X-Git-Url: https://git.xonotic.org/?p=xonotic%2Fgmqcc.git;a=blobdiff_plain;f=util.c;h=b7e3c7c7295625ac395fdccb49656edccb7b8491;hp=c3d3d4fa2d6cc9c220784c079122f146e508ff1f;hb=160e7cf7eebd7fa173fb739aca00143097a3518b;hpb=9f742271b1224c29b25f21fd9680b9031edf0b92 diff --git a/util.c b/util.c index c3d3d4f..b7e3c7c 100644 --- a/util.c +++ b/util.c @@ -1,5 +1,5 @@ /* - * Copyright (C) 2012 + * Copyright (C) 2012, 2013 * Dale Weiler * Wolfgang Bumiller * @@ -25,10 +25,13 @@ #include #include "gmqcc.h" +/* TODO: remove globals ... */ uint64_t mem_ab = 0; uint64_t mem_db = 0; uint64_t mem_at = 0; uint64_t mem_dt = 0; +uint64_t mem_pk = 0; +uint64_t mem_hw = 0; struct memblock_t { const char *file; @@ -38,10 +41,16 @@ struct memblock_t { struct memblock_t *prev; }; +#define PEAK_MEM \ + do { \ + if (mem_hw > mem_pk) \ + mem_pk = mem_hw; \ + } while (0) + static struct memblock_t *mem_start = NULL; void *util_memory_a(size_t byte, unsigned int line, const char *file) { - struct memblock_t *info = malloc(sizeof(struct memblock_t) + byte); + struct memblock_t *info = (struct memblock_t*)malloc(sizeof(struct memblock_t) + byte); void *data = (void*)(info+1); if (!info) return NULL; info->line = line; @@ -53,21 +62,23 @@ void *util_memory_a(size_t byte, unsigned int line, const char *file) { mem_start->prev = info; mem_start = info; - util_debug("MEM", "allocation: % 8u (bytes) address 0x%08X @ %s:%u\n", byte, data, file, line); mem_at++; mem_ab += info->byte; + mem_hw += info->byte; + + PEAK_MEM; return data; } -void util_memory_d(void *ptrn, unsigned int line, const char *file) { +void util_memory_d(void *ptrn) { struct memblock_t *info = NULL; if (!ptrn) return; info = ((struct memblock_t*)ptrn - 1); - util_debug("MEM", "released: % 8u (bytes) address 0x%08X @ %s:%u\n", info->byte, ptrn, file, line); mem_db += info->byte; + mem_hw -= info->byte; mem_dt++; if (info->prev) @@ -88,18 +99,16 @@ void *util_memory_r(void *ptrn, size_t byte, unsigned int line, const char *file if (!ptrn) return util_memory_a(byte, line, file); if (!byte) { - util_memory_d(ptrn, line, file); + util_memory_d(ptrn); return NULL; } oldinfo = ((struct memblock_t*)ptrn - 1); newinfo = ((struct memblock_t*)malloc(sizeof(struct memblock_t) + byte)); - util_debug("MEM", "reallocation: % 8u -> %u (bytes) address 0x%08X -> 0x%08X @ %s:%u\n", oldinfo->byte, byte, ptrn, (void*)(newinfo+1), file, line); - /* new data */ if (!newinfo) { - util_memory_d(oldinfo+1, line, file); + util_memory_d(oldinfo+1); return NULL; } @@ -125,131 +134,123 @@ void *util_memory_r(void *ptrn, size_t byte, unsigned int line, const char *file mem_start = newinfo; mem_ab -= oldinfo->byte; + mem_hw -= oldinfo->byte; mem_ab += newinfo->byte; + mem_hw += newinfo->byte; + + PEAK_MEM; free(oldinfo); return newinfo+1; } +static void util_dumpmem(struct memblock_t *memory, uint16_t cols) { + uint32_t i, j; + for (i = 0; i < memory->byte + ((memory->byte % cols) ? (cols - memory->byte % cols) : 0); i++) { + if (i % cols == 0) con_out(" 0x%06X: ", i); + if (i < memory->byte) con_out("%02X " , 0xFF & ((char*)(memory + 1))[i]); + else con_out(" "); + + if ((uint16_t)(i % cols) == (cols - 1)) { + for (j = i - (cols - 1); j <= i; j++) { + con_out("%c", + (j >= memory->byte) + ? ' ' + : (isprint(((char*)(memory + 1))[j])) + ? 0xFF & ((char*)(memory + 1)) [j] + : '.' + ); + } + con_out("\n"); + } + } +} + void util_meminfo() { struct memblock_t *info; - if (!opts.memchk) - return; - for (info = mem_start; info; info = info->next) { - util_debug("MEM", "lost: % 8u (bytes) at %s:%u\n", - info->byte, - info->file, - info->line); + if (OPTS_OPTION_BOOL(OPTION_DEBUG)) { + for (info = mem_start; info; info = info->next) { + con_out("lost: %u (bytes) at %s:%u\n", + info->byte, + info->file, + info->line); + + util_dumpmem(info, OPTS_OPTION_U16(OPTION_MEMDUMPCOLS)); + } } - util_debug("MEM", "Memory information:\n\ - Total allocations: %llu\n\ - Total deallocations: %llu\n\ - Total allocated: %llu (bytes)\n\ - Total deallocated: %llu (bytes)\n\ - Leaks found: lost %llu (bytes) in %d allocations\n", - mem_at, mem_dt, - mem_ab, mem_db, - (mem_ab - mem_db), - (mem_at - mem_dt) - ); + if (OPTS_OPTION_BOOL(OPTION_DEBUG) || + OPTS_OPTION_BOOL(OPTION_MEMCHK)) { + con_out("Memory information:\n\ + Total allocations: %llu\n\ + Total deallocations: %llu\n\ + Total allocated: %f (MB)\n\ + Total deallocated: %f (MB)\n\ + Total peak memory: %f (MB)\n\ + Total leaked memory: %f (MB) in %llu allocations\n", + mem_at, + mem_dt, + (float)(mem_ab) / 1048576.0f, + (float)(mem_db) / 1048576.0f, + (float)(mem_pk) / 1048576.0f, + (float)(mem_ab - mem_db) / 1048576.0f, + + /* could be more clever */ + (mem_at - mem_dt) + ); + } } /* * Some string utility functions, because strdup uses malloc, and we want * to track all memory (without replacing malloc). */ -char *util_strdup(const char *s) { +char *_util_Estrdup(const char *s, const char *file, size_t line) { size_t len = 0; char *ptr = NULL; + /* in case of -DNOTRACK */ + (void)file; + (void)line; + if (!s) return NULL; - if ((len = strlen(s)) && (ptr = mem_a(len+1))) { + if ((len = strlen(s)) && (ptr = (char*)mem_af(len+1, line, file))) { memcpy(ptr, s, len); ptr[len] = '\0'; } return ptr; } -/* - * Remove quotes from a string, escapes from \ in string - * as well. This function shouldn't be used to create a - * char array that is later freed (it uses pointer arith) - */ -char *util_strrq(const char *s) { - char *dst = (char*)s; - char *src = (char*)s; - char chr; - while ((chr = *src++) != '\0') { - if (chr == '\\') { - *dst++ = chr; - if ((chr = *src++) == '\0') - break; - *dst++ = chr; - } else if (chr != '"') - *dst++ = chr; - } - *dst = '\0'; - return dst; -} - -/* - * Chops a substring from an existing string by creating a - * copy of it and null terminating it at the required position. - */ -char *util_strchp(const char *s, const char *e) { - const char *c = NULL; - if (!s || !e) - return NULL; - - c = s; - while (c != e) - c++; +char *_util_Estrdup_empty(const char *s, const char *file, size_t line) { + size_t len = 0; + char *ptr = NULL; - return util_strdup(s); -} + /* in case of -DNOTRACK */ + (void)file; + (void)line; -/* - * Returns true if string is all uppercase, otherwise - * it returns false. - */ -bool util_strupper(const char *str) { - while (*str) { - if(!isupper(*str)) - return false; - str++; - } - return true; -} + if (!s) + return NULL; -/* - * Returns true if string is all digits, otherwise - * it returns false. - */ -bool util_strdigit(const char *str) { - while (*str) { - if(!isdigit(*str)) - return false; - str++; + len = strlen(s); + if ((ptr = (char*)mem_af(len+1, line, file))) { + memcpy(ptr, s, len); + ptr[len] = '\0'; } - return true; -} - -bool util_strncmpexact(const char *src, const char *ned, size_t len) { - return (!strncmp(src, ned, len) && !src[len]); + return ptr; } void util_debug(const char *area, const char *ms, ...) { va_list va; - if (!opts.debug) + if (!OPTS_OPTION_BOOL(OPTION_DEBUG)) return; - if (!strcmp(area, "MEM") && !opts.memchk) + if (!strcmp(area, "MEM") && !OPTS_OPTION_BOOL(OPTION_MEMCHK)) return; va_start(va, ms); @@ -259,26 +260,75 @@ void util_debug(const char *area, const char *ms, ...) { } /* - * Endianess swapping, all data must be stored little-endian. This - * reorders by stride and length, much nicer than other functions for - * certian-sized types like short or int. + * only required if big endian .. otherwise no need to swap + * data. */ -void util_endianswap(void *m, int s, int l) { - size_t w = 0; - size_t i = 0; +#if PLATFORM_BYTE_ORDER == GMQCC_BYTE_ORDER_BIG + static GMQCC_INLINE void util_swap16(uint16_t *d, size_t l) { + while (l--) { + d[l] = (d[l] << 8) | (d[l] >> 8); + } + } - /* ignore if we're already LE */ - if(*((char *)&s)) - return; + static GMQCC_INLINE void util_swap32(uint32_t *d, size_t l) { + while (l--) { + uint32_t v; + v = ((d[l] << 8) & 0xFF00FF00) | ((d[l] >> 8) & 0x00FF00FF); + d[l] = (v << 16) | (v >> 16); + } + } - for(; w < (size_t)l; w++) { - for(; i < (size_t)(s << 1); i++) { - unsigned char *p = (unsigned char *)m+w*s; - unsigned char t = p[i]; - p[i] = p[s-i-1]; - p[s-i-1] = t; + /* Some strange system doesn't like constants that big, AND doesn't recognize an ULL suffix + * so let's go the safe way + */ + static GMQCC_INLINE void util_swap64(uint32_t *d, size_t l) { + /* + while (l--) { + uint64_t v; + v = ((d[l] << 8) & 0xFF00FF00FF00FF00) | ((d[l] >> 8) & 0x00FF00FF00FF00FF); + v = ((v << 16) & 0xFFFF0000FFFF0000) | ((v >> 16) & 0x0000FFFF0000FFFF); + d[l] = (v << 32) | (v >> 32); + } + */ + size_t i; + for (i = 0; i < l; i += 2) { + uint32_t v1 = d[i]; + d[i] = d[i+1]; + d[i+1] = v1; + util_swap32(d+i, 2); } } +#endif + +void util_endianswap(void *_data, size_t length, unsigned int typesize) { +# if PLATFORM_BYTE_ORDER == -1 /* runtime check */ + if (*((char*)&typesize)) + return; +#else + /* prevent unused warnings */ + (void) _data; + (void) length; + (void) typesize; + +# if PLATFORM_BYTE_ORDER == GMQCC_BYTE_ORDER_LITTLE + return; +# else + switch (typesize) { + case 1: return; + case 2: + util_swap16((uint16_t*)_data, length>>1); + return; + case 4: + util_swap32((uint32_t*)_data, length>>2); + return; + case 8: + util_swap64((uint32_t*)_data, length>>3); + return; + + default: exit(EXIT_FAILURE); /* please blow the fuck up! */ + } +# endif +#endif } /* @@ -295,7 +345,7 @@ void util_endianswap(void *m, int s, int l) { * In most traditional CRC algorithms on uses a reflected table driven method where a value * or register is reflected if it's bits are swapped around it's center. For example: * take the bits 0101 is the 4-bit reflection of 1010, and respectfully 0011 would be the - * reflection of 1100. Quakle however expects a NON-Reflected CRC on the output, but still + * reflection of 1100. Quake however expects a NON-Reflected CRC on the output, but still * requires a final XOR on the values (0xFFFF and 0x0000) this is a standard CCITT CRC-16 * which I respectfully as a programmer don't agree with. * @@ -303,58 +353,6 @@ void util_endianswap(void *m, int s, int l) { * but those are what Quake seems to request. */ -/* - * This is an implementation of CRC32 & CRC16. The polynomials have been - * offline computed for faster generation at the cost of larger code size. - * - * CRC32 Polynomial: 0xEDB88320 - * CRC16 Polynomial: 0x00001021 - */ -static const uint32_t util_crc32_table[] = { - 0x00000000, 0x77073096, 0xEE0E612C, 0x990951BA, 0x076DC419, 0x706AF48F, - 0xE963A535, 0x9E6495A3, 0x0EDB8832, 0x79DCB8A4, 0xE0D5E91E, 0x97D2D988, - 0x09B64C2B, 0x7EB17CBD, 0xE7B82D07, 0x90BF1D91, 0x1DB71064, 0x6AB020F2, - 0xF3B97148, 0x84BE41DE, 0x1ADAD47D, 0x6DDDE4EB, 0xF4D4B551, 0x83D385C7, - 0x136C9856, 0x646BA8C0, 0xFD62F97A, 0x8A65C9EC, 0x14015C4F, 0x63066CD9, - 0xFA0F3D63, 0x8D080DF5, 0x3B6E20C8, 0x4C69105E, 0xD56041E4, 0xA2677172, - 0x3C03E4D1, 0x4B04D447, 0xD20D85FD, 0xA50AB56B, 0x35B5A8FA, 0x42B2986C, - 0xDBBBC9D6, 0xACBCF940, 0x32D86CE3, 0x45DF5C75, 0xDCD60DCF, 0xABD13D59, - 0x26D930AC, 0x51DE003A, 0xC8D75180, 0xBFD06116, 0x21B4F4B5, 0x56B3C423, - 0xCFBA9599, 0xB8BDA50F, 0x2802B89E, 0x5F058808, 0xC60CD9B2, 0xB10BE924, - 0x2F6F7C87, 0x58684C11, 0xC1611DAB, 0xB6662D3D, 0x76DC4190, 0x01DB7106, - 0x98D220BC, 0xEFD5102A, 0x71B18589, 0x06B6B51F, 0x9FBFE4A5, 0xE8B8D433, - 0x7807C9A2, 0x0F00F934, 0x9609A88E, 0xE10E9818, 0x7F6A0DBB, 0x086D3D2D, - 0x91646C97, 0xE6635C01, 0x6B6B51F4, 0x1C6C6162, 0x856530D8, 0xF262004E, - 0x6C0695ED, 0x1B01A57B, 0x8208F4C1, 0xF50FC457, 0x65B0D9C6, 0x12B7E950, - 0x8BBEB8EA, 0xFCB9887C, 0x62DD1DDF, 0x15DA2D49, 0x8CD37CF3, 0xFBD44C65, - 0x4DB26158, 0x3AB551CE, 0xA3BC0074, 0xD4BB30E2, 0x4ADFA541, 0x3DD895D7, - 0xA4D1C46D, 0xD3D6F4FB, 0x4369E96A, 0x346ED9FC, 0xAD678846, 0xDA60B8D0, - 0x44042D73, 0x33031DE5, 0xAA0A4C5F, 0xDD0D7CC9, 0x5005713C, 0x270241AA, - 0xBE0B1010, 0xC90C2086, 0x5768B525, 0x206F85B3, 0xB966D409, 0xCE61E49F, - 0x5EDEF90E, 0x29D9C998, 0xB0D09822, 0xC7D7A8B4, 0x59B33D17, 0x2EB40D81, - 0xB7BD5C3B, 0xC0BA6CAD, 0xEDB88320, 0x9ABFB3B6, 0x03B6E20C, 0x74B1D29A, - 0xEAD54739, 0x9DD277AF, 0x04DB2615, 0x73DC1683, 0xE3630B12, 0x94643B84, - 0x0D6D6A3E, 0x7A6A5AA8, 0xE40ECF0B, 0x9309FF9D, 0x0A00AE27, 0x7D079EB1, - 0xF00F9344, 0x8708A3D2, 0x1E01F268, 0x6906C2FE, 0xF762575D, 0x806567CB, - 0x196C3671, 0x6E6B06E7, 0xFED41B76, 0x89D32BE0, 0x10DA7A5A, 0x67DD4ACC, - 0xF9B9DF6F, 0x8EBEEFF9, 0x17B7BE43, 0x60B08ED5, 0xD6D6A3E8, 0xA1D1937E, - 0x38D8C2C4, 0x4FDFF252, 0xD1BB67F1, 0xA6BC5767, 0x3FB506DD, 0x48B2364B, - 0xD80D2BDA, 0xAF0A1B4C, 0x36034AF6, 0x41047A60, 0xDF60EFC3, 0xA867DF55, - 0x316E8EEF, 0x4669BE79, 0xCB61B38C, 0xBC66831A, 0x256FD2A0, 0x5268E236, - 0xCC0C7795, 0xBB0B4703, 0x220216B9, 0x5505262F, 0xC5BA3BBE, 0xB2BD0B28, - 0x2BB45A92, 0x5CB36A04, 0xC2D7FFA7, 0xB5D0CF31, 0x2CD99E8B, 0x5BDEAE1D, - 0x9B64C2B0, 0xEC63F226, 0x756AA39C, 0x026D930A, 0x9C0906A9, 0xEB0E363F, - 0x72076785, 0x05005713, 0x95BF4A82, 0xE2B87A14, 0x7BB12BAE, 0x0CB61B38, - 0x92D28E9B, 0xE5D5BE0D, 0x7CDCEFB7, 0x0BDBDF21, 0x86D3D2D4, 0xF1D4E242, - 0x68DDB3F8, 0x1FDA836E, 0x81BE16CD, 0xF6B9265B, 0x6FB077E1, 0x18B74777, - 0x88085AE6, 0xFF0F6A70, 0x66063BCA, 0x11010B5C, 0x8F659EFF, 0xF862AE69, - 0x616BFFD3, 0x166CCF45, 0xA00AE278, 0xD70DD2EE, 0x4E048354, 0x3903B3C2, - 0xA7672661, 0xD06016F7, 0x4969474D, 0x3E6E77DB, 0xAED16A4A, 0xD9D65ADC, - 0x40DF0B66, 0x37D83BF0, 0xA9BCAE53, 0xDEBB9EC5, 0x47B2CF7F, 0x30B5FFE9, - 0xBDBDF21C, 0xCABAC28A, 0x53B39330, 0x24B4A3A6, 0xBAD03605, 0xCDD70693, - 0x54DE5729, 0x23D967BF, 0xB3667A2E, 0xC4614AB8, 0x5D681B02, 0x2A6F2B94, - 0xB40BBE37, 0xC30C8EA1, 0x5A05DF1B, 0x2D02EF8D -}; static const uint16_t util_crc16_table[] = { 0x0000, 0x1021, 0x2042, 0x3063, 0x4084, 0x50A5, 0x60C6, 0x70E7, 0x8108, 0x9129, 0xA14A, 0xB16B, @@ -401,80 +399,22 @@ static const uint16_t util_crc16_table[] = { 0x2E93, 0x3EB2, 0x0ED1, 0x1EF0 }; -/* - * Implements a CRC function for X worth bits using (uint[X]_t) - * as type. and util_crc[X]_table. - - * Quake expects a non-reflective CRC. - */ -#define CRC(X) \ -uint##X##_t util_crc##X(uint##X##_t current, const char *k, size_t len) { \ - register uint##X##_t h= current; \ - for (; len; --len, ++k) \ - h = util_crc##X##_table[(h>>8)^((unsigned char)*k)]^(h<<8); \ - return h; \ +/* Non - Reflected */ +uint16_t util_crc16(uint16_t current, const char *k, size_t len) { + register uint16_t h = current; + for (; len; --len, ++k) + h = util_crc16_table[(h>>8)^((unsigned char)*k)]^(h<<8); + return h; } -CRC(32) -CRC(16) -#undef CRC -/* -#define CRC(X) \ -uint##X##_t util_crc##X(const char *k, int len, const short clamp) { \ - register uint##X##_t h= (uint##X##_t)0xFFFFFFFF; \ - for (; len; --len, ++k) \ - h = util_crc##X##_table[(h^((unsigned char)*k))&0xFF]^(h>>8); \ - return (~h)%clamp; \ -} -*/ - - -/* - * Implements libc getline for systems that don't have it, which is - * assmed all. This works the same as getline(). - */ -int util_getline(char **lineptr, size_t *n, FILE *stream) { - int chr; - int ret; - char *pos; - - if (!lineptr || !n || !stream) - return -1; - if (!*lineptr) { - if (!(*lineptr = (char*)mem_a((*n=64)))) - return -1; - } - - chr = *n; - pos = *lineptr; - - for (;;) { - int c = getc(stream); - - if (chr < 2) { - *n += (*n > 16) ? *n : 64; - chr = *n + *lineptr - pos; - if (!(*lineptr = (char*)mem_r(*lineptr,*n))) - return -1; - pos = *n - chr + *lineptr; - } - - if (ferror(stream)) - return -1; - if (c == EOF) { - if (pos == *lineptr) - return -1; - else - break; - } - - *pos++ = c; - chr--; - if (c == '\n') - break; - } - *pos = '\0'; - return (ret = pos - *lineptr); +/* Reflective Varation (for reference) */ +#if 0 +uint16_t util_crc16(const char *k, int len, const short clamp) { + register uint16_t h= (uint16_t)0xFFFFFFFF; + for (; len; --len, ++k) + h = util_crc16_table[(h^((unsigned char)*k))&0xFF]^(h>>8); + return (~h)%clamp; } +#endif size_t util_strtocmd(const char *in, char *out, size_t outsz) { size_t sz = 1; @@ -492,26 +432,17 @@ size_t util_strtononcmd(const char *in, char *out, size_t outsz) { return sz-1; } - -FILE *util_fopen(const char *filename, const char *mode) -{ -#ifdef _MSC_VER - FILE *out; - if (fopen_s(&out, filename, mode) != 0) - return NULL; - return out; -#else - return fopen(filename, mode); -#endif -} - +/* TODO: rewrite ... when I redo the ve cleanup */ void _util_vec_grow(void **a, size_t i, size_t s) { - size_t m = *a ? 2*_vec_beg(*a)+i : i+1; - void *p = mem_r((*a ? _vec_raw(*a) : NULL), s * m + sizeof(size_t)*2); + vector_t *d = vec_meta(*a); + size_t m = *a ? 2 * d->allocated +i : i+1; + void *p = mem_r((*a ? d : NULL), s * m + sizeof(vector_t)); + if (!*a) - ((size_t*)p)[1] = 0; - *a = (void*)((size_t*)p + 2); - _vec_beg(*a) = m; + ((vector_t*)p)->used = 0; + *a = (vector_t*)p + 1; + + vec_meta(*a)->allocated = m; } /* @@ -525,61 +456,11 @@ typedef struct hash_node_t { struct hash_node_t *next; /* next node (linked list) */ } hash_node_t; -/* - * x86 and x86_64 optimized murmur hash functions for the hashtable - * we have individual implementations for optimal performance. - * - * Forced inlined as we wrap these up in the actual utility function - * below. These should be autovectorized by gcc. - */ -#ifdef __x86_64__ -GMQCC_INLINE uint32_t util_hthashfunc(hash_table_t *ht, const char *key, register size_t seed) { - const uint64_t mix = 0xC6A4A7935BD1E995UL; - const int rot = 47; - size_t size = strlen(key); - uint64_t hash = seed ^ (size - mix); - uint64_t alias = 0; - const uint64_t *beg = (const uint64_t*)key; - const uint64_t *end = beg + (size / 8); - const unsigned char *final = NULL; - - while (beg != end) { - alias = *beg++; - - alias *= mix; - alias ^= alias >> rot; - alias *= mix; - - hash ^= alias; - hash *= mix; - } - - final = (const unsigned char *)beg; - - switch (size & 7) { - case 7: hash ^= (uint64_t)(final[6]) << 48; - case 6: hash ^= (uint64_t)(final[5]) << 40; - case 5: hash ^= (uint64_t)(final[4]) << 32; - case 4: hash ^= (uint64_t)(final[3]) << 24; - case 3: hash ^= (uint64_t)(final[2]) << 16; - case 2: hash ^= (uint64_t)(final[1]) << 8; - case 1: hash ^= (uint64_t)(final[0]); - hash *= mix; - } - - hash ^= hash >> rot; - hash *= mix; - hash ^= hash >> rot; - - return (uint32_t)(hash % ht->size); -} - -#else -GMQCC_INLINE uint32_t util_hthashfunc(hash_table_t *ht, const char *key, register size_t seed) { +GMQCC_INLINE size_t util_hthash(hash_table_t *ht, const char *key) { const uint32_t mix = 0x5BD1E995; const uint32_t rot = 24; size_t size = strlen(key); - uint32_t hash = seed ^ size; + uint32_t hash = 0x1EF0 /* LICRC TAB */ ^ size; uint32_t alias = 0; const unsigned char *data = (const unsigned char*)key; @@ -608,28 +489,15 @@ GMQCC_INLINE uint32_t util_hthashfunc(hash_table_t *ht, const char *key, registe hash *= mix; hash ^= hash >> 15; - return hash % ht->size; -} -#endif - -/* we use the crc table as seeds for the murmur hash :P */ -size_t util_hthash(hash_table_t *ht, const char *key) { - static size_t seed = 0; - register size_t hash = util_hthashfunc(ht, key, util_crc32_table[seed]); - - /* reset seed */ - if (seed >= sizeof(util_crc32_table) / sizeof(*util_crc32_table)) - seed = 0; - - return hash; + return (size_t) (hash % ht->size); } hash_node_t *_util_htnewpair(const char *key, void *value) { hash_node_t *node; - if (!(node = mem_a(sizeof(hash_node_t)))) + if (!(node = (hash_node_t*)mem_a(sizeof(hash_node_t)))) return NULL; - if (!(node->key = util_strdup(key))) { + if (!(node->key = util_strdupe(key))) { mem_d(node); return NULL; } @@ -652,10 +520,10 @@ hash_table_t *util_htnew(size_t size) { if (size < 1) return NULL; - if (!(hashtable = mem_a(sizeof(hash_table_t)))) + if (!(hashtable = (hash_table_t*)mem_a(sizeof(hash_table_t)))) return NULL; - if (!(hashtable->table = mem_a(sizeof(hash_node_t*) * size))) { + if (!(hashtable->table = (hash_node_t**)mem_a(sizeof(hash_node_t*) * size))) { mem_d(hashtable); return NULL; } @@ -714,11 +582,45 @@ void *util_htget(hash_table_t *ht, const char *key) { return util_htgeth(ht, key, util_hthash(ht, key)); } +void *code_util_str_htgeth(hash_table_t *ht, const char *key, size_t bin) { + hash_node_t *pair; + size_t len, keylen; + int cmp; + + keylen = strlen(key); + + pair = ht->table[bin]; + while (pair && pair->key) { + len = strlen(pair->key); + if (len < keylen) { + pair = pair->next; + continue; + } + if (keylen == len) { + cmp = strcmp(key, pair->key); + if (cmp == 0) + return pair->value; + if (cmp < 0) + return NULL; + pair = pair->next; + continue; + } + cmp = strcmp(key, pair->key + len - keylen); + if (cmp == 0) { + uintptr_t up = (uintptr_t)pair->value; + up += len - keylen; + return (void*)up; + } + pair = pair->next; + } + return NULL; +} + /* * Free all allocated data in a hashtable, this is quite the amount * of work. */ -void util_htdel(hash_table_t *ht) { +void util_htrem(hash_table_t *ht, void (*callback)(void *data)) { size_t i = 0; for (; i < ht->size; i++) { hash_node_t *n = ht->table[i]; @@ -728,6 +630,8 @@ void util_htdel(hash_table_t *ht) { while (n) { if (n->key) mem_d(n->key); + if (callback) + callback(n->value); p = n; n = n->next; mem_d(p); @@ -738,3 +642,273 @@ void util_htdel(hash_table_t *ht) { mem_d(ht->table); mem_d(ht); } + +void util_htrmh(hash_table_t *ht, const char *key, size_t bin, void (*cb)(void*)) { + hash_node_t **pair = &ht->table[bin]; + hash_node_t *tmp; + + while (*pair && (*pair)->key && strcmp(key, (*pair)->key) > 0) + pair = &(*pair)->next; + + tmp = *pair; + if (!tmp || !tmp->key || strcmp(key, tmp->key) != 0) + return; + + if (cb) + (*cb)(tmp->value); + + *pair = tmp->next; + mem_d(tmp->key); + mem_d(tmp); +} + +void util_htrm(hash_table_t *ht, const char *key, void (*cb)(void*)) { + util_htrmh(ht, key, util_hthash(ht, key), cb); +} + +void util_htdel(hash_table_t *ht) { + util_htrem(ht, NULL); +} + +/* + * Portable implementation of vasprintf/asprintf. Assumes vsnprintf + * exists, otherwise compiler error. + * + * TODO: fix for MSVC .... + */ +int util_vasprintf(char **dat, const char *fmt, va_list args) { + int ret; + int len; + char *tmp = NULL; + + /* + * For visuals tido _vsnprintf doesn't tell you the length of a + * formatted string if it overflows. However there is a MSVC + * intrinsic (which is documented wrong) called _vcsprintf which + * will return the required amount to allocate. + */ + #ifdef _MSC_VER + char *str; + if ((len = _vscprintf(fmt, args)) < 0) { + *dat = NULL; + return -1; + } + + tmp = mem_a(len + 1); + if ((ret = _vsnprintf(tmp, len+1, fmt, args)) != len) { + mem_d(tmp); + *dat = NULL; + return -1; + } + *dat = tmp; + return len; + #else + /* + * For everything else we have a decent conformint vsnprintf that + * returns the number of bytes needed. We give it a try though on + * a short buffer, since efficently speaking, it could be nice to + * above a second vsnprintf call. + */ + char buf[128]; + va_list cpy; + va_copy(cpy, args); + len = vsnprintf(buf, sizeof(buf), fmt, cpy); + va_end (cpy); + + if (len < (int)sizeof(buf)) { + *dat = util_strdup(buf); + return len; + } + + /* not large enough ... */ + tmp = (char*)mem_a(len + 1); + if ((ret = vsnprintf(tmp, len + 1, fmt, args)) != len) { + mem_d(tmp); + *dat = NULL; + return -1; + } + + *dat = tmp; + return len; + #endif +} +int util_asprintf(char **ret, const char *fmt, ...) { + va_list args; + int read; + va_start(args, fmt); + read = util_vasprintf(ret, fmt, args); + va_end (args); + + return read; +} + +/* + * Implementation of the Mersenne twister PRNG (pseudo random numer + * generator). Implementation of MT19937. Has a period of 2^19937-1 + * which is a Mersenne Prime (hence the name). + * + * Implemented from specification and original paper: + * http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/ARTICLES/mt.pdf + * + * This code is placed in the public domain by me personally + * (Dale Weiler, a.k.a graphitemaster). + */ + +#define MT_SIZE 624 +#define MT_PERIOD 397 +#define MT_SPACE (MT_SIZE - MT_PERIOD) + +static uint32_t mt_state[MT_SIZE]; +static size_t mt_index = 0; + +static GMQCC_INLINE void mt_generate() { + /* + * The loop has been unrolled here: the original paper and implemenation + * Called for the following code: + * for (register unsigned i = 0; i < MT_SIZE; ++i) { + * register uint32_t load; + * load = (0x80000000 & mt_state[i]) // most significant 32nd bit + * load |= (0x7FFFFFFF & mt_state[(i + 1) % MT_SIZE]) // least significant 31nd bit + * + * mt_state[i] = mt_state[(i + MT_PERIOD) % MT_SIZE] ^ (load >> 1); + * + * if (load & 1) mt_state[i] ^= 0x9908B0DF; + * } + * + * This essentially is a waste: we have two modulus operations, and + * a branch that is executed every iteration from [0, MT_SIZE). + * + * Please see: http://www.quadibloc.com/crypto/co4814.htm for more + * information on how this clever trick works. + */ + static const uint32_t matrix[2] = { + 0x00000000, + 0x9908B0Df + }; + /* + * This register gives up a little more speed by instructing the compiler + * to force these into CPU registers (they're counters for indexing mt_state + * which we can force the compiler to generate prefetch instructions for) + */ + register uint32_t y; + register uint32_t i; + + /* + * Said loop has been unrolled for MT_SPACE (226 iterations), opposed + * to [0, MT_SIZE) (634 iterations). + */ + for (i = 0; i < MT_SPACE; ++i) { + y = (0x80000000 & mt_state[i]) | (0x7FFFFFF & mt_state[i + 1]); + mt_state[i] = mt_state[i + MT_PERIOD] ^ (y >> 1) ^ matrix[y & 1]; + + i ++; /* loop unroll */ + + y = (0x80000000 & mt_state[i]) | (0x7FFFFFF & mt_state[i + 1]); + mt_state[i] = mt_state[i + MT_PERIOD] ^ (y >> 1) ^ matrix[y & 1]; + } + + /* + * collapsing the walls unrolled (evenly dividing 396 [632-227 = 396 + * = 2*2*3*3*11]) + */ + i = MT_SPACE; + while (i < MT_SIZE - 1) { + /* + * We expand this 11 times .. manually, no macros are required + * here. This all fits in the CPU cache. + */ + y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]); + mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1]; + ++i; + y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]); + mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1]; + ++i; + y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]); + mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1]; + ++i; + y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]); + mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1]; + ++i; + y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]); + mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1]; + ++i; + y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]); + mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1]; + ++i; + y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]); + mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1]; + ++i; + y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]); + mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1]; + ++i; + y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]); + mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1]; + ++i; + y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]); + mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1]; + ++i; + y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]); + mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1]; + ++i; + } + + /* i = mt_state[623] */ + y = (0x80000000 & mt_state[MT_SIZE - 1]) | (0x7FFFFFFF & mt_state[MT_SIZE - 1]); + mt_state[MT_SIZE - 1] = mt_state[MT_PERIOD - 1] ^ (y >> 1) ^ matrix[y & 1]; +} + +void util_seed(uint32_t value) { + /* + * We seed the mt_state with a LCG (linear congruential generator) + * We're operating exactly on exactly m=32, so there is no need to + * use modulus. + * + * The multipler of choice is 0x6C07865, also knows as the Borosh- + * Niederreiter multipler used for modulus 2^32. More can be read + * about this in Knuth's TAOCP Volume 2, page 106. + * + * If you don't own TAOCP something is wrong with you :-) .. so I + * also provided a link to the original paper by Borosh and + * Niederreiter. It's called "Optional Multipliers for PRNG by The + * Linear Congruential Method" (1983). + * http://en.wikipedia.org/wiki/Linear_congruential_generator + * + * From said page, it says the following: + * "A common Mersenne twister implementation, interestingly enough + * used an LCG to generate seed data." + * + * Remarks: + * The data we're operating on is 32-bits for the mt_state array, so + * there is no masking required with 0xFFFFFFFF + */ + register size_t i; + + mt_state[0] = value; + for (i = 1; i < MT_SIZE; ++i) + mt_state[i] = 0x6C078965 * (mt_state[i - 1] ^ mt_state[i - 1] >> 30) + i; +} + +uint32_t util_rand() { + register uint32_t y; + + /* + * This is inlined with any sane compiler (I checked) + * for some reason though, SubC seems to be generating invalid + * code when it inlines this. + */ + if (!mt_index) + mt_generate(); + + y = mt_state[mt_index]; + + /* Standard tempering */ + y ^= y >> 11; /* +7 */ + y ^= y << 7 & 0x9D2C5680; /* +4 */ + y ^= y << 15 & 0xEFC60000; /* -4 */ + y ^= y >> 18; /* -7 */ + + if(++mt_index == MT_SIZE) + mt_index = 0; + + return y; +}