ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/deliantra/server/include/util.h
(Generate patch)

Comparing deliantra/server/include/util.h (file contents):
Revision 1.84 by root, Wed Dec 31 17:35:37 2008 UTC vs.
Revision 1.115 by root, Tue Apr 26 14:41:36 2011 UTC

1/* 1/*
2 * This file is part of Deliantra, the Roguelike Realtime MMORPG. 2 * This file is part of Deliantra, the Roguelike Realtime MMORPG.
3 * 3 *
4 * Copyright (©) 2005,2006,2007,2008 Marc Alexander Lehmann / Robin Redeker / the Deliantra team 4 * Copyright (©) 2005,2006,2007,2008,2009,2010,2011 Marc Alexander Lehmann / Robin Redeker / the Deliantra team
5 * 5 *
6 * Deliantra is free software: you can redistribute it and/or modify 6 * Deliantra is free software: you can redistribute it and/or modify it under
7 * it under the terms of the GNU General Public License as published by 7 * the terms of the Affero GNU General Public License as published by the
8 * the Free Software Foundation, either version 3 of the License, or 8 * Free Software Foundation, either version 3 of the License, or (at your
9 * (at your option) any later version. 9 * option) any later version.
10 * 10 *
11 * This program is distributed in the hope that it will be useful, 11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of 12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details. 14 * GNU General Public License for more details.
15 * 15 *
16 * You should have received a copy of the GNU General Public License 16 * You should have received a copy of the Affero GNU General Public License
17 * along with this program. If not, see <http://www.gnu.org/licenses/>. 17 * and the GNU General Public License along with this program. If not, see
18 * <http://www.gnu.org/licenses/>.
18 * 19 *
19 * The authors can be reached via e-mail to <support@deliantra.net> 20 * The authors can be reached via e-mail to <support@deliantra.net>
20 */ 21 */
21 22
22#ifndef UTIL_H__ 23#ifndef UTIL_H__
23#define UTIL_H__ 24#define UTIL_H__
24 25
26#include <compiler.h>
27
25#define DEBUG_POISON 0x00 // poison memory before freeing it if != 0 28#define DEBUG_POISON 0x00 // poison memory before freeing it if != 0
26#define DEBUG_SALLOC 0 // add a debug wrapper around all sallocs 29#define DEBUG_SALLOC 0 // add a debug wrapper around all sallocs
27#define PREFER_MALLOC 0 // use malloc and not the slice allocator 30#define PREFER_MALLOC 0 // use malloc and not the slice allocator
28
29#if __GNUC__ >= 3
30# define is_constant(c) __builtin_constant_p (c)
31# define expect(expr,value) __builtin_expect ((expr),(value))
32# define prefetch(addr,rw,locality) __builtin_prefetch (addr, rw, locality)
33#else
34# define is_constant(c) 0
35# define expect(expr,value) (expr)
36# define prefetch(addr,rw,locality)
37#endif
38
39#if __GNUC__ < 4 || (__GNUC__ == 4 || __GNUC_MINOR__ < 4)
40# define decltype(x) typeof(x)
41#endif
42
43// put into ifs if you are very sure that the expression
44// is mostly true or mosty false. note that these return
45// booleans, not the expression.
46#define expect_false(expr) expect ((expr) ? 1 : 0, 0)
47#define expect_true(expr) expect ((expr) ? 1 : 0, 1)
48 31
49#include <pthread.h> 32#include <pthread.h>
50 33
51#include <cstddef> 34#include <cstddef>
52#include <cmath> 35#include <cmath>
… …
72#endif 55#endif
73 56
74// use C0X decltype for auto declarations until ISO C++ sanctifies them (if ever) 57// use C0X decltype for auto declarations until ISO C++ sanctifies them (if ever)
75#define auto(var,expr) decltype(expr) var = (expr) 58#define auto(var,expr) decltype(expr) var = (expr)
76 59
60#if cplusplus_does_not_suck
61// does not work for local types (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm)
62template<typename T, int N>
63static inline int array_length (const T (&arr)[N])
64{
65 return N;
66}
67#else
68#define array_length(name) (sizeof (name) / sizeof (name [0]))
69#endif
70
77// very ugly macro that basically declares and initialises a variable 71// very ugly macro that basically declares and initialises a variable
78// that is in scope for the next statement only 72// that is in scope for the next statement only
79// works only for stuff that can be assigned 0 and converts to false 73// works only for stuff that can be assigned 0 and converts to false
80// (note: works great for pointers) 74// (note: works great for pointers)
81// most ugly macro I ever wrote 75// most ugly macro I ever wrote
… …
110// sign returns -1 or +1 104// sign returns -1 or +1
111template<typename T> 105template<typename T>
112static inline T sign (T v) { return v < 0 ? -1 : +1; } 106static inline T sign (T v) { return v < 0 ? -1 : +1; }
113// relies on 2c representation 107// relies on 2c representation
114template<> 108template<>
115inline sint8 sign (sint8 v) { return 1 - (sint8 (uint8 (v) >> 7) * 2); } 109inline sint8 sign (sint8 v) { return 1 - (sint8 (uint8 (v) >> 7) * 2); }
110template<>
111inline sint16 sign (sint16 v) { return 1 - (sint16 (uint16 (v) >> 15) * 2); }
112template<>
113inline sint32 sign (sint32 v) { return 1 - (sint32 (uint32 (v) >> 31) * 2); }
116 114
117// sign0 returns -1, 0 or +1 115// sign0 returns -1, 0 or +1
118template<typename T> 116template<typename T>
119static inline T sign0 (T v) { return v ? sign (v) : 0; } 117static inline T sign0 (T v) { return v ? sign (v) : 0; }
120 118
119//clashes with C++0x
120template<typename T, typename U>
121static inline T copysign (T a, U b) { return a > 0 ? b : -b; }
122
123// div* only work correctly for div > 0
121// div, with correct rounding (< 0.5 downwards, >=0.5 upwards) 124// div, with correct rounding (< 0.5 downwards, >=0.5 upwards)
122template<typename T> static inline T div (T val, T div) { return (val + div / 2) / div; } 125template<typename T> static inline T div (T val, T div)
126{
127 return expect_false (val < 0) ? - ((-val + (div - 1) / 2) / div) : (val + div / 2) / div;
128}
129
130template<> inline float div (float val, float div) { return val / div; }
131template<> inline double div (double val, double div) { return val / div; }
132
123// div, round-up 133// div, round-up
124template<typename T> static inline T div_ru (T val, T div) { return (val + div - 1) / div; } 134template<typename T> static inline T div_ru (T val, T div)
135{
136 return expect_false (val < 0) ? - ((-val ) / div) : (val + div - 1) / div;
137}
125// div, round-down 138// div, round-down
126template<typename T> static inline T div_rd (T val, T div) { return (val ) / div; } 139template<typename T> static inline T div_rd (T val, T div)
140{
141 return expect_false (val < 0) ? - ((-val + (div - 1) ) / div) : (val ) / div;
142}
127 143
144// lerp* only work correctly for min_in < max_in
145// Linear intERPolate, scales val from min_in..max_in to min_out..max_out
128template<typename T> 146template<typename T>
129static inline T 147static inline T
130lerp (T val, T min_in, T max_in, T min_out, T max_out) 148lerp (T val, T min_in, T max_in, T min_out, T max_out)
131{ 149{
132 return min_out + div <T> ((val - min_in) * (max_out - min_out), max_in - min_in); 150 return min_out + div <T> ((val - min_in) * (max_out - min_out), max_in - min_in);
… …
191 int32_t d = b - a; 209 int32_t d = b - a;
192 d &= d >> 31; 210 d &= d >> 31;
193 return b - d; 211 return b - d;
194} 212}
195 213
196// this is much faster than crossfires original algorithm 214// this is much faster than crossfire's original algorithm
197// on modern cpus 215// on modern cpus
198inline int 216inline int
199isqrt (int n) 217isqrt (int n)
200{ 218{
201 return (int)sqrtf ((float)n); 219 return (int)sqrtf ((float)n);
220}
221
222// this is kind of like the ^^ operator, if it would exist, without sequence point.
223// more handy than it looks like, due to the implicit !! done on its arguments
224inline bool
225logical_xor (bool a, bool b)
226{
227 return a != b;
228}
229
230inline bool
231logical_implies (bool a, bool b)
232{
233 return a <= b;
202} 234}
203 235
204// this is only twice as fast as naive sqrtf (dx*dy+dy*dy) 236// this is only twice as fast as naive sqrtf (dx*dy+dy*dy)
205#if 0 237#if 0
206// and has a max. error of 6 in the range -100..+100. 238// and has a max. error of 6 in the range -100..+100.
… …
220#else 252#else
221 return dx_ + dy_ - min (dx_, dy_) * 5 / 8; 253 return dx_ + dy_ - min (dx_, dy_) * 5 / 8;
222#endif 254#endif
223} 255}
224 256
257// can be substantially faster than floor, if your value range allows for it
258template<typename T>
259inline T
260fastfloor (T x)
261{
262 return std::floor (x);
263}
264
265inline float
266fastfloor (float x)
267{
268 return sint32(x) - (x < 0);
269}
270
271inline double
272fastfloor (double x)
273{
274 return sint64(x) - (x < 0);
275}
276
225/* 277/*
226 * absdir(int): Returns a number between 1 and 8, which represent 278 * absdir(int): Returns a number between 1 and 8, which represent
227 * the "absolute" direction of a number (it actually takes care of 279 * the "absolute" direction of a number (it actually takes care of
228 * "overflow" in previous calculations of a direction). 280 * "overflow" in previous calculations of a direction).
229 */ 281 */
… …
231absdir (int d) 283absdir (int d)
232{ 284{
233 return ((d - 1) & 7) + 1; 285 return ((d - 1) & 7) + 1;
234} 286}
235 287
288// avoid ctz name because netbsd or freebsd spams it's namespace with it
289#if GCC_VERSION(3,4)
290static inline int least_significant_bit (uint32_t x)
291{
292 return __builtin_ctz (x);
293}
294#else
295int least_significant_bit (uint32_t x);
296#endif
297
298#define for_all_bits_sparse_32(mask, idxvar) \
299 for (uint32_t idxvar, mask_ = mask; \
300 mask_ && ((idxvar = least_significant_bit (mask_)), mask_ &= ~(1 << idxvar), 1);)
301
236extern ssize_t slice_alloc; // statistics 302extern ssize_t slice_alloc; // statistics
237 303
238void *salloc_ (int n) throw (std::bad_alloc); 304void *salloc_ (int n) throw (std::bad_alloc);
239void *salloc_ (int n, void *src) throw (std::bad_alloc); 305void *salloc_ (int n, void *src) throw (std::bad_alloc);
240 306
… …
384 { 450 {
385 p->~Tp (); 451 p->~Tp ();
386 } 452 }
387}; 453};
388 454
389// P. L'Ecuyer, “Maximally Equidistributed Combined Tausworthe Generators”, Mathematics of Computation, 65, 213 (1996), 203–213.
390// http://www.iro.umontreal.ca/~lecuyer/myftp/papers/tausme.ps
391// http://www.iro.umontreal.ca/~lecuyer/myftp/papers/tausme2.ps
392struct tausworthe_random_generator
393{
394 uint32_t state [4];
395
396 void operator =(const tausworthe_random_generator &src)
397 {
398 state [0] = src.state [0];
399 state [1] = src.state [1];
400 state [2] = src.state [2];
401 state [3] = src.state [3];
402 }
403
404 void seed (uint32_t seed);
405 uint32_t next ();
406};
407
408// Xorshift RNGs, George Marsaglia
409// http://www.jstatsoft.org/v08/i14/paper
410// this one is about 40% faster than the tausworthe one above (i.e. not much),
411// despite the inlining, and has the issue of only creating 2**32-1 numbers.
412struct xorshift_random_generator
413{
414 uint32_t x, y;
415
416 void operator =(const xorshift_random_generator &src)
417 {
418 x = src.x;
419 y = src.y;
420 }
421
422 void seed (uint32_t seed)
423 {
424 x = seed;
425 y = seed * 69069U;
426 }
427
428 uint32_t next ()
429 {
430 uint32_t t = x ^ (x << 10);
431 x = y;
432 y = y ^ (y >> 13) ^ t ^ (t >> 10);
433 return y;
434 }
435};
436
437template<class generator>
438struct random_number_generator : generator
439{
440 // uniform distribution, 0 .. max (0, num - 1)
441 uint32_t operator ()(uint32_t num)
442 {
443 return !is_constant (num) ? get_range (num) // non-constant
444 : num & (num - 1) ? (this->next () * (uint64_t)num) >> 32U // constant, non-power-of-two
445 : this->next () & (num - 1); // constant, power-of-two
446 }
447
448 // return a number within (min .. max)
449 int operator () (int r_min, int r_max)
450 {
451 return is_constant (r_min) && is_constant (r_max) && r_min <= r_max
452 ? r_min + operator ()(r_max - r_min + 1)
453 : get_range (r_min, r_max);
454 }
455
456 double operator ()()
457 {
458 return this->next () / (double)0xFFFFFFFFU;
459 }
460
461protected:
462 uint32_t get_range (uint32_t r_max);
463 int get_range (int r_min, int r_max);
464};
465
466typedef random_number_generator<tausworthe_random_generator> rand_gen;
467
468extern rand_gen rndm, rmg_rndm;
469
470INTERFACE_CLASS (attachable) 455INTERFACE_CLASS (attachable)
471struct refcnt_base 456struct refcnt_base
472{ 457{
473 typedef int refcnt_t; 458 typedef int refcnt_t;
474 mutable refcnt_t ACC (RW, refcnt); 459 mutable refcnt_t ACC (RW, refcnt);
… …
535typedef refptr<maptile> maptile_ptr; 520typedef refptr<maptile> maptile_ptr;
536typedef refptr<object> object_ptr; 521typedef refptr<object> object_ptr;
537typedef refptr<archetype> arch_ptr; 522typedef refptr<archetype> arch_ptr;
538typedef refptr<client> client_ptr; 523typedef refptr<client> client_ptr;
539typedef refptr<player> player_ptr; 524typedef refptr<player> player_ptr;
525typedef refptr<region> region_ptr;
526
527#define STRHSH_NULL 2166136261
528
529static inline uint32_t
530strhsh (const char *s)
531{
532 // use FNV-1a hash (http://isthe.com/chongo/tech/comp/fnv/)
533 // it is about twice as fast as the one-at-a-time one,
534 // with good distribution.
535 // FNV-1a is faster on many cpus because the multiplication
536 // runs concurrently with the looping logic.
537 // we modify the hash a bit to improve its distribution
538 uint32_t hash = STRHSH_NULL;
539
540 while (*s)
541 hash = (hash ^ *s++) * 16777619U;
542
543 return hash ^ (hash >> 16);
544}
545
546static inline uint32_t
547memhsh (const char *s, size_t len)
548{
549 uint32_t hash = STRHSH_NULL;
550
551 while (len--)
552 hash = (hash ^ *s++) * 16777619U;
553
554 return hash;
555}
540 556
541struct str_hash 557struct str_hash
542{ 558{
543 std::size_t operator ()(const char *s) const 559 std::size_t operator ()(const char *s) const
544 { 560 {
545#if 0
546 uint32_t hash = 0;
547
548 /* use the one-at-a-time hash function, which supposedly is
549 * better than the djb2-like one used by perl5.005, but
550 * certainly is better then the bug used here before.
551 * see http://burtleburtle.net/bob/hash/doobs.html
552 */
553 while (*s)
554 {
555 hash += *s++;
556 hash += hash << 10;
557 hash ^= hash >> 6;
558 }
559
560 hash += hash << 3;
561 hash ^= hash >> 11;
562 hash += hash << 15;
563#else
564 // use FNV-1a hash (http://isthe.com/chongo/tech/comp/fnv/)
565 // it is about twice as fast as the one-at-a-time one,
566 // with good distribution.
567 // FNV-1a is faster on many cpus because the multiplication
568 // runs concurrent with the looping logic.
569 uint32_t hash = 2166136261;
570
571 while (*s)
572 hash = (hash ^ *s++) * 16777619;
573#endif
574
575 return hash; 561 return strhsh (s);
562 }
563
564 std::size_t operator ()(const shstr &s) const
565 {
566 return strhsh (s);
576 } 567 }
577}; 568};
578 569
579struct str_equal 570struct str_equal
580{ 571{
… …
672 { 663 {
673 erase (&obj); 664 erase (&obj);
674 } 665 }
675}; 666};
676 667
668/////////////////////////////////////////////////////////////////////////////
669
670// something like a vector or stack, but without
671// out of bounds checking
672template<typename T>
673struct fixed_stack
674{
675 T *data;
676 int size;
677 int max;
678
679 fixed_stack ()
680 : size (0), data (0)
681 {
682 }
683
684 fixed_stack (int max)
685 : size (0), max (max)
686 {
687 data = salloc<T> (max);
688 }
689
690 void reset (int new_max)
691 {
692 sfree (data, max);
693 size = 0;
694 max = new_max;
695 data = salloc<T> (max);
696 }
697
698 void free ()
699 {
700 sfree (data, max);
701 data = 0;
702 }
703
704 ~fixed_stack ()
705 {
706 sfree (data, max);
707 }
708
709 T &operator[](int idx)
710 {
711 return data [idx];
712 }
713
714 void push (T v)
715 {
716 data [size++] = v;
717 }
718
719 T &pop ()
720 {
721 return data [--size];
722 }
723
724 T remove (int idx)
725 {
726 T v = data [idx];
727
728 data [idx] = data [--size];
729
730 return v;
731 }
732};
733
734/////////////////////////////////////////////////////////////////////////////
735
677// basically does what strncpy should do, but appends "..." to strings exceeding length 736// basically does what strncpy should do, but appends "..." to strings exceeding length
737// returns the number of bytes actually used (including \0)
678void assign (char *dst, const char *src, int maxlen); 738int assign (char *dst, const char *src, int maxsize);
679 739
680// type-safe version of assign 740// type-safe version of assign
681template<int N> 741template<int N>
682inline void assign (char (&dst)[N], const char *src) 742inline int assign (char (&dst)[N], const char *src)
683{ 743{
684 assign ((char *)&dst, src, N); 744 return assign ((char *)&dst, src, N);
685} 745}
686 746
687typedef double tstamp; 747typedef double tstamp;
688 748
689// return current time as timestamp 749// return current time as timestamp
690tstamp now (); 750tstamp now ();
691 751
692int similar_direction (int a, int b); 752int similar_direction (int a, int b);
693 753
694// like sprintf, but returns a "static" buffer 754// like v?sprintf, but returns a "static" buffer
695const char *format (const char *format, ...); 755char *vformat (const char *format, va_list ap);
756char *format (const char *format, ...) attribute ((format (printf, 1, 2)));
757
758// safety-check player input which will become object->msg
759bool msg_is_safe (const char *msg);
696 760
697///////////////////////////////////////////////////////////////////////////// 761/////////////////////////////////////////////////////////////////////////////
698// threads, very very thin wrappers around pthreads 762// threads, very very thin wrappers around pthreads
699 763
700struct thread 764struct thread

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines