2 * Copyright (C) 2012, 2013
6 * Permission is hereby granted, free of charge, to any person obtaining a copy of
7 * this software and associated documentation files (the "Software"), to deal in
8 * the Software without restriction, including without limitation the rights to
9 * use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
10 * of the Software, and to permit persons to whom the Software is furnished to do
11 * so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included in all
14 * copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
28 /* TODO: remove globals ... */
38 struct memblock_t *next;
39 struct memblock_t *prev;
42 static struct memblock_t *mem_start = NULL;
44 void *util_memory_a(size_t byte, unsigned int line, const char *file) {
45 struct memblock_t *info = (struct memblock_t*)malloc(sizeof(struct memblock_t) + byte);
46 void *data = (void*)(info+1);
47 if (!info) return NULL;
52 info->next = mem_start;
54 mem_start->prev = info;
63 void util_memory_d(void *ptrn) {
64 struct memblock_t *info = NULL;
67 info = ((struct memblock_t*)ptrn - 1);
73 info->prev->next = info->next;
75 info->next->prev = info->prev;
76 if (info == mem_start)
77 mem_start = info->next;
82 void *util_memory_r(void *ptrn, size_t byte, unsigned int line, const char *file) {
83 struct memblock_t *oldinfo = NULL;
85 struct memblock_t *newinfo;
88 return util_memory_a(byte, line, file);
94 oldinfo = ((struct memblock_t*)ptrn - 1);
95 newinfo = ((struct memblock_t*)malloc(sizeof(struct memblock_t) + byte));
99 util_memory_d(oldinfo+1);
104 memcpy(newinfo+1, oldinfo+1, oldinfo->byte);
108 oldinfo->prev->next = oldinfo->next;
110 oldinfo->next->prev = oldinfo->prev;
111 if (oldinfo == mem_start)
112 mem_start = oldinfo->next;
115 newinfo->line = line;
116 newinfo->byte = byte;
117 newinfo->file = file;
118 newinfo->prev = NULL;
119 newinfo->next = mem_start;
121 mem_start->prev = newinfo;
124 mem_ab -= oldinfo->byte;
125 mem_ab += newinfo->byte;
132 void util_meminfo() {
133 struct memblock_t *info;
138 for (info = mem_start; info; info = info->next) {
139 util_debug("MEM", "lost: % 8u (bytes) at %s:%u\n",
145 util_debug("MEM", "Memory information:\n\
146 Total allocations: %llu\n\
147 Total deallocations: %llu\n\
148 Total allocated: %llu (bytes)\n\
149 Total deallocated: %llu (bytes)\n\
150 Leaks found: lost %llu (bytes) in %d allocations\n",
159 * Some string utility functions, because strdup uses malloc, and we want
160 * to track all memory (without replacing malloc).
162 char *util_strdup(const char *s) {
169 if ((len = strlen(s)) && (ptr = (char*)mem_a(len+1))) {
176 void util_debug(const char *area, const char *ms, ...) {
181 if (!strcmp(area, "MEM") && !opts.memchk)
185 con_out ("[%s] ", area);
191 * only required if big endian .. otherwise no need to swap
194 #if PLATFORM_BYTE_ORDER == GMQCC_BYTE_ORDER_BIG
195 static GMQCC_INLINE void util_swap16(uint16_t *d, size_t l) {
197 d[l] = (d[l] << 8) | (d[l] >> 8);
201 static GMQCC_INLINE void util_swap32(uint32_t *d, size_t l) {
204 v = ((d[l] << 8) & 0xFF00FF00) | ((d[l] >> 8) & 0x00FF00FF);
205 d[l] = (v << 16) | (v >> 16);
209 /* Some strange system doesn't like constants that big, AND doesn't recognize an ULL suffix
210 * so let's go the safe way
212 static GMQCC_INLINE void util_swap64(uint32_t *d, size_t l) {
216 v = ((d[l] << 8) & 0xFF00FF00FF00FF00) | ((d[l] >> 8) & 0x00FF00FF00FF00FF);
217 v = ((v << 16) & 0xFFFF0000FFFF0000) | ((v >> 16) & 0x0000FFFF0000FFFF);
218 d[l] = (v << 32) | (v >> 32);
222 for (i = 0; i < l; i += 2) {
231 void util_endianswap(void *_data, size_t length, unsigned int typesize) {
232 # if PLATFORM_BYTE_ORDER == -1 /* runtime check */
233 if (*((char*)&typesize))
236 /* prevent unused warnings */
241 # if PLATFORM_BYTE_ORDER == GMQCC_BYTE_ORDER_LITTLE
247 util_swap16((uint16_t*)_data, length>>1);
250 util_swap32((uint32_t*)_data, length>>2);
253 util_swap64((uint32_t*)_data, length>>3);
256 default: abort(); /* please blow the fuck up! */
263 * CRC algorithms vary in the width of the polynomial, the value of said polynomial,
264 * the initial value used for the register, weather the bits of each byte are reflected
265 * before being processed, weather the algorithm itself feeds input bytes through the
266 * register or XORs them with a byte from one end and then straight into the table, as
267 * well as (but not limited to the idea of reflected versions) where the final register
268 * value becomes reversed, and finally weather the value itself is used to XOR the final
269 * register value. AS such you can already imagine how painfully annoying CRCs are,
270 * of course we stand to target Quake, which expects it's certian set of rules for proper
271 * calculation of a CRC.
273 * In most traditional CRC algorithms on uses a reflected table driven method where a value
274 * or register is reflected if it's bits are swapped around it's center. For example:
275 * take the bits 0101 is the 4-bit reflection of 1010, and respectfully 0011 would be the
276 * reflection of 1100. Quake however expects a NON-Reflected CRC on the output, but still
277 * requires a final XOR on the values (0xFFFF and 0x0000) this is a standard CCITT CRC-16
278 * which I respectfully as a programmer don't agree with.
280 * So now you know what we target, and why we target it, despite how unsettling it may seem
281 * but those are what Quake seems to request.
284 static const uint16_t util_crc16_table[] = {
285 0x0000, 0x1021, 0x2042, 0x3063, 0x4084, 0x50A5,
286 0x60C6, 0x70E7, 0x8108, 0x9129, 0xA14A, 0xB16B,
287 0xC18C, 0xD1AD, 0xE1CE, 0xF1EF, 0x1231, 0x0210,
288 0x3273, 0x2252, 0x52B5, 0x4294, 0x72F7, 0x62D6,
289 0x9339, 0x8318, 0xB37B, 0xA35A, 0xD3BD, 0xC39C,
290 0xF3FF, 0xE3DE, 0x2462, 0x3443, 0x0420, 0x1401,
291 0x64E6, 0x74C7, 0x44A4, 0x5485, 0xA56A, 0xB54B,
292 0x8528, 0x9509, 0xE5EE, 0xF5CF, 0xC5AC, 0xD58D,
293 0x3653, 0x2672, 0x1611, 0x0630, 0x76D7, 0x66F6,
294 0x5695, 0x46B4, 0xB75B, 0xA77A, 0x9719, 0x8738,
295 0xF7DF, 0xE7FE, 0xD79D, 0xC7BC, 0x48C4, 0x58E5,
296 0x6886, 0x78A7, 0x0840, 0x1861, 0x2802, 0x3823,
297 0xC9CC, 0xD9ED, 0xE98E, 0xF9AF, 0x8948, 0x9969,
298 0xA90A, 0xB92B, 0x5AF5, 0x4AD4, 0x7AB7, 0x6A96,
299 0x1A71, 0x0A50, 0x3A33, 0x2A12, 0xDBFD, 0xCBDC,
300 0xFBBF, 0xEB9E, 0x9B79, 0x8B58, 0xBB3B, 0xAB1A,
301 0x6CA6, 0x7C87, 0x4CE4, 0x5CC5, 0x2C22, 0x3C03,
302 0x0C60, 0x1C41, 0xEDAE, 0xFD8F, 0xCDEC, 0xDDCD,
303 0xAD2A, 0xBD0B, 0x8D68, 0x9D49, 0x7E97, 0x6EB6,
304 0x5ED5, 0x4EF4, 0x3E13, 0x2E32, 0x1E51, 0x0E70,
305 0xFF9F, 0xEFBE, 0xDFDD, 0xCFFC, 0xBF1B, 0xAF3A,
306 0x9F59, 0x8F78, 0x9188, 0x81A9, 0xB1CA, 0xA1EB,
307 0xD10C, 0xC12D, 0xF14E, 0xE16F, 0x1080, 0x00A1,
308 0x30C2, 0x20E3, 0x5004, 0x4025, 0x7046, 0x6067,
309 0x83B9, 0x9398, 0xA3FB, 0xB3DA, 0xC33D, 0xD31C,
310 0xE37F, 0xF35E, 0x02B1, 0x1290, 0x22F3, 0x32D2,
311 0x4235, 0x5214, 0x6277, 0x7256, 0xB5EA, 0xA5CB,
312 0x95A8, 0x8589, 0xF56E, 0xE54F, 0xD52C, 0xC50D,
313 0x34E2, 0x24C3, 0x14A0, 0x0481, 0x7466, 0x6447,
314 0x5424, 0x4405, 0xA7DB, 0xB7FA, 0x8799, 0x97B8,
315 0xE75F, 0xF77E, 0xC71D, 0xD73C, 0x26D3, 0x36F2,
316 0x0691, 0x16B0, 0x6657, 0x7676, 0x4615, 0x5634,
317 0xD94C, 0xC96D, 0xF90E, 0xE92F, 0x99C8, 0x89E9,
318 0xB98A, 0xA9AB, 0x5844, 0x4865, 0x7806, 0x6827,
319 0x18C0, 0x08E1, 0x3882, 0x28A3, 0xCB7D, 0xDB5C,
320 0xEB3F, 0xFB1E, 0x8BF9, 0x9BD8, 0xABBB, 0xBB9A,
321 0x4A75, 0x5A54, 0x6A37, 0x7A16, 0x0AF1, 0x1AD0,
322 0x2AB3, 0x3A92, 0xFD2E, 0xED0F, 0xDD6C, 0xCD4D,
323 0xBDAA, 0xAD8B, 0x9DE8, 0x8DC9, 0x7C26, 0x6C07,
324 0x5C64, 0x4C45, 0x3CA2, 0x2C83, 0x1CE0, 0x0CC1,
325 0xEF1F, 0xFF3E, 0xCF5D, 0xDF7C, 0xAF9B, 0xBFBA,
326 0x8FD9, 0x9FF8, 0x6E17, 0x7E36, 0x4E55, 0x5E74,
327 0x2E93, 0x3EB2, 0x0ED1, 0x1EF0
330 /* Non - Reflected */
331 uint16_t util_crc16(uint16_t current, const char *k, size_t len) {
332 register uint16_t h = current;
333 for (; len; --len, ++k)
334 h = util_crc16_table[(h>>8)^((unsigned char)*k)]^(h<<8);
337 /* Reflective Varation (for reference) */
339 uint16_t util_crc16(const char *k, int len, const short clamp) {
340 register uint16_t h= (uint16_t)0xFFFFFFFF;
341 for (; len; --len, ++k)
342 h = util_crc16_table[(h^((unsigned char)*k))&0xFF]^(h>>8);
347 size_t util_strtocmd(const char *in, char *out, size_t outsz) {
349 for (; *in && sz < outsz; ++in, ++out, ++sz)
350 *out = (*in == '-') ? '_' : (isalpha(*in) && !isupper(*in)) ? *in + 'A' - 'a': *in;
355 size_t util_strtononcmd(const char *in, char *out, size_t outsz) {
357 for (; *in && sz < outsz; ++in, ++out, ++sz)
358 *out = (*in == '_') ? '-' : (isalpha(*in) && isupper(*in)) ? *in + 'a' - 'A' : *in;
363 /* TODO: rewrite ... when I redo the ve cleanup */
364 void _util_vec_grow(void **a, size_t i, size_t s) {
365 vector_t *d = vec_meta(*a);
366 size_t m = *a ? 2 * d->allocated +i : i+1;
367 void *p = mem_r((*a ? d : NULL), s * m + sizeof(vector_t));
370 ((vector_t*)p)->used = 0;
371 *a = (vector_t*)p + 1;
373 vec_meta(*a)->allocated = m;
377 * Hash table for generic data, based on dynamic memory allocations
378 * all around. This is the internal interface, please look for
379 * EXPOSED INTERFACE comment below
381 typedef struct hash_node_t {
382 char *key; /* the key for this node in table */
383 void *value; /* pointer to the data as void* */
384 struct hash_node_t *next; /* next node (linked list) */
387 GMQCC_INLINE size_t util_hthash(hash_table_t *ht, const char *key) {
388 const uint32_t mix = 0x5BD1E995;
389 const uint32_t rot = 24;
390 size_t size = strlen(key);
391 uint32_t hash = 0x1EF0 /* LICRC TAB */ ^ size;
393 const unsigned char *data = (const unsigned char*)key;
396 alias = *(uint32_t*)data;
399 alias ^= alias >> rot;
410 case 3: hash ^= data[2] << 16;
411 case 2: hash ^= data[1] << 8;
412 case 1: hash ^= data[0];
420 return (size_t) (hash % ht->size);
423 hash_node_t *_util_htnewpair(const char *key, void *value) {
425 if (!(node = (hash_node_t*)mem_a(sizeof(hash_node_t))))
428 if (!(node->key = util_strdup(key))) {
440 * EXPOSED INTERFACE for the hashtable implementation
441 * util_htnew(size) -- to make a new hashtable
442 * util_htset(table, key, value, sizeof(value)) -- to set something in the table
443 * util_htget(table, key) -- to get something from the table
444 * util_htdel(table) -- to delete the table
446 hash_table_t *util_htnew(size_t size) {
447 hash_table_t *hashtable = NULL;
451 if (!(hashtable = (hash_table_t*)mem_a(sizeof(hash_table_t))))
454 if (!(hashtable->table = (hash_node_t**)mem_a(sizeof(hash_node_t*) * size))) {
459 hashtable->size = size;
460 memset(hashtable->table, 0, sizeof(hash_node_t*) * size);
465 void util_htseth(hash_table_t *ht, const char *key, size_t bin, void *value) {
466 hash_node_t *newnode = NULL;
467 hash_node_t *next = NULL;
468 hash_node_t *last = NULL;
470 next = ht->table[bin];
472 while (next && next->key && strcmp(key, next->key) > 0)
473 last = next, next = next->next;
475 /* already in table, do a replace */
476 if (next && next->key && strcmp(key, next->key) == 0) {
479 /* not found, grow a pair man :P */
480 newnode = _util_htnewpair(key, value);
481 if (next == ht->table[bin]) {
482 newnode->next = next;
483 ht->table[bin] = newnode;
485 last->next = newnode;
487 newnode->next = next;
488 last->next = newnode;
493 void util_htset(hash_table_t *ht, const char *key, void *value) {
494 util_htseth(ht, key, util_hthash(ht, key), value);
497 void *util_htgeth(hash_table_t *ht, const char *key, size_t bin) {
498 hash_node_t *pair = ht->table[bin];
500 while (pair && pair->key && strcmp(key, pair->key) > 0)
503 if (!pair || !pair->key || strcmp(key, pair->key) != 0)
509 void *util_htget(hash_table_t *ht, const char *key) {
510 return util_htgeth(ht, key, util_hthash(ht, key));
513 void *code_util_str_htgeth(hash_table_t *ht, const char *key, size_t bin) {
518 keylen = strlen(key);
520 pair = ht->table[bin];
521 while (pair && pair->key) {
522 len = strlen(pair->key);
528 cmp = strcmp(key, pair->key);
536 cmp = strcmp(key, pair->key + len - keylen);
538 uintptr_t up = (uintptr_t)pair->value;
548 * Free all allocated data in a hashtable, this is quite the amount
551 void util_htdel(hash_table_t *ht) {
553 for (; i < ht->size; i++) {
554 hash_node_t *n = ht->table[i];
573 * A basic implementation of a hash-set. Unlike a hashtable, a hash
574 * set doesn't maintain key-value pairs. It simply maintains a key
575 * that can be set, removed, and checked for.
577 * See EXPOSED interface comment below
579 #define GMQCC_HASHSET_PRIME0 0x0049
580 #define GMQCC_HASHSET_PRIME1 0x1391
582 static int util_hsput(hash_set_t *set, void *item) {
583 size_t hash = (size_t)item; /* shouldn't drop the bits */
586 /* a == 0 || a == 1 */
590 iter = set->mask & (GMQCC_HASHSET_PRIME0 * hash);
592 /* while (set->items[iter] != 0 && set->items[iter] != 1) */
593 while (!(set->items[iter] >> 1)) {
594 if (set->items[iter] == hash)
597 iter = set->mask & (iter + GMQCC_HASHSET_PRIME1);
601 set->items[iter] = hash;
606 static void util_hsupdate(hash_set_t *set) {
611 /* time to rehash? */
612 if ((float)set->total >= (size_t)((double)set->capacity * 0.85)) {
617 set->capacity = (size_t)(1 << set->bits);
618 set->mask = set->capacity - 1;
619 set->items = mem_a(set->capacity * sizeof(size_t));
622 /*assert(set->items);*/
625 * this shouldn't be slow? if so unroll it a little perhaps
626 * (shouldn't be though)
628 for (itr = 0; itr < end; itr++)
629 util_hsput(set, (void*)old[itr]);
636 * EXPOSED interface: all of these functions are exposed to the outside
637 * for use. The stuff above is static because it's the "internal" mechanics
638 * for syncronizing the set for updating, and putting data into the set.
640 int util_hsadd(hash_set_t *set, void *item) {
641 int run = util_hsput(set, item); /* inlined */
647 /* remove item in set */
648 int util_hsrem(hash_set_t *set, void *item) {
649 size_t hash = (size_t)item;
650 size_t iter = set->mask & (GMQCC_HASHSET_PRIME0 * hash);
652 while (set->items[iter]) {
653 if (set->items[iter] == hash) {
654 set->items[iter] = 1;
659 iter = set->mask & (iter + GMQCC_HASHSET_PRIME1);
665 /* check if item is set */
666 int util_hshas(hash_set_t *set, void *item) {
667 size_t hash = (size_t)item;
668 size_t iter = set->mask & (GMQCC_HASHSET_PRIME0 * hash);
670 while (set->items[iter]) {
671 if (set->items[iter] == hash)
674 iter = set->mask & (iter + GMQCC_HASHSET_PRIME1);
680 hash_set_t *util_hsnew(void) {
683 if (!(set = mem_a(sizeof(hash_set_t))))
688 set->capacity = (size_t)(1 << set->bits);
689 set->mask = set->capacity - 1;
690 set->items = mem_a(set->capacity * sizeof(size_t));
700 void util_hsdel(hash_set_t *set) {
708 #undef GMQCC_HASHSET_PRIME0
709 #undef GMQCC_HASHSET_PRIME1
713 * Portable implementation of vasprintf/asprintf. Assumes vsnprintf
714 * exists, otherwise compiler error.
716 * TODO: fix for MSVC ....
718 int util_vasprintf(char **dat, const char *fmt, va_list args) {
724 * For visuals tido _vsnprintf doesn't tell you the length of a
725 * formatted string if it overflows. However there is a MSVC
726 * intrinsic (which is documented wrong) called _vcsprintf which
727 * will return the required amount to allocate.
731 if ((len = _vscprintf(fmt, args)) < 0) {
736 tmp = mem_a(len + 1);
737 if ((ret = _vsnprintf(tmp, len+1, fmt, args)) != len) {
746 * For everything else we have a decent conformint vsnprintf that
747 * returns the number of bytes needed. We give it a try though on
748 * a short buffer, since efficently speaking, it could be nice to
749 * above a second vsnprintf call.
754 len = vsnprintf(buf, sizeof(buf), fmt, cpy);
757 if (len < (int)sizeof(buf)) {
758 *dat = util_strdup(buf);
762 /* not large enough ... */
763 tmp = mem_a(len + 1);
764 if ((ret = vsnprintf(tmp, len + 1, fmt, args)) != len) {
774 int util_asprintf(char **ret, const char *fmt, ...) {
778 read = util_vasprintf(ret, fmt, args);
785 * Implementation of the Mersenne twister PRNG (pseudo random numer
786 * generator). Implementation of MT19937. Has a period of 2^19937-1
787 * which is a Mersenne Prime (hence the name).
789 * Implemented from specification and original paper:
790 * http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/ARTICLES/mt.pdf
792 * This code is placed in the public domain by me personally
793 * (Dale Weiler, a.k.a graphitemaster).
797 #define MT_PERIOD 397
798 #define MT_SPACE (MT_SIZE - MT_PERIOD)
800 static uint32_t mt_state[MT_SIZE];
801 static size_t mt_index = 0;
803 static GMQCC_INLINE void mt_generate() {
805 * The loop has been unrolled here: the original paper and implemenation
806 * Called for the following code:
807 * for (register unsigned i = 0; i < MT_SIZE; ++i) {
808 * register uint32_t load;
809 * load = (0x80000000 & mt_state[i]) // most significant 32nd bit
810 * load |= (0x7FFFFFFF & mt_state[(i + 1) % MT_SIZE]) // least significant 31nd bit
812 * mt_state[i] = mt_state[(i + MT_PERIOD) % MT_SIZE] ^ (load >> 1);
814 * if (load & 1) mt_state[i] ^= 0x9908B0DF;
817 * This essentially is a waste: we have two modulus operations, and
818 * a branch that is executed every iteration from [0, MT_SIZE).
820 * Please see: http://www.quadibloc.com/crypto/co4814.htm for more
821 * information on how this clever trick works.
823 static const uint32_t matrix[2] = {
828 * This register gives up a little more speed by instructing the compiler
829 * to force these into CPU registers (they're counters for indexing mt_state
830 * which we can force the compiler to generate prefetch instructions for)
836 * Said loop has been unrolled for MT_SPACE (226 iterations), opposed
837 * to [0, MT_SIZE) (634 iterations).
839 for (i = 0; i < MT_SPACE; ++i) {
840 y = (0x80000000 & mt_state[i]) | (0x7FFFFFF & mt_state[i + 1]);
841 mt_state[i] = mt_state[i + MT_PERIOD] ^ (y >> 1) ^ matrix[y & 1];
843 i ++; /* loop unroll */
845 y = (0x80000000 & mt_state[i]) | (0x7FFFFFF & mt_state[i + 1]);
846 mt_state[i] = mt_state[i + MT_PERIOD] ^ (y >> 1) ^ matrix[y & 1];
850 * collapsing the walls unrolled (evenly dividing 396 [632-227 = 396
854 while (i < MT_SIZE - 1) {
856 * We expand this 11 times .. manually, no macros are required
857 * here. This all fits in the CPU cache.
859 y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]);
860 mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1];
862 y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]);
863 mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1];
865 y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]);
866 mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1];
868 y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]);
869 mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1];
871 y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]);
872 mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1];
874 y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]);
875 mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1];
877 y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]);
878 mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1];
880 y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]);
881 mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1];
883 y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]);
884 mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1];
886 y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]);
887 mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1];
889 y = (0x80000000 & mt_state[i]) | (0x7FFFFFFF & mt_state[i + 1]);
890 mt_state[i] = mt_state[i - MT_SPACE] ^ (y >> 1) ^ matrix[y & 1];
894 /* i = mt_state[623] */
895 y = (0x80000000 & mt_state[MT_SIZE - 1]) | (0x7FFFFFFF & mt_state[MT_SIZE - 1]);
896 mt_state[MT_SIZE - 1] = mt_state[MT_PERIOD - 1] ^ (y >> 1) ^ matrix[y & 1];
899 void util_seed(uint32_t value) {
901 * We seed the mt_state with a LCG (linear congruential generator)
902 * We're operating exactly on exactly m=32, so there is no need to
905 * The multipler of choice is 0x6C07865, also knows as the Borosh-
906 * Niederreiter multipler used for modulus 2^32. More can be read
907 * about this in Knuth's TAOCP Volume 2, page 106.
909 * If you don't own TAOCP something is wrong with you :-) .. so I
910 * also provided a link to the original paper by Borosh and
911 * Niederreiter. It's called "Optional Multipliers for PRNG by The
912 * Linear Congruential Method" (1983).
913 * http://en.wikipedia.org/wiki/Linear_congruential_generator
915 * From said page, it says the following:
916 * "A common Mersenne twister implementation, interestingly enough
917 * used an LCG to generate seed data."
920 * The data we're operating on is 32-bits for the mt_state array, so
921 * there is no masking required with 0xFFFFFFFF
926 for (i = 1; i < MT_SIZE; ++i)
927 mt_state[i] = 0x6C078965 * (mt_state[i - 1] ^ mt_state[i - 1] >> 30) + i;
930 uint32_t util_rand() {
934 * This is inlined with any sane compiler (I checked)
935 * for some reason though, SubC seems to be generating invalid
936 * code when it inlines this.
941 y = mt_state[mt_index];
943 /* Standard tempering */
944 y ^= y >> 11; /* +7 */
945 y ^= y << 7 & 0x9D2C5680; /* +4 */
946 y ^= y << 15 & 0xEFC60000; /* -4 */
947 y ^= y >> 18; /* -7 */
949 if(++mt_index == MT_SIZE)