat (fdin, &statbuf) < 0) { + perror("Fstat error"); + return 1; + } + + maxCompressSize = statbuf.st_size + LDM_HEADER_SIZE; + + /* Go to the location corresponding to the last byte. */ + /* TODO: fallocate? */ + if (lseek(fdout, maxCompressSize - 1, SEEK_SET) == -1) { + perror("lseek error"); + return 1; + } + + /* Write a dummy byte at the last location. */ + if (write(fdout, "", 1) != 1) { + perror("write error"); + return 1; + } + + /* mmap the input file. */ + if ((src = mmap(0, statbuf.st_size, PROT_READ, MAP_SHARED, fdin, 0)) + == (caddr_t) - 1) { + perror("mmap error for input"); + return 1; + } + + /* mmap the output file */ + if ((dst = mmap(0, maxCompressSize, PROT_READ | PROT_WRITE, + MAP_SHARED, fdout, 0)) == (caddr_t) - 1) { + perror("mmap error for output"); + return 1; + } + +#ifdef ZSTD + compressSize = ZSTD_compress(dst, statbuf.st_size, + src, statbuf.st_size, 1); +#else + compressSize = LDM_HEADER_SIZE + + LDM_compress(src, statbuf.st_size, + dst + LDM_HEADER_SIZE, statbuf.st_size); + + // Write compress and decompress size to header + // TODO: should depend on LDM_DECOMPRESS_SIZE write32 + memcpy(dst, &compressSize, 4); + memcpy(dst + 4, &(statbuf.st_size), 4); + +#ifdef DEBUG + printf("Compressed size: %zu\n", compressSize); + printf("Decompressed size: %zu\n", (size_t)statbuf.st_size); +#endif +#endif + + // Truncate file to compressSize. + ftruncate(fdout, compressSize); + + printf("%25s : %6u -> %7u - %s (%.1f%%)\n", fname, + (unsigned)statbuf.st_size, (unsigned)compressSize, oname, + (double)compressSize / (statbuf.st_size) * 100); + + // Close files. + close(fdin); + close(fdout); + return 0; +} + +/* Decompress file compressed using LDM_compress. + * The input file should have the LDM_HEADER followed by payload. + * Returns 0 if succesful, and an error code otherwise. + */ +static int decompress(const char *fname, const char *oname) { + int fdin, fdout; + struct stat statbuf; + char *src, *dst; + size_t compressSize, decompressSize, outSize; + + /* Open the input file. */ + if ((fdin = open(fname, O_RDONLY)) < 0) { + perror("Error in file opening"); + return 1; + } + + /* Open the output file. */ + if ((fdout = open(oname, O_RDWR | O_CREAT | O_TRUNC, (mode_t)0600)) < 0) { + perror("Can't create output file"); + return 1; + } + + /* Find the size of the input file. */ + if (fstat (fdin, &statbuf) < 0) { + perror("Fstat error"); + return 1; + } + + /* mmap the input file. */ + if ((src = mmap(0, statbuf.st_size, PROT_READ, MAP_SHARED, fdin, 0)) + == (caddr_t) - 1) { + perror("mmap error for input"); + return 1; + } + + /* Read the header. */ + LDM_readHeader(src, &compressSize, &decompressSize); + +#ifdef DEBUG + printf("Size, compressSize, decompressSize: %zu %zu %zu\n", + (size_t)statbuf.st_size, compressSize, decompressSize); +#endif + + /* Go to the location corresponding to the last byte. */ + if (lseek(fdout, decompressSize - 1, SEEK_SET) == -1) { + perror("lseek error"); + return 1; + } + + /* write a dummy byte at the last location */ + if (write(fdout, "", 1) != 1) { + perror("write error"); + return 1; + } + + /* mmap the output file */ + if ((dst = mmap(0, decompressSize, PROT_READ | PROT_WRITE, + MAP_SHARED, fdout, 0)) == (caddr_t) - 1) { + perror("mmap error for output"); + return 1; + } + +#ifdef ZSTD + outSize = ZSTD_decompress(dst, decomrpessed_size, + src + LDM_HEADER_SIZE, + statbuf.st_size - LDM_HEADER_SIZE); +#else + outSize = LDM_decompress( + src + LDM_HEADER_SIZE, statbuf.st_size - LDM_HEADER_SIZE, + dst, decompressSize); + + printf("Ret size out: %zu\n", outSize); + #endif + ftruncate(fdout, outSize); + + close(fdin); + close(fdout); + return 0; +} + +/* Compare two files. + * Returns 0 iff they are the same. + */ +static int compare(FILE *fp0, FILE *fp1) { + int result = 0; + while (result == 0) { + char b0[1024]; + char b1[1024]; + const size_t r0 = fread(b0, 1, sizeof(b0), fp0); + const size_t r1 = fread(b1, 1, sizeof(b1), fp1); + + result = (int)r0 - (int)r1; + + if (0 == r0 || 0 == r1) break; + + if (0 == result) result = memcmp(b0, b1, r0); + } + return result; +} + +/* Verify the input file is the same as the decompressed file. */ +static void verify(const char *inpFilename, const char *decFilename) { + FILE *inpFp = fopen(inpFilename, "rb"); + FILE *decFp = fopen(decFilename, "rb"); + + printf("verify : %s <-> %s\n", inpFilename, decFilename); + { + const int cmp = compare(inpFp, decFp); + if(0 == cmp) { + printf("verify : OK\n"); + } else { + printf("verify : NG\n"); + } + } + + fclose(decFp); + fclose(inpFp); +} + +int main(int argc, const char *argv[]) { + const char * const exeName = argv[0]; + char inpFilename[256] = { 0 }; + char ldmFilename[256] = { 0 }; + char decFilename[256] = { 0 }; + + if (argc < 2) { + printf("Wrong arguments\n"); + printf("Usage:\n"); + printf("%s FILE\n", exeName); + return 1; + } + + snprintf(inpFilename, 256, "%s", argv[1]); + snprintf(ldmFilename, 256, "%s.ldm", argv[1]); + snprintf(decFilename, 256, "%s.ldm.dec", argv[1]); + + printf("inp = [%s]\n", inpFilename); + printf("ldm = [%s]\n", ldmFilename); + printf("dec = [%s]\n", decFilename); + + + /* Compress */ + { + struct timeval tv1, tv2; + gettimeofday(&tv1, NULL); + if (compress(inpFilename, ldmFilename)) { + printf("Compress error"); + return 1; + } + gettimeofday(&tv2, NULL); + printf("Total time = %f seconds\n", + (double) (tv2.tv_usec - tv1.tv_usec) / 1000000 + + (double) (tv2.tv_sec - tv1.tv_sec)); + } + + /* Decompress */ + { + struct timeval tv1, tv2; + gettimeofday(&tv1, NULL); + if (decompress(ldmFilename, decFilename)) { + printf("Decompress error"); + return 1; + } + gettimeofday(&tv2, NULL); + printf("Total time = %f seconds\n", + (double) (tv2.tv_usec - tv1.tv_usec) / 1000000 + + (double) (tv2.tv_sec - tv1.tv_sec)); + } + /* verify */ + verify(inpFilename, decFilename); + return 0; +} + + +#if 0 +static size_t compress_file(FILE *in, FILE *out, size_t *size_in, + size_t *size_out) { + char *src, *buf = NULL; + size_t r = 1; + size_t size, n, k, count_in = 0, count_out = 0, offset, frame_size = 0; + + src = malloc(BUF_SIZE); + if (!src) { + printf("Not enough memory\n"); + goto cleanup; + } + + size = BUF_SIZE + LDM_HEADER_SIZE; + buf = malloc(size); + if (!buf) { + printf("Not enough memory\n"); + goto cleanup; + } + + + for (;;) { + k = fread(src, 1, BUF_SIZE, in); + if (k == 0) + break; + count_in += k; + + n = LDM_compress(src, buf, k, BUF_SIZE); + + // n = k; + // offset += n; + offset = k; + count_out += k; + +// k = fwrite(src, 1, offset, out); + + k = fwrite(buf, 1, offset, out); + if (k < offset) { + if (ferror(out)) + printf("Write failed\n"); + else + printf("Short write\n"); + goto cleanup; + } + + } + *size_in = count_in; + *size_out = count_out; + r = 0; + cleanup: + free(src); + free(buf); + return r; +} + +static size_t decompress_file(FILE *in, FILE *out) { + void *src = malloc(BUF_SIZE); + void *dst = NULL; + size_t dst_capacity = BUF_SIZE; + size_t ret = 1; + size_t bytes_written = 0; + + if (!src) { + perror("decompress_file(src)"); + goto cleanup; + } + + while (ret != 0) { + /* Load more input */ + size_t src_size = fread(src, 1, BUF_SIZE, in); + void *src_ptr = src; + void *src_end = src_ptr + src_size; + if (src_size == 0 || ferror(in)) { + printf("(TODO): Decompress: not enough input or error reading file\n"); + //TODO + ret = 0; + goto cleanup; + } + + /* Allocate destination buffer if it hasn't been allocated already */ + if (!dst) { + dst = malloc(dst_capacity); + if (!dst) { + perror("decompress_file(dst)"); + goto cleanup; + } + } + + // TODO + + /* Decompress: + * Continue while there is more input to read. + */ + while (src_ptr != src_end && ret != 0) { + // size_t dst_size = src_size; + size_t dst_size = LDM_decompress(src, dst, src_size, dst_capacity); + size_t written = fwrite(dst, 1, dst_size, out); +// printf("Writing %zu bytes\n", dst_size); + bytes_written += dst_size; + if (written != dst_size) { + printf("Decompress: Failed to write to file\n"); + goto cleanup; + } + src_ptr += src_size; + src_size = src_end - src_ptr; + } + + /* Update input */ + + } + + printf("Wrote %zu bytes\n", bytes_written); + + cleanup: + free(src); + free(dst); + + return ret; +} + +int main2(int argc, char *argv[]) { + char inpFilename[256] = { 0 }; + char ldmFilename[256] = { 0 }; + char decFilename[256] = { 0 }; + + if (argc < 2) { + printf("Please specify input filename\n"); + return 0; + } + snprintf(inpFilename, 256, "%s", argv[1]); + snprintf(ldmFilename, 256, "%s.ldm", argv[1]); + snprintf(decFilename, 256, "%s.ldm.dec", argv[1]); + + printf("inp = [%s]\n", inpFilename); + printf("ldm = [%s]\n", ldmFilename); + printf("dec = [%s]\n", decFilename); + + /* compress */ + { + FILE *inpFp = fopen(inpFilename, "rb"); + FILE *outFp = fopen(ldmFilename, "wb"); + size_t sizeIn = 0; + size_t sizeOut = 0; + size_t ret; + printf("compress : %s -> %s\n", inpFilename, ldmFilename); + ret = compress_file(inpFp, outFp, &sizeIn, &sizeOut); + if (ret) { + printf("compress : failed with code %zu\n", ret); + return ret; + } + printf("%s: %zu → %zu bytes, %.1f%%\n", + inpFilename, sizeIn, sizeOut, + (double)sizeOut / sizeIn * 100); + printf("compress : done\n"); + + fclose(outFp); + fclose(inpFp); + } + + /* decompress */ + { + FILE *inpFp = fopen(ldmFilename, "rb"); + FILE *outFp = fopen(decFilename, "wb"); + size_t ret; + + printf("decompress : %s -> %s\n", ldmFilename, decFilename); + ret = decompress_file(inpFp, outFp); + if (ret) { + printf("decompress : failed with code %zu\n", ret); + return ret; + } + printf("decompress : done\n"); + + fclose(outFp); + fclose(inpFp); + } + + /* verify */ + { + FILE *inpFp = fopen(inpFilename, "rb"); + FILE *decFp = fopen(decFilename, "rb"); + + printf("verify : %s <-> %s\n", inpFilename, decFilename); + const int cmp = compare(inpFp, decFp); + if(0 == cmp) { + printf("verify : OK\n"); + } else { + printf("verify : NG\n"); + } + + fclose(decFp); + fclose(inpFp); + } + return 0; +} +#endif + diff --git a/contrib/long_distance_matching/versions/v0.3/util.c b/contrib/long_distance_matching/versions/v0.3/util.c new file mode 100644 index 000000000..9ea4ca1e5 --- /dev/null +++ b/contrib/long_distance_matching/versions/v0.3/util.c @@ -0,0 +1,64 @@ +#include +#include +#include +#include + +#include "util.h" + +typedef uint8_t BYTE; +typedef uint16_t U16; +typedef uint32_t U32; +typedef int32_t S32; +typedef uint64_t U64; + +unsigned LDM_isLittleEndian(void) { + const union { U32 u; BYTE c[4]; } one = { 1 }; + return one.c[0]; +} + +U16 LDM_read16(const void *memPtr) { + U16 val; + memcpy(&val, memPtr, sizeof(val)); + return val; +} + +U16 LDM_readLE16(const void *memPtr) { + if (LDM_isLittleEndian()) { + return LDM_read16(memPtr); + } else { + const BYTE *p = (const BYTE *)memPtr; + return (U16)((U16)p[0] + (p[1] << 8)); + } +} + +void LDM_write16(void *memPtr, U16 value){ + memcpy(memPtr, &value, sizeof(value)); +} + +void LDM_write32(void *memPtr, U32 value) { + memcpy(memPtr, &value, sizeof(value)); +} + +void LDM_writeLE16(void *memPtr, U16 value) { + if (LDM_isLittleEndian()) { + LDM_write16(memPtr, value); + } else { + BYTE* p = (BYTE *)memPtr; + p[0] = (BYTE) value; + p[1] = (BYTE)(value>>8); + } +} + +U32 LDM_read32(const void *ptr) { + return *(const U32 *)ptr; +} + +U64 LDM_read64(const void *ptr) { + return *(const U64 *)ptr; +} + +void LDM_copy8(void *dst, const void *src) { + memcpy(dst, src, 8); +} + + diff --git a/contrib/long_distance_matching/versions/v0.3/util.h b/contrib/long_distance_matching/versions/v0.3/util.h new file mode 100644 index 000000000..90726412e --- /dev/null +++ b/contrib/long_distance_matching/versions/v0.3/util.h @@ -0,0 +1,23 @@ +#ifndef LDM_UTIL_H +#define LDM_UTIL_H + +unsigned LDM_isLittleEndian(void); + +uint16_t LDM_read16(const void *memPtr); + +uint16_t LDM_readLE16(const void *memPtr); + +void LDM_write16(void *memPtr, uint16_t value); + +void LDM_write32(void *memPtr, uint32_t value); + +void LDM_writeLE16(void *memPtr, uint16_t value); + +uint32_t LDM_read32(const void *ptr); + +uint64_t LDM_read64(const void *ptr); + +void LDM_copy8(void *dst, const void *src); + + +#endif /* LDM_UTIL_H */