|
|
1 | /* |
|
|
2 | * This file is part of Deliantra, the Roguelike Realtime MMORPG. |
|
|
3 | * |
|
|
4 | * Copyright (©) 2005,2006,2007,2008,2009,2010 Marc Alexander Lehmann / Robin Redeker / the Deliantra team |
|
|
5 | * |
|
|
6 | * Deliantra is free software: you can redistribute it and/or modify it under |
|
|
7 | * the terms of the Affero GNU General Public License as published by the |
|
|
8 | * Free Software Foundation, either version 3 of the License, or (at your |
|
|
9 | * option) any later version. |
|
|
10 | * |
|
|
11 | * This program is distributed in the hope that it will be useful, |
|
|
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
|
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
|
|
14 | * GNU General Public License for more details. |
|
|
15 | * |
|
|
16 | * You should have received a copy of the Affero GNU General Public License |
|
|
17 | * and the GNU General Public License along with this program. If not, see |
|
|
18 | * <http://www.gnu.org/licenses/>. |
|
|
19 | * |
|
|
20 | * The authors can be reached via e-mail to <support@deliantra.net> |
|
|
21 | */ |
|
|
22 | |
1 | #ifndef UTIL_H__ |
23 | #ifndef UTIL_H__ |
2 | #define UTIL_H__ |
24 | #define UTIL_H__ |
3 | |
25 | |
4 | //#define PREFER_MALLOC |
26 | #include <compiler.h> |
5 | |
27 | |
6 | #if __GNUC__ >= 3 |
28 | #define DEBUG_POISON 0x00 // poison memory before freeing it if != 0 |
7 | # define is_constant(c) __builtin_constant_p (c) |
29 | #define DEBUG_SALLOC 0 // add a debug wrapper around all sallocs |
8 | #else |
30 | #define PREFER_MALLOC 0 // use malloc and not the slice allocator |
9 | # define is_constant(c) 0 |
31 | |
10 | #endif |
32 | #include <pthread.h> |
11 | |
33 | |
12 | #include <cstddef> |
34 | #include <cstddef> |
13 | #include <cmath> |
35 | #include <cmath> |
14 | #include <new> |
36 | #include <new> |
15 | #include <vector> |
37 | #include <vector> |
… | |
… | |
17 | #include <glib.h> |
39 | #include <glib.h> |
18 | |
40 | |
19 | #include <shstr.h> |
41 | #include <shstr.h> |
20 | #include <traits.h> |
42 | #include <traits.h> |
21 | |
43 | |
|
|
44 | #if DEBUG_SALLOC |
|
|
45 | # define g_slice_alloc0(s) debug_slice_alloc0(s) |
|
|
46 | # define g_slice_alloc(s) debug_slice_alloc(s) |
|
|
47 | # define g_slice_free1(s,p) debug_slice_free1(s,p) |
|
|
48 | void *g_slice_alloc (unsigned long size); |
|
|
49 | void *g_slice_alloc0 (unsigned long size); |
|
|
50 | void g_slice_free1 (unsigned long size, void *ptr); |
|
|
51 | #elif PREFER_MALLOC |
|
|
52 | # define g_slice_alloc0(s) calloc (1, (s)) |
|
|
53 | # define g_slice_alloc(s) malloc ((s)) |
|
|
54 | # define g_slice_free1(s,p) free ((p)) |
|
|
55 | #endif |
|
|
56 | |
22 | // use a gcc extension for auto declarations until ISO C++ sanctifies them |
57 | // use C0X decltype for auto declarations until ISO C++ sanctifies them (if ever) |
23 | #define AUTODECL(var,expr) typeof(expr) var = (expr) |
58 | #define auto(var,expr) decltype(expr) var = (expr) |
24 | |
59 | |
|
|
60 | // could use the sizeof (arr) /( sizeof (arr [0]) here, but C++ is |
|
|
61 | // much more obfuscated... :) |
|
|
62 | |
|
|
63 | template<typename T, int N> |
|
|
64 | inline int array_length (const T (&arr)[N]) |
|
|
65 | { |
|
|
66 | return N; |
|
|
67 | } |
|
|
68 | |
25 | // very ugly macro that basicaly declares and initialises a variable |
69 | // very ugly macro that basically declares and initialises a variable |
26 | // that is in scope for the next statement only |
70 | // that is in scope for the next statement only |
27 | // works only for stuff that can be assigned 0 and converts to false |
71 | // works only for stuff that can be assigned 0 and converts to false |
28 | // (note: works great for pointers) |
72 | // (note: works great for pointers) |
29 | // most ugly macro I ever wrote |
73 | // most ugly macro I ever wrote |
30 | #define declvar(type, name, value) if (type name = 0) { } else if (((name) = (value)), 1) |
74 | #define statementvar(type, name, value) if (type name = 0) { } else if (((name) = (value)), 1) |
31 | |
75 | |
32 | // in range including end |
76 | // in range including end |
33 | #define IN_RANGE_INC(val,beg,end) \ |
77 | #define IN_RANGE_INC(val,beg,end) \ |
34 | ((unsigned int)(val) - (unsigned int)(beg) <= (unsigned int)(end) - (unsigned int)(beg)) |
78 | ((unsigned int)(val) - (unsigned int)(beg) <= (unsigned int)(end) - (unsigned int)(beg)) |
35 | |
79 | |
36 | // in range excluding end |
80 | // in range excluding end |
37 | #define IN_RANGE_EXC(val,beg,end) \ |
81 | #define IN_RANGE_EXC(val,beg,end) \ |
38 | ((unsigned int)(val) - (unsigned int)(beg) < (unsigned int)(end) - (unsigned int)(beg)) |
82 | ((unsigned int)(val) - (unsigned int)(beg) < (unsigned int)(end) - (unsigned int)(beg)) |
39 | |
83 | |
|
|
84 | void cleanup (const char *cause, bool make_core = false); |
40 | void fork_abort (const char *msg); |
85 | void fork_abort (const char *msg); |
41 | |
86 | |
42 | // rationale for using (U) not (T) is to reduce signed/unsigned issues, |
87 | // rationale for using (U) not (T) is to reduce signed/unsigned issues, |
43 | // as a is often a constant while b is the variable. it is still a bug, though. |
88 | // as a is often a constant while b is the variable. it is still a bug, though. |
44 | template<typename T, typename U> static inline T min (T a, U b) { return (U)a < b ? (U)a : b; } |
89 | template<typename T, typename U> static inline T min (T a, U b) { return (U)a < b ? (U)a : b; } |
45 | template<typename T, typename U> static inline T max (T a, U b) { return (U)a > b ? (U)a : b; } |
90 | template<typename T, typename U> static inline T max (T a, U b) { return (U)a > b ? (U)a : b; } |
46 | template<typename T, typename U, typename V> static inline T clamp (T v, U a, V b) { return v < (T)a ? (T)a : v >(T)b ? (T)b : v; } |
91 | template<typename T, typename U, typename V> static inline T clamp (T v, U a, V b) { return v < (T)a ? (T)a : v >(T)b ? (T)b : v; } |
47 | |
92 | |
|
|
93 | template<typename T, typename U> static inline void min_it (T &v, U m) { v = min (v, (T)m); } |
|
|
94 | template<typename T, typename U> static inline void max_it (T &v, U m) { v = max (v, (T)m); } |
|
|
95 | template<typename T, typename U, typename V> static inline void clamp_it (T &v, U a, V b) { v = clamp (v, (T)a, (T)b); } |
|
|
96 | |
48 | template<typename T, typename U> static inline void swap (T& a, U& b) { T t=a; a=(T)b; b=(U)t; } |
97 | template<typename T, typename U> static inline void swap (T& a, U& b) { T t=a; a=(T)b; b=(U)t; } |
49 | |
98 | |
|
|
99 | template<typename T, typename U, typename V> static inline T min (T a, U b, V c) { return min (a, min (b, c)); } |
|
|
100 | template<typename T, typename U, typename V> static inline T max (T a, U b, V c) { return max (a, max (b, c)); } |
|
|
101 | |
|
|
102 | // sign returns -1 or +1 |
|
|
103 | template<typename T> |
|
|
104 | static inline T sign (T v) { return v < 0 ? -1 : +1; } |
|
|
105 | // relies on 2c representation |
|
|
106 | template<> |
|
|
107 | inline sint8 sign (sint8 v) { return 1 - (sint8 (uint8 (v) >> 7) * 2); } |
|
|
108 | template<> |
|
|
109 | inline sint16 sign (sint16 v) { return 1 - (sint16 (uint16 (v) >> 15) * 2); } |
|
|
110 | template<> |
|
|
111 | inline sint32 sign (sint32 v) { return 1 - (sint32 (uint32 (v) >> 31) * 2); } |
|
|
112 | |
|
|
113 | // sign0 returns -1, 0 or +1 |
|
|
114 | template<typename T> |
|
|
115 | static inline T sign0 (T v) { return v ? sign (v) : 0; } |
|
|
116 | |
|
|
117 | template<typename T, typename U> |
|
|
118 | static inline T copysign (T a, U b) { return a > 0 ? b : -b; } |
|
|
119 | |
|
|
120 | // div* only work correctly for div > 0 |
|
|
121 | // div, with correct rounding (< 0.5 downwards, >=0.5 upwards) |
|
|
122 | template<typename T> static inline T div (T val, T div) |
|
|
123 | { |
|
|
124 | return expect_false (val < 0) ? - ((-val + (div - 1) / 2) / div) : (val + div / 2) / div; |
|
|
125 | } |
|
|
126 | // div, round-up |
|
|
127 | template<typename T> static inline T div_ru (T val, T div) |
|
|
128 | { |
|
|
129 | return expect_false (val < 0) ? - ((-val ) / div) : (val + div - 1) / div; |
|
|
130 | } |
|
|
131 | // div, round-down |
|
|
132 | template<typename T> static inline T div_rd (T val, T div) |
|
|
133 | { |
|
|
134 | return expect_false (val < 0) ? - ((-val + (div - 1) ) / div) : (val ) / div; |
|
|
135 | } |
|
|
136 | |
|
|
137 | // lerp* only work correctly for min_in < max_in |
|
|
138 | // Linear intERPolate, scales val from min_in..max_in to min_out..max_out |
|
|
139 | template<typename T> |
|
|
140 | static inline T |
|
|
141 | lerp (T val, T min_in, T max_in, T min_out, T max_out) |
|
|
142 | { |
|
|
143 | return min_out + div <T> ((val - min_in) * (max_out - min_out), max_in - min_in); |
|
|
144 | } |
|
|
145 | |
|
|
146 | // lerp, round-down |
|
|
147 | template<typename T> |
|
|
148 | static inline T |
|
|
149 | lerp_rd (T val, T min_in, T max_in, T min_out, T max_out) |
|
|
150 | { |
|
|
151 | return min_out + div_rd<T> ((val - min_in) * (max_out - min_out), max_in - min_in); |
|
|
152 | } |
|
|
153 | |
|
|
154 | // lerp, round-up |
|
|
155 | template<typename T> |
|
|
156 | static inline T |
|
|
157 | lerp_ru (T val, T min_in, T max_in, T min_out, T max_out) |
|
|
158 | { |
|
|
159 | return min_out + div_ru<T> ((val - min_in) * (max_out - min_out), max_in - min_in); |
|
|
160 | } |
|
|
161 | |
|
|
162 | // lots of stuff taken from FXT |
|
|
163 | |
|
|
164 | /* Rotate right. This is used in various places for checksumming */ |
|
|
165 | //TODO: that sucks, use a better checksum algo |
|
|
166 | static inline uint32_t |
|
|
167 | rotate_right (uint32_t c, uint32_t count = 1) |
|
|
168 | { |
|
|
169 | return (c << (32 - count)) | (c >> count); |
|
|
170 | } |
|
|
171 | |
|
|
172 | static inline uint32_t |
|
|
173 | rotate_left (uint32_t c, uint32_t count = 1) |
|
|
174 | { |
|
|
175 | return (c >> (32 - count)) | (c << count); |
|
|
176 | } |
|
|
177 | |
|
|
178 | // Return abs(a-b) |
|
|
179 | // Both a and b must not have the most significant bit set |
|
|
180 | static inline uint32_t |
|
|
181 | upos_abs_diff (uint32_t a, uint32_t b) |
|
|
182 | { |
|
|
183 | long d1 = b - a; |
|
|
184 | long d2 = (d1 & (d1 >> 31)) << 1; |
|
|
185 | |
|
|
186 | return d1 - d2; // == (b - d) - (a + d); |
|
|
187 | } |
|
|
188 | |
|
|
189 | // Both a and b must not have the most significant bit set |
|
|
190 | static inline uint32_t |
|
|
191 | upos_min (uint32_t a, uint32_t b) |
|
|
192 | { |
|
|
193 | int32_t d = b - a; |
|
|
194 | d &= d >> 31; |
|
|
195 | return a + d; |
|
|
196 | } |
|
|
197 | |
|
|
198 | // Both a and b must not have the most significant bit set |
|
|
199 | static inline uint32_t |
|
|
200 | upos_max (uint32_t a, uint32_t b) |
|
|
201 | { |
|
|
202 | int32_t d = b - a; |
|
|
203 | d &= d >> 31; |
|
|
204 | return b - d; |
|
|
205 | } |
|
|
206 | |
50 | // this is much faster than crossfires original algorithm |
207 | // this is much faster than crossfire's original algorithm |
51 | // on modern cpus |
208 | // on modern cpus |
52 | inline int |
209 | inline int |
53 | isqrt (int n) |
210 | isqrt (int n) |
54 | { |
211 | { |
55 | return (int)sqrtf ((float)n); |
212 | return (int)sqrtf ((float)n); |
|
|
213 | } |
|
|
214 | |
|
|
215 | // this is kind of like the ^^ operator, if it would exist, without sequence point. |
|
|
216 | // more handy than it looks like, due to the implicit !! done on its arguments |
|
|
217 | inline bool |
|
|
218 | logical_xor (bool a, bool b) |
|
|
219 | { |
|
|
220 | return a != b; |
|
|
221 | } |
|
|
222 | |
|
|
223 | inline bool |
|
|
224 | logical_implies (bool a, bool b) |
|
|
225 | { |
|
|
226 | return a <= b; |
56 | } |
227 | } |
57 | |
228 | |
58 | // this is only twice as fast as naive sqrtf (dx*dy+dy*dy) |
229 | // this is only twice as fast as naive sqrtf (dx*dy+dy*dy) |
59 | #if 0 |
230 | #if 0 |
60 | // and has a max. error of 6 in the range -100..+100. |
231 | // and has a max. error of 6 in the range -100..+100. |
… | |
… | |
85 | absdir (int d) |
256 | absdir (int d) |
86 | { |
257 | { |
87 | return ((d - 1) & 7) + 1; |
258 | return ((d - 1) & 7) + 1; |
88 | } |
259 | } |
89 | |
260 | |
|
|
261 | // avoid ctz name because netbsd or freebsd spams it's namespace with it |
|
|
262 | #if GCC_VERSION(3,4) |
|
|
263 | static inline int least_significant_bit (uint32_t x) |
|
|
264 | { |
|
|
265 | return __builtin_ctz (x); |
|
|
266 | } |
|
|
267 | #else |
|
|
268 | int least_significant_bit (uint32_t x); |
|
|
269 | #endif |
|
|
270 | |
|
|
271 | #define for_all_bits_sparse_32(mask, idxvar) \ |
|
|
272 | for (uint32_t idxvar, mask_ = mask; \ |
|
|
273 | mask_ && ((idxvar = least_significant_bit (mask_)), mask_ &= ~(1 << idxvar), 1);) |
|
|
274 | |
|
|
275 | extern ssize_t slice_alloc; // statistics |
|
|
276 | |
|
|
277 | void *salloc_ (int n) throw (std::bad_alloc); |
|
|
278 | void *salloc_ (int n, void *src) throw (std::bad_alloc); |
|
|
279 | |
|
|
280 | // strictly the same as g_slice_alloc, but never returns 0 |
|
|
281 | template<typename T> |
|
|
282 | inline T *salloc (int n = 1) throw (std::bad_alloc) { return (T *)salloc_ (n * sizeof (T)); } |
|
|
283 | |
|
|
284 | // also copies src into the new area, like "memdup" |
|
|
285 | // if src is 0, clears the memory |
|
|
286 | template<typename T> |
|
|
287 | inline T *salloc (int n, T *src) throw (std::bad_alloc) { return (T *)salloc_ (n * sizeof (T), (void *)src); } |
|
|
288 | |
|
|
289 | // clears the memory |
|
|
290 | template<typename T> |
|
|
291 | inline T *salloc0(int n = 1) throw (std::bad_alloc) { return (T *)salloc_ (n * sizeof (T), 0); } |
|
|
292 | |
|
|
293 | // for symmetry |
|
|
294 | template<typename T> |
|
|
295 | inline void sfree (T *ptr, int n = 1) throw () |
|
|
296 | { |
|
|
297 | if (expect_true (ptr)) |
|
|
298 | { |
|
|
299 | slice_alloc -= n * sizeof (T); |
|
|
300 | if (DEBUG_POISON) memset (ptr, DEBUG_POISON, n * sizeof (T)); |
|
|
301 | g_slice_free1 (n * sizeof (T), (void *)ptr); |
|
|
302 | assert (slice_alloc >= 0);//D |
|
|
303 | } |
|
|
304 | } |
|
|
305 | |
|
|
306 | // nulls the pointer |
|
|
307 | template<typename T> |
|
|
308 | inline void sfree0 (T *&ptr, int n = 1) throw () |
|
|
309 | { |
|
|
310 | sfree<T> (ptr, n); |
|
|
311 | ptr = 0; |
|
|
312 | } |
|
|
313 | |
90 | // makes dynamically allocated objects zero-initialised |
314 | // makes dynamically allocated objects zero-initialised |
91 | struct zero_initialised |
315 | struct zero_initialised |
92 | { |
316 | { |
93 | void *operator new (size_t s, void *p) |
317 | void *operator new (size_t s, void *p) |
94 | { |
318 | { |
… | |
… | |
96 | return p; |
320 | return p; |
97 | } |
321 | } |
98 | |
322 | |
99 | void *operator new (size_t s) |
323 | void *operator new (size_t s) |
100 | { |
324 | { |
101 | return g_slice_alloc0 (s); |
325 | return salloc0<char> (s); |
102 | } |
326 | } |
103 | |
327 | |
104 | void *operator new[] (size_t s) |
328 | void *operator new[] (size_t s) |
105 | { |
329 | { |
106 | return g_slice_alloc0 (s); |
330 | return salloc0<char> (s); |
107 | } |
331 | } |
108 | |
332 | |
109 | void operator delete (void *p, size_t s) |
333 | void operator delete (void *p, size_t s) |
110 | { |
334 | { |
111 | g_slice_free1 (s, p); |
335 | sfree ((char *)p, s); |
112 | } |
336 | } |
113 | |
337 | |
114 | void operator delete[] (void *p, size_t s) |
338 | void operator delete[] (void *p, size_t s) |
115 | { |
339 | { |
116 | g_slice_free1 (s, p); |
340 | sfree ((char *)p, s); |
117 | } |
341 | } |
118 | }; |
342 | }; |
119 | |
343 | |
120 | void *salloc_ (int n) throw (std::bad_alloc); |
344 | // makes dynamically allocated objects zero-initialised |
121 | void *salloc_ (int n, void *src) throw (std::bad_alloc); |
345 | struct slice_allocated |
122 | |
|
|
123 | // strictly the same as g_slice_alloc, but never returns 0 |
|
|
124 | template<typename T> |
|
|
125 | inline T *salloc (int n = 1) throw (std::bad_alloc) { return (T *)salloc_ (n * sizeof (T)); } |
|
|
126 | |
|
|
127 | // also copies src into the new area, like "memdup" |
|
|
128 | // if src is 0, clears the memory |
|
|
129 | template<typename T> |
|
|
130 | inline T *salloc (int n, T *src) throw (std::bad_alloc) { return (T *)salloc_ (n * sizeof (T), (void *)src); } |
|
|
131 | |
|
|
132 | // clears the memory |
|
|
133 | template<typename T> |
|
|
134 | inline T *salloc0(int n = 1) throw (std::bad_alloc) { return (T *)salloc_ (n * sizeof (T), 0); } |
|
|
135 | |
|
|
136 | // for symmetry |
|
|
137 | template<typename T> |
|
|
138 | inline void sfree (T *ptr, int n = 1) throw () |
|
|
139 | { |
346 | { |
140 | #ifdef PREFER_MALLOC |
347 | void *operator new (size_t s, void *p) |
141 | free (ptr); |
348 | { |
142 | #else |
349 | return p; |
143 | g_slice_free1 (n * sizeof (T), (void *)ptr); |
350 | } |
144 | #endif |
351 | |
145 | } |
352 | void *operator new (size_t s) |
|
|
353 | { |
|
|
354 | return salloc<char> (s); |
|
|
355 | } |
|
|
356 | |
|
|
357 | void *operator new[] (size_t s) |
|
|
358 | { |
|
|
359 | return salloc<char> (s); |
|
|
360 | } |
|
|
361 | |
|
|
362 | void operator delete (void *p, size_t s) |
|
|
363 | { |
|
|
364 | sfree ((char *)p, s); |
|
|
365 | } |
|
|
366 | |
|
|
367 | void operator delete[] (void *p, size_t s) |
|
|
368 | { |
|
|
369 | sfree ((char *)p, s); |
|
|
370 | } |
|
|
371 | }; |
146 | |
372 | |
147 | // a STL-compatible allocator that uses g_slice |
373 | // a STL-compatible allocator that uses g_slice |
148 | // boy, this is verbose |
374 | // boy, this is verbose |
149 | template<typename Tp> |
375 | template<typename Tp> |
150 | struct slice_allocator |
376 | struct slice_allocator |
… | |
… | |
162 | { |
388 | { |
163 | typedef slice_allocator<U> other; |
389 | typedef slice_allocator<U> other; |
164 | }; |
390 | }; |
165 | |
391 | |
166 | slice_allocator () throw () { } |
392 | slice_allocator () throw () { } |
167 | slice_allocator (const slice_allocator &o) throw () { } |
393 | slice_allocator (const slice_allocator &) throw () { } |
168 | template<typename Tp2> |
394 | template<typename Tp2> |
169 | slice_allocator (const slice_allocator<Tp2> &) throw () { } |
395 | slice_allocator (const slice_allocator<Tp2> &) throw () { } |
170 | |
396 | |
171 | ~slice_allocator () { } |
397 | ~slice_allocator () { } |
172 | |
398 | |
… | |
… | |
181 | void deallocate (pointer p, size_type n) |
407 | void deallocate (pointer p, size_type n) |
182 | { |
408 | { |
183 | sfree<Tp> (p, n); |
409 | sfree<Tp> (p, n); |
184 | } |
410 | } |
185 | |
411 | |
186 | size_type max_size ()const throw () |
412 | size_type max_size () const throw () |
187 | { |
413 | { |
188 | return size_t (-1) / sizeof (Tp); |
414 | return size_t (-1) / sizeof (Tp); |
189 | } |
415 | } |
190 | |
416 | |
191 | void construct (pointer p, const Tp &val) |
417 | void construct (pointer p, const Tp &val) |
… | |
… | |
202 | // P. L'Ecuyer, “Maximally Equidistributed Combined Tausworthe Generators”, Mathematics of Computation, 65, 213 (1996), 203–213. |
428 | // P. L'Ecuyer, “Maximally Equidistributed Combined Tausworthe Generators”, Mathematics of Computation, 65, 213 (1996), 203–213. |
203 | // http://www.iro.umontreal.ca/~lecuyer/myftp/papers/tausme.ps |
429 | // http://www.iro.umontreal.ca/~lecuyer/myftp/papers/tausme.ps |
204 | // http://www.iro.umontreal.ca/~lecuyer/myftp/papers/tausme2.ps |
430 | // http://www.iro.umontreal.ca/~lecuyer/myftp/papers/tausme2.ps |
205 | struct tausworthe_random_generator |
431 | struct tausworthe_random_generator |
206 | { |
432 | { |
207 | // generator |
|
|
208 | uint32_t state [4]; |
433 | uint32_t state [4]; |
209 | |
434 | |
210 | void operator =(const tausworthe_random_generator &src) |
435 | void operator =(const tausworthe_random_generator &src) |
211 | { |
436 | { |
212 | state [0] = src.state [0]; |
437 | state [0] = src.state [0]; |
… | |
… | |
215 | state [3] = src.state [3]; |
440 | state [3] = src.state [3]; |
216 | } |
441 | } |
217 | |
442 | |
218 | void seed (uint32_t seed); |
443 | void seed (uint32_t seed); |
219 | uint32_t next (); |
444 | uint32_t next (); |
|
|
445 | }; |
220 | |
446 | |
221 | // uniform distribution |
447 | // Xorshift RNGs, George Marsaglia |
|
|
448 | // http://www.jstatsoft.org/v08/i14/paper |
|
|
449 | // this one is about 40% faster than the tausworthe one above (i.e. not much), |
|
|
450 | // despite the inlining, and has the issue of only creating 2**32-1 numbers. |
|
|
451 | // see also http://www.iro.umontreal.ca/~lecuyer/myftp/papers/xorshift.pdf |
|
|
452 | struct xorshift_random_generator |
|
|
453 | { |
|
|
454 | uint32_t x, y; |
|
|
455 | |
|
|
456 | void operator =(const xorshift_random_generator &src) |
|
|
457 | { |
|
|
458 | x = src.x; |
|
|
459 | y = src.y; |
|
|
460 | } |
|
|
461 | |
|
|
462 | void seed (uint32_t seed) |
|
|
463 | { |
|
|
464 | x = seed; |
|
|
465 | y = seed * 69069U; |
|
|
466 | } |
|
|
467 | |
|
|
468 | uint32_t next () |
|
|
469 | { |
|
|
470 | uint32_t t = x ^ (x << 10); |
|
|
471 | x = y; |
|
|
472 | y = y ^ (y >> 13) ^ t ^ (t >> 10); |
|
|
473 | return y; |
|
|
474 | } |
|
|
475 | }; |
|
|
476 | |
|
|
477 | template<class generator> |
|
|
478 | struct random_number_generator : generator |
|
|
479 | { |
|
|
480 | // uniform distribution, 0 .. max (0, num - 1) |
222 | uint32_t operator ()(uint32_t r_max) |
481 | uint32_t operator ()(uint32_t num) |
223 | { |
482 | { |
224 | return is_constant (r_max) |
483 | return !is_constant (num) ? get_range (num) // non-constant |
225 | ? this->next () % r_max |
484 | : num & (num - 1) ? (this->next () * (uint64_t)num) >> 32U // constant, non-power-of-two |
226 | : get_range (r_max); |
485 | : this->next () & (num - 1); // constant, power-of-two |
227 | } |
486 | } |
228 | |
487 | |
229 | // return a number within (min .. max) |
488 | // return a number within the closed interval [min .. max] |
230 | int operator () (int r_min, int r_max) |
489 | int operator () (int r_min, int r_max) |
231 | { |
490 | { |
232 | return is_constant (r_min) && is_constant (r_max) |
491 | return is_constant (r_min) && is_constant (r_max) && r_min <= r_max |
233 | ? r_min + (*this) (max (r_max - r_min + 1, 1)) |
492 | ? r_min + operator ()(r_max - r_min + 1) |
234 | : get_range (r_min, r_max); |
493 | : get_range (r_min, r_max); |
235 | } |
494 | } |
236 | |
495 | |
|
|
496 | // return a number within the closed interval [0..1] |
237 | double operator ()() |
497 | double operator ()() |
238 | { |
498 | { |
239 | return this->next () / (double)0xFFFFFFFFU; |
499 | return this->next () / (double)0xFFFFFFFFU; |
240 | } |
500 | } |
241 | |
501 | |
242 | protected: |
502 | protected: |
243 | uint32_t get_range (uint32_t r_max); |
503 | uint32_t get_range (uint32_t r_max); |
244 | int get_range (int r_min, int r_max); |
504 | int get_range (int r_min, int r_max); |
245 | }; |
505 | }; |
246 | |
506 | |
247 | typedef tausworthe_random_generator rand_gen; |
507 | typedef random_number_generator<tausworthe_random_generator> rand_gen; |
248 | |
508 | |
249 | extern rand_gen rndm; |
509 | extern rand_gen rndm, rmg_rndm; |
|
|
510 | |
|
|
511 | INTERFACE_CLASS (attachable) |
|
|
512 | struct refcnt_base |
|
|
513 | { |
|
|
514 | typedef int refcnt_t; |
|
|
515 | mutable refcnt_t ACC (RW, refcnt); |
|
|
516 | |
|
|
517 | MTH void refcnt_inc () const { ++refcnt; } |
|
|
518 | MTH void refcnt_dec () const { --refcnt; } |
|
|
519 | |
|
|
520 | refcnt_base () : refcnt (0) { } |
|
|
521 | }; |
|
|
522 | |
|
|
523 | // to avoid branches with more advanced compilers |
|
|
524 | extern refcnt_base::refcnt_t refcnt_dummy; |
250 | |
525 | |
251 | template<class T> |
526 | template<class T> |
252 | struct refptr |
527 | struct refptr |
253 | { |
528 | { |
|
|
529 | // p if not null |
|
|
530 | refcnt_base::refcnt_t *refcnt_ref () { return p ? &p->refcnt : &refcnt_dummy; } |
|
|
531 | |
|
|
532 | void refcnt_dec () |
|
|
533 | { |
|
|
534 | if (!is_constant (p)) |
|
|
535 | --*refcnt_ref (); |
|
|
536 | else if (p) |
|
|
537 | --p->refcnt; |
|
|
538 | } |
|
|
539 | |
|
|
540 | void refcnt_inc () |
|
|
541 | { |
|
|
542 | if (!is_constant (p)) |
|
|
543 | ++*refcnt_ref (); |
|
|
544 | else if (p) |
|
|
545 | ++p->refcnt; |
|
|
546 | } |
|
|
547 | |
254 | T *p; |
548 | T *p; |
255 | |
549 | |
256 | refptr () : p(0) { } |
550 | refptr () : p(0) { } |
257 | refptr (const refptr<T> &p) : p(p.p) { if (p) p->refcnt_inc (); } |
551 | refptr (const refptr<T> &p) : p(p.p) { refcnt_inc (); } |
258 | refptr (T *p) : p(p) { if (p) p->refcnt_inc (); } |
552 | refptr (T *p) : p(p) { refcnt_inc (); } |
259 | ~refptr () { if (p) p->refcnt_dec (); } |
553 | ~refptr () { refcnt_dec (); } |
260 | |
554 | |
261 | const refptr<T> &operator =(T *o) |
555 | const refptr<T> &operator =(T *o) |
262 | { |
556 | { |
|
|
557 | // if decrementing ever destroys we need to reverse the order here |
263 | if (p) p->refcnt_dec (); |
558 | refcnt_dec (); |
264 | p = o; |
559 | p = o; |
265 | if (p) p->refcnt_inc (); |
560 | refcnt_inc (); |
266 | |
|
|
267 | return *this; |
561 | return *this; |
268 | } |
562 | } |
269 | |
563 | |
270 | const refptr<T> &operator =(const refptr<T> o) |
564 | const refptr<T> &operator =(const refptr<T> &o) |
271 | { |
565 | { |
272 | *this = o.p; |
566 | *this = o.p; |
273 | return *this; |
567 | return *this; |
274 | } |
568 | } |
275 | |
569 | |
276 | T &operator * () const { return *p; } |
570 | T &operator * () const { return *p; } |
277 | T *operator ->() const { return p; } |
571 | T *operator ->() const { return p; } |
278 | |
572 | |
279 | operator T *() const { return p; } |
573 | operator T *() const { return p; } |
280 | }; |
574 | }; |
281 | |
575 | |
282 | typedef refptr<maptile> maptile_ptr; |
576 | typedef refptr<maptile> maptile_ptr; |
283 | typedef refptr<object> object_ptr; |
577 | typedef refptr<object> object_ptr; |
284 | typedef refptr<archetype> arch_ptr; |
578 | typedef refptr<archetype> arch_ptr; |
285 | typedef refptr<client> client_ptr; |
579 | typedef refptr<client> client_ptr; |
286 | typedef refptr<player> player_ptr; |
580 | typedef refptr<player> player_ptr; |
|
|
581 | typedef refptr<region> region_ptr; |
|
|
582 | |
|
|
583 | #define STRHSH_NULL 2166136261 |
|
|
584 | |
|
|
585 | static inline uint32_t |
|
|
586 | strhsh (const char *s) |
|
|
587 | { |
|
|
588 | // use FNV-1a hash (http://isthe.com/chongo/tech/comp/fnv/) |
|
|
589 | // it is about twice as fast as the one-at-a-time one, |
|
|
590 | // with good distribution. |
|
|
591 | // FNV-1a is faster on many cpus because the multiplication |
|
|
592 | // runs concurrently with the looping logic. |
|
|
593 | uint32_t hash = STRHSH_NULL; |
|
|
594 | |
|
|
595 | while (*s) |
|
|
596 | hash = (hash ^ *s++) * 16777619U; |
|
|
597 | |
|
|
598 | return hash; |
|
|
599 | } |
|
|
600 | |
|
|
601 | static inline uint32_t |
|
|
602 | memhsh (const char *s, size_t len) |
|
|
603 | { |
|
|
604 | uint32_t hash = STRHSH_NULL; |
|
|
605 | |
|
|
606 | while (len--) |
|
|
607 | hash = (hash ^ *s++) * 16777619U; |
|
|
608 | |
|
|
609 | return hash; |
|
|
610 | } |
287 | |
611 | |
288 | struct str_hash |
612 | struct str_hash |
289 | { |
613 | { |
290 | std::size_t operator ()(const char *s) const |
614 | std::size_t operator ()(const char *s) const |
291 | { |
615 | { |
292 | unsigned long hash = 0; |
|
|
293 | |
|
|
294 | /* use the one-at-a-time hash function, which supposedly is |
|
|
295 | * better than the djb2-like one used by perl5.005, but |
|
|
296 | * certainly is better then the bug used here before. |
|
|
297 | * see http://burtleburtle.net/bob/hash/doobs.html |
|
|
298 | */ |
|
|
299 | while (*s) |
|
|
300 | { |
|
|
301 | hash += *s++; |
|
|
302 | hash += hash << 10; |
|
|
303 | hash ^= hash >> 6; |
|
|
304 | } |
|
|
305 | |
|
|
306 | hash += hash << 3; |
|
|
307 | hash ^= hash >> 11; |
|
|
308 | hash += hash << 15; |
|
|
309 | |
|
|
310 | return hash; |
616 | return strhsh (s); |
|
|
617 | } |
|
|
618 | |
|
|
619 | std::size_t operator ()(const shstr &s) const |
|
|
620 | { |
|
|
621 | return strhsh (s); |
311 | } |
622 | } |
312 | }; |
623 | }; |
313 | |
624 | |
314 | struct str_equal |
625 | struct str_equal |
315 | { |
626 | { |
… | |
… | |
317 | { |
628 | { |
318 | return !strcmp (a, b); |
629 | return !strcmp (a, b); |
319 | } |
630 | } |
320 | }; |
631 | }; |
321 | |
632 | |
|
|
633 | // Mostly the same as std::vector, but insert/erase can reorder |
|
|
634 | // the elements, making append(=insert)/remove O(1) instead of O(n). |
|
|
635 | // |
|
|
636 | // NOTE: only some forms of erase are available |
322 | template<class T> |
637 | template<class T> |
323 | struct unordered_vector : std::vector<T, slice_allocator<T> > |
638 | struct unordered_vector : std::vector<T, slice_allocator<T> > |
324 | { |
639 | { |
325 | typedef typename unordered_vector::iterator iterator; |
640 | typedef typename unordered_vector::iterator iterator; |
326 | |
641 | |
… | |
… | |
336 | { |
651 | { |
337 | erase ((unsigned int )(i - this->begin ())); |
652 | erase ((unsigned int )(i - this->begin ())); |
338 | } |
653 | } |
339 | }; |
654 | }; |
340 | |
655 | |
341 | template<class T, int T::* index> |
656 | // This container blends advantages of linked lists |
|
|
657 | // (efficiency) with vectors (random access) by |
|
|
658 | // by using an unordered vector and storing the vector |
|
|
659 | // index inside the object. |
|
|
660 | // |
|
|
661 | // + memory-efficient on most 64 bit archs |
|
|
662 | // + O(1) insert/remove |
|
|
663 | // + free unique (but varying) id for inserted objects |
|
|
664 | // + cache-friendly iteration |
|
|
665 | // - only works for pointers to structs |
|
|
666 | // |
|
|
667 | // NOTE: only some forms of erase/insert are available |
|
|
668 | typedef int object_vector_index; |
|
|
669 | |
|
|
670 | template<class T, object_vector_index T::*indexmember> |
342 | struct object_vector : std::vector<T *, slice_allocator<T *> > |
671 | struct object_vector : std::vector<T *, slice_allocator<T *> > |
343 | { |
672 | { |
|
|
673 | typedef typename object_vector::iterator iterator; |
|
|
674 | |
|
|
675 | bool contains (const T *obj) const |
|
|
676 | { |
|
|
677 | return obj->*indexmember; |
|
|
678 | } |
|
|
679 | |
|
|
680 | iterator find (const T *obj) |
|
|
681 | { |
|
|
682 | return obj->*indexmember |
|
|
683 | ? this->begin () + obj->*indexmember - 1 |
|
|
684 | : this->end (); |
|
|
685 | } |
|
|
686 | |
|
|
687 | void push_back (T *obj) |
|
|
688 | { |
|
|
689 | std::vector<T *, slice_allocator<T *> >::push_back (obj); |
|
|
690 | obj->*indexmember = this->size (); |
|
|
691 | } |
|
|
692 | |
344 | void insert (T *obj) |
693 | void insert (T *obj) |
345 | { |
694 | { |
346 | assert (!(obj->*index)); |
|
|
347 | push_back (obj); |
695 | push_back (obj); |
348 | obj->*index = this->size (); |
|
|
349 | } |
696 | } |
350 | |
697 | |
351 | void insert (T &obj) |
698 | void insert (T &obj) |
352 | { |
699 | { |
353 | insert (&obj); |
700 | insert (&obj); |
354 | } |
701 | } |
355 | |
702 | |
356 | void erase (T *obj) |
703 | void erase (T *obj) |
357 | { |
704 | { |
358 | assert (obj->*index); |
|
|
359 | int pos = obj->*index; |
705 | unsigned int pos = obj->*indexmember; |
360 | obj->*index = 0; |
706 | obj->*indexmember = 0; |
361 | |
707 | |
362 | if (pos < this->size ()) |
708 | if (pos < this->size ()) |
363 | { |
709 | { |
364 | (*this)[pos - 1] = (*this)[this->size () - 1]; |
710 | (*this)[pos - 1] = (*this)[this->size () - 1]; |
365 | (*this)[pos - 1]->*index = pos; |
711 | (*this)[pos - 1]->*indexmember = pos; |
366 | } |
712 | } |
367 | |
713 | |
368 | this->pop_back (); |
714 | this->pop_back (); |
369 | } |
715 | } |
370 | |
716 | |
371 | void erase (T &obj) |
717 | void erase (T &obj) |
372 | { |
718 | { |
373 | errase (&obj); |
719 | erase (&obj); |
374 | } |
720 | } |
375 | }; |
721 | }; |
376 | |
722 | |
377 | // basically does what strncpy should do, but appends "..." to strings exceeding length |
723 | // basically does what strncpy should do, but appends "..." to strings exceeding length |
|
|
724 | // returns the number of bytes actually used (including \0) |
378 | void assign (char *dst, const char *src, int maxlen); |
725 | int assign (char *dst, const char *src, int maxsize); |
379 | |
726 | |
380 | // type-safe version of assign |
727 | // type-safe version of assign |
381 | template<int N> |
728 | template<int N> |
382 | inline void assign (char (&dst)[N], const char *src) |
729 | inline int assign (char (&dst)[N], const char *src) |
383 | { |
730 | { |
384 | assign ((char *)&dst, src, N); |
731 | return assign ((char *)&dst, src, N); |
385 | } |
732 | } |
386 | |
733 | |
387 | typedef double tstamp; |
734 | typedef double tstamp; |
388 | |
735 | |
389 | // return current time as timestampe |
736 | // return current time as timestamp |
390 | tstamp now (); |
737 | tstamp now (); |
391 | |
738 | |
392 | int similar_direction (int a, int b); |
739 | int similar_direction (int a, int b); |
393 | |
740 | |
|
|
741 | // like v?sprintf, but returns a "static" buffer |
|
|
742 | char *vformat (const char *format, va_list ap); |
|
|
743 | char *format (const char *format, ...) attribute ((format (printf, 1, 2))); |
|
|
744 | |
|
|
745 | // safety-check player input which will become object->msg |
|
|
746 | bool msg_is_safe (const char *msg); |
|
|
747 | |
|
|
748 | ///////////////////////////////////////////////////////////////////////////// |
|
|
749 | // threads, very very thin wrappers around pthreads |
|
|
750 | |
|
|
751 | struct thread |
|
|
752 | { |
|
|
753 | pthread_t id; |
|
|
754 | |
|
|
755 | void start (void *(*start_routine)(void *), void *arg = 0); |
|
|
756 | |
|
|
757 | void cancel () |
|
|
758 | { |
|
|
759 | pthread_cancel (id); |
|
|
760 | } |
|
|
761 | |
|
|
762 | void *join () |
|
|
763 | { |
|
|
764 | void *ret; |
|
|
765 | |
|
|
766 | if (pthread_join (id, &ret)) |
|
|
767 | cleanup ("pthread_join failed", 1); |
|
|
768 | |
|
|
769 | return ret; |
|
|
770 | } |
|
|
771 | }; |
|
|
772 | |
|
|
773 | // note that mutexes are not classes |
|
|
774 | typedef pthread_mutex_t smutex; |
|
|
775 | |
|
|
776 | #if __linux && defined (PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP) |
|
|
777 | #define SMUTEX_INITIALISER PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP |
|
|
778 | #else |
|
|
779 | #define SMUTEX_INITIALISER PTHREAD_MUTEX_INITIALIZER |
394 | #endif |
780 | #endif |
395 | |
781 | |
|
|
782 | #define SMUTEX(name) smutex name = SMUTEX_INITIALISER |
|
|
783 | #define SMUTEX_LOCK(name) pthread_mutex_lock (&(name)) |
|
|
784 | #define SMUTEX_UNLOCK(name) pthread_mutex_unlock (&(name)) |
|
|
785 | |
|
|
786 | typedef pthread_cond_t scond; |
|
|
787 | |
|
|
788 | #define SCOND(name) scond name = PTHREAD_COND_INITIALIZER |
|
|
789 | #define SCOND_SIGNAL(name) pthread_cond_signal (&(name)) |
|
|
790 | #define SCOND_BROADCAST(name) pthread_cond_broadcast (&(name)) |
|
|
791 | #define SCOND_WAIT(name,mutex) pthread_cond_wait (&(name), &(mutex)) |
|
|
792 | |
|
|
793 | #endif |
|
|
794 | |