From: Stella Lau Date: Tue, 11 Jul 2017 16:23:44 +0000 (-0700) Subject: Save v3 X-Git-Tag: v1.3.1^2~12^2~7^2~44 X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=f6c5d07fe295997d9f01461d2c40714e10aa5827;p=thirdparty%2Fzstd.git Save v3 --- diff --git a/contrib/long_distance_matching/ldm.c b/contrib/long_distance_matching/ldm.c index 42a4affd7..1dedf5c37 100644 --- a/contrib/long_distance_matching/ldm.c +++ b/contrib/long_distance_matching/ldm.c @@ -7,9 +7,9 @@ #include "ldm.h" #include "util.h" -#define HASH_EVERY 7 +#define HASH_EVERY 1 -#define LDM_MEMORY_USAGE 20 +#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) @@ -17,7 +17,7 @@ #define LDM_OFFSET_SIZE 4 -#define WINDOW_SIZE (1 << 24) +#define WINDOW_SIZE (1 << 20) #define MAX_WINDOW_SIZE 31 #define HASH_SIZE 4 #define LDM_HASH_LENGTH 4 @@ -28,7 +28,7 @@ #define RUN_BITS (8-ML_BITS) #define RUN_MASK ((1U< +#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/v3/ldm.h b/contrib/long_distance_matching/versions/v3/ldm.h new file mode 100644 index 000000000..287d444dd --- /dev/null +++ b/contrib/long_distance_matching/versions/v3/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/v3/main-ldm.c b/contrib/long_distance_matching/versions/v3/main-ldm.c new file mode 100644 index 000000000..724d735dd --- /dev/null +++ b/contrib/long_distance_matching/versions/v3/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/v3/util.c b/contrib/long_distance_matching/versions/v3/util.c new file mode 100644 index 000000000..9ea4ca1e5 --- /dev/null +++ b/contrib/long_distance_matching/versions/v3/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/v3/util.h b/contrib/long_distance_matching/versions/v3/util.h new file mode 100644 index 000000000..90726412e --- /dev/null +++ b/contrib/long_distance_matching/versions/v3/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 */