X-Git-Url: https://git.xonotic.org/?p=xonotic%2Fgmqcc.git;a=blobdiff_plain;f=util.c;h=ecae50516514e45db673cd10966ac408a010b0d1;hp=5650fcda892fd1ec3475d8c94c9f22cc21154674;hb=a68f0fcb355db42acabe72da5939fbd1b04f6016;hpb=7833efd6e12c8f9c4cc65f05b05a2a10537754fd diff --git a/util.c b/util.c index 5650fcd..ecae505 100644 --- a/util.c +++ b/util.c @@ -1,5 +1,5 @@ /* - * Copyright (C) 2012 + * Copyright (C) 2012, 2013 * Dale Weiler * Wolfgang Bumiller * @@ -21,235 +21,45 @@ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -#include -#include -#include "gmqcc.h" - -uint64_t mem_ab = 0; -uint64_t mem_db = 0; -uint64_t mem_at = 0; -uint64_t mem_dt = 0; - -struct memblock_t { - const char *file; - unsigned int line; - size_t byte; - struct memblock_t *next; - struct memblock_t *prev; -}; - -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); - void *data = (void*)(info+1); - if (!info) return NULL; - info->line = line; - info->byte = byte; - info->file = file; - info->prev = NULL; - info->next = mem_start; - if (mem_start) - 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; - - return data; -} - -void util_memory_d(void *ptrn, unsigned int line, const char *file) { - 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_dt++; - - if (info->prev) - info->prev->next = info->next; - if (info->next) - info->next->prev = info->prev; - if (info == mem_start) - mem_start = info->next; - - free(info); -} - -void *util_memory_r(void *ptrn, size_t byte, unsigned int line, const char *file) { - struct memblock_t *oldinfo = NULL; - - struct memblock_t *newinfo; - - if (!ptrn) - return util_memory_a(byte, line, file); - if (!byte) { - util_memory_d(ptrn, line, file); - return NULL; - } +#include +#include - 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); - return NULL; - } - - /* copy old */ - memcpy(newinfo+1, oldinfo+1, oldinfo->byte); - - /* free old */ - if (oldinfo->prev) - oldinfo->prev->next = oldinfo->next; - if (oldinfo->next) - oldinfo->next->prev = oldinfo->prev; - if (oldinfo == mem_start) - mem_start = oldinfo->next; - - /* fill info */ - newinfo->line = line; - newinfo->byte = byte; - newinfo->file = file; - newinfo->prev = NULL; - newinfo->next = mem_start; - if (mem_start) - mem_start->prev = newinfo; - mem_start = newinfo; - - mem_ab -= oldinfo->byte; - mem_ab += newinfo->byte; - - free(oldinfo); - - return newinfo+1; -} - -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); - } - - 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) - ); -} - -/* - * Some string utility functions, because strdup uses malloc, and we want - * to track all memory (without replacing malloc). - */ -char *util_strdup(const char *s) { - size_t len = 0; - char *ptr = NULL; - - if (!s) - return NULL; - - if ((len = strlen(s)) && (ptr = mem_a(len+1))) { - 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++; - - return util_strdup(s); -} - -/* - * 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; -} +#include "gmqcc.h" /* - * Returns true if string is all digits, otherwise - * it returns false. + * Initially this was handled with a table in the gmqcc.h header, but + * much to my surprise the contents of the table was duplicated for + * each translation unit, causing all these strings to be duplicated + * for every .c file it was included into. This method culls back on + * it. This is a 'utility' function because the executor also depends + * on this for dissasembled bytecode. */ -bool util_strdigit(const char *str) { - while (*str) { - if(!isdigit(*str)) - return false; - str++; - } - return true; -} - -bool util_strncmpexact(const char *src, const char *ned, size_t len) { - return (!strncmp(src, ned, len) && !src[len]); -} +const char *util_instr_str[VINSTR_END] = { + "DONE", "MUL_F", "MUL_V", "MUL_FV", + "MUL_VF", "DIV_F", "ADD_F", "ADD_V", + "SUB_F", "SUB_V", "EQ_F", "EQ_V", + "EQ_S", "EQ_E", "EQ_FNC", "NE_F", + "NE_V", "NE_S", "NE_E", "NE_FNC", + "LE", "GE", "LT", "GT", + "LOAD_F", "LOAD_V", "LOAD_S", "LOAD_ENT", + "LOAD_FLD", "LOAD_FNC", "ADDRESS", "STORE_F", + "STORE_V", "STORE_S", "STORE_ENT", "STORE_FLD", + "STORE_FNC", "STOREP_F", "STOREP_V", "STOREP_S", + "STOREP_ENT", "STOREP_FLD", "STOREP_FNC", "RETURN", + "NOT_F", "NOT_V", "NOT_S", "NOT_ENT", + "NOT_FNC", "IF", "IFNOT", "CALL0", + "CALL1", "CALL2", "CALL3", "CALL4", + "CALL5", "CALL6", "CALL7", "CALL8", + "STATE", "GOTO", "AND", "OR", + "BITAND", "BITOR" +}; 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 +69,75 @@ 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 +154,7 @@ 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 +162,6 @@ 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, @@ -401,340 +208,374 @@ 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. +/* 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; +} +/* 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 - * Quake expects a non-reflective CRC. +/* + * modifier is the match to make and the transpsition from it, while add is the upper-value that determines the + * transposion from uppercase to lower case. */ -#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; \ -} -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; \ +static GMQCC_INLINE size_t util_strtransform(const char *in, char *out, size_t outsz, const char *mod, int add) { + size_t sz = 1; + for (; *in && sz < outsz; ++in, ++out, ++sz) { + *out = (*in == mod[0]) + ? mod[1] + : (util_isalpha(*in) && util_isupper(*in + add)) + ? *in + add + : *in; + } + *out = 0; + return sz-1; } -*/ +size_t util_strtocmd(const char *in, char *out, size_t outsz) { + return util_strtransform(in, out, outsz, "-_", 'A'-'a'); +} +size_t util_strtononcmd(const char *in, char *out, size_t outsz) { + return util_strtransform(in, out, outsz, "_-", 'a'-'A'); +} +size_t util_optimizationtostr(const char *in, char *out, size_t outsz) { + return util_strtransform(in, out, outsz, "_ ", 'a'-'A'); +} /* - * Implements libc getline for systems that don't have it, which is - * assmed all. This works the same as getline(). + * Portable implementation of vasprintf/asprintf. Assumes vsnprintf + * exists, otherwise compiler error. + * + * TODO: fix for MSVC .... */ -int util_getline(char **lineptr, size_t *n, FILE *stream) { - int chr; +int util_vasprintf(char **dat, const char *fmt, va_list args) { int ret; - char *pos; - - if (!lineptr || !n || !stream) - return -1; - if (!*lineptr) { - if (!(*lineptr = (char*)mem_a((*n=64)))) + 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 + if ((len = _vscprintf(fmt, args)) < 0) { + *dat = NULL; 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)) + tmp = (char*)mem_a(len + 1); + if ((ret = _vsnprintf_s(tmp, len+1, len+1, fmt, args)) != len) { + mem_d(tmp); + *dat = NULL; return -1; - if (c == EOF) { - if (pos == *lineptr) - return -1; - else - break; + } + *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; } - *pos++ = c; - chr--; - if (c == '\n') - break; - } - *pos = '\0'; - return (ret = pos - *lineptr); -} - -size_t util_strtocmd(const char *in, char *out, size_t outsz) { - size_t sz = 1; - for (; *in && sz < outsz; ++in, ++out, ++sz) - *out = (*in == '-') ? '_' : (isalpha(*in) && !isupper(*in)) ? *in + 'A' - 'a': *in; - *out = 0; - return sz-1; -} + /* 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; + } -size_t util_strtononcmd(const char *in, char *out, size_t outsz) { - size_t sz = 1; - for (; *in && sz < outsz; ++in, ++out, ++sz) - *out = (*in == '_') ? '-' : (isalpha(*in) && isupper(*in)) ? *in + 'a' - 'A' : *in; - *out = 0; - return sz-1; + *dat = tmp; + return len; + #endif } - - -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 -} - -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); - if (!*a) - ((size_t*)p)[1] = 0; - *a = (void*)((size_t*)p + 2); - _vec_beg(*a) = m; +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; } /* - * Hash table for generic data, based on dynamic memory allocations - * all around. This is the internal interface, please look for - * EXPOSED INTERFACE comment below - */ -typedef struct hash_node_t { - char *key; /* the key for this node in table */ - void *value; /* pointer to the data as void* */ - 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. + * These are various re-implementations (wrapping the real ones) of + * string functions that MSVC consideres unsafe. We wrap these up and + * use the safe varations on MSVC. */ -#ifdef __x86_64__ -GMQCC_INLINE uint32_t util_hthashfunc(hash_table_t *ht, const char *key, size_t seed) { - const uint64_t mix = 0xC6A4A7935BD1E995ULL; - 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; +#ifdef _MSC_VER + static char **util_strerror_allocated() { + static char **data = NULL; + return data; } - 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; + static void util_strerror_cleanup(void) { + size_t i; + char **data = util_strerror_allocated(); + for (i = 0; i < vec_size(data); i++) + mem_d(data[i]); + vec_free(data); } - hash ^= hash >> rot; - hash *= mix; - hash ^= hash >> rot; + const char *util_strerror(int num) { + char *allocated = NULL; + static bool install = false; + static size_t tries = 0; + char **vector = util_strerror_allocated(); - return (uint32_t)(hash % ht->size); -} + /* try installing cleanup handler */ + while (!install) { + if (tries == 32) + return "(unknown)"; -#else -GMQCC_INLINE uint32_t util_hthashfunc(hash_table_t *ht, const char *key, size_t seed) { - const uint32_t mix = 0x5BD1E995; - const uint32_t rot = 24; - size_t size = strlen(key); - uint32_t hash = seed ^ size; - uint32_t alias = 0; - const unsigned char *data = (const unsigned char*)key; - - while (size >= 4) { - alias = *(uint32_t*)data; - - alias *= mix; - alias ^= alias >> rot; - alias *= mix; - - hash *= mix; - hash ^= alias; - - data += 4; - size -= 4; - } + install = !atexit(&util_strerror_cleanup); + tries ++; + } - switch (size) { - case 3: hash ^= data[2] << 16; - case 2: hash ^= data[1] << 8; - case 1: hash ^= data[0]; - hash *= mix; + allocated = (char*)mem_a(4096); /* A page must be enough */ + strerror_s(allocated, 4096, num); + + vec_push(vector, allocated); + return (const char *)allocated; } - hash ^= hash >> 13; - hash *= mix; - hash ^= hash >> 15; + int util_snprintf(char *src, size_t bytes, const char *format, ...) { + int rt; + va_list va; + va_start(va, format); - return hash % ht->size; -} -#endif + rt = vsprintf_s(src, bytes, format, va); + va_end (va); -/* 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]); + return rt; + } - /* reset seed */ - if (seed >= sizeof(util_crc32_table) / sizeof(*util_crc32_table)) - seed = 0; + char *util_strcat(char *dest, const char *src) { + strcat_s(dest, strlen(src), src); + return dest; + } - return hash; -} + char *util_strncpy(char *dest, const char *src, size_t num) { + strncpy_s(dest, num, src, num); + return dest; + } +#else + const char *util_strerror(int num) { + return strerror(num); + } -hash_node_t *_util_htnewpair(const char *key, void *value) { - hash_node_t *node; - if (!(node = mem_a(sizeof(hash_node_t)))) - return NULL; + int util_snprintf(char *src, size_t bytes, const char *format, ...) { + int rt; + va_list va; + va_start(va, format); + rt = vsnprintf(src, bytes, format, va); + va_end (va); - if (!(node->key = util_strdup(key))) { - mem_d(node); - return NULL; + return rt; } - node->value = value; - node->next = NULL; + char *util_strcat(char *dest, const char *src) { + return strcat(dest, src); + } - return node; -} + char *util_strncpy(char *dest, const char *src, size_t num) { + return strncpy(dest, src, num); + } + +#endif /*! _MSC_VER */ /* - * EXPOSED INTERFACE for the hashtable implementation - * util_htnew(size) -- to make a new hashtable - * util_htset(table, key, value, sizeof(value)) -- to set something in the table - * util_htget(table, key) -- to get something from the table - * util_htdel(table) -- to delete the table + * 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). */ -hash_table_t *util_htnew(size_t size) { - hash_table_t *hashtable = NULL; - if (size < 1) - return NULL; - - if (!(hashtable = mem_a(sizeof(hash_table_t)))) - return NULL; - if (!(hashtable->table = mem_a(sizeof(hash_node_t*) * size))) { - mem_d(hashtable); - return NULL; +#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(void) { + /* + * 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-1; ++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]; } - hashtable->size = size; - memset(hashtable->table, 0, sizeof(hash_node_t*) * size); - - return hashtable; -} - -void util_htseth(hash_table_t *ht, const char *key, size_t bin, void *value) { - hash_node_t *newnode = NULL; - hash_node_t *next = NULL; - hash_node_t *last = NULL; - - next = ht->table[bin]; - - while (next && next->key && strcmp(key, next->key) > 0) - last = next, next = next->next; - - /* already in table, do a replace */ - if (next && next->key && strcmp(key, next->key) == 0) { - next->value = value; - } else { - /* not found, grow a pair man :P */ - newnode = _util_htnewpair(key, value); - if (next == ht->table[bin]) { - newnode->next = next; - ht->table[bin] = newnode; - } else if (!next) { - last->next = newnode; - } else { - newnode->next = next; - last->next = newnode; - } + /* + * collapsing the walls unrolled (evenly dividing 396 [632-227 = 396 + * = 2*2*3*3*11]) + */ + i = MT_SPACE; + while (i < MT_SIZE-2) { + /* + * 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_htset(hash_table_t *ht, const char *key, void *value) { - util_htseth(ht, key, util_hthash(ht, key), value); +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; } -void *util_htgeth(hash_table_t *ht, const char *key, size_t bin) { - hash_node_t *pair = ht->table[bin]; +uint32_t util_rand() { + register uint32_t y; - while (pair && pair->key && strcmp(key, pair->key) > 0) - pair = pair->next; + /* + * 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(); - if (!pair || !pair->key || strcmp(key, pair->key) != 0) - return NULL; + y = mt_state[mt_index]; - return pair->value; -} + /* Standard tempering */ + y ^= y >> 11; /* +7 */ + y ^= y << 7 & 0x9D2C5680; /* +4 */ + y ^= y << 15 & 0xEFC60000; /* -4 */ + y ^= y >> 18; /* -7 */ -void *util_htget(hash_table_t *ht, const char *key) { - return util_htgeth(ht, key, util_hthash(ht, key)); -} + if(++mt_index == MT_SIZE) + mt_index = 0; -/* - * Free all allocated data in a hashtable, this is quite the amount - * of work. - */ -void util_htdel(hash_table_t *ht) { - size_t i = 0; - for (; i < ht->size; i++) { - hash_node_t *n = ht->table[i]; - hash_node_t *p; - - /* free in list */ - while (n) { - if (n->key) - mem_d(n->key); - p = n; - n = n->next; - mem_d(p); - } - - } - /* free table */ - mem_d(ht->table); - mem_d(ht); + return y; }