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 (©) 2017,2018 Marc Alexander Lehmann / the Deliantra team |
4 | * Copyright (©) 2005,2006,2007,2008 Marc Alexander Lehmann / Robin Redeker / the Deliantra team |
5 | * Copyright (©) 2005,2006,2007,2008,2009,2010,2011,2012,2013,2014,2015,2016 Marc Alexander Lehmann / Robin Redeker / the Deliantra team |
5 | * |
6 | * |
6 | * Deliantra is free software: you can redistribute it and/or modify |
7 | * 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 |
8 | * 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 |
9 | * Free Software Foundation, either version 3 of the License, or (at your |
9 | * (at your option) any later version. |
10 | * option) any later version. |
10 | * |
11 | * |
11 | * This program is distributed in the hope that it will be useful, |
12 | * This program is distributed in the hope that it will be useful, |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
14 | * GNU General Public License for more details. |
15 | * GNU General Public License for more details. |
15 | * |
16 | * |
16 | * You should have received a copy of the GNU General Public License |
17 | * 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/>. |
18 | * and the GNU General Public License along with this program. If not, see |
|
|
19 | * <http://www.gnu.org/licenses/>. |
18 | * |
20 | * |
19 | * The authors can be reached via e-mail to <support@deliantra.net> |
21 | * The authors can be reached via e-mail to <support@deliantra.net> |
20 | */ |
22 | */ |
21 | |
23 | |
22 | #ifndef UTIL_H__ |
24 | #ifndef UTIL_H__ |
23 | #define UTIL_H__ |
25 | #define UTIL_H__ |
24 | |
26 | |
25 | #define DEBUG_POISON 0x00 // poison memory before freeing it if != 0 |
27 | #define DEBUG_POISON 0x00 // poison memory before freeing it if != 0 |
26 | #define DEBUG_SALLOC 0 // add a debug wrapper around all sallocs |
28 | #define DEBUG_SALLOC 0 // add a debug wrapper around all sallocs |
27 | #define PREFER_MALLOC 0 // use malloc and not the slice allocator |
29 | #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 | # define noinline __attribute__((__noinline__)) |
|
|
34 | #else |
|
|
35 | # define is_constant(c) 0 |
|
|
36 | # define expect(expr,value) (expr) |
|
|
37 | # define prefetch(addr,rw,locality) |
|
|
38 | # define noinline |
|
|
39 | #endif |
|
|
40 | |
|
|
41 | #if __GNUC__ < 4 || (__GNUC__ == 4 || __GNUC_MINOR__ < 4) |
|
|
42 | # define decltype(x) typeof(x) |
|
|
43 | #endif |
|
|
44 | |
|
|
45 | // put into ifs if you are very sure that the expression |
|
|
46 | // is mostly true or mosty false. note that these return |
|
|
47 | // booleans, not the expression. |
|
|
48 | #define expect_false(expr) expect ((expr) ? 1 : 0, 0) |
|
|
49 | #define expect_true(expr) expect ((expr) ? 1 : 0, 1) |
|
|
50 | |
30 | |
51 | #include <pthread.h> |
31 | #include <pthread.h> |
52 | |
32 | |
53 | #include <cstddef> |
33 | #include <cstddef> |
54 | #include <cmath> |
34 | #include <cmath> |
55 | #include <new> |
35 | #include <new> |
56 | #include <vector> |
36 | #include <vector> |
57 | |
37 | |
58 | #include <glib.h> |
38 | #include <glib.h> |
59 | |
39 | |
|
|
40 | #include <flat_hash_map.hpp> |
|
|
41 | |
60 | #include <shstr.h> |
42 | #include <shstr.h> |
61 | #include <traits.h> |
43 | #include <traits.h> |
|
|
44 | |
|
|
45 | #include "ecb.h" |
62 | |
46 | |
63 | #if DEBUG_SALLOC |
47 | #if DEBUG_SALLOC |
64 | # define g_slice_alloc0(s) debug_slice_alloc0(s) |
48 | # define g_slice_alloc0(s) debug_slice_alloc0(s) |
65 | # define g_slice_alloc(s) debug_slice_alloc(s) |
49 | # define g_slice_alloc(s) debug_slice_alloc(s) |
66 | # define g_slice_free1(s,p) debug_slice_free1(s,p) |
50 | # define g_slice_free1(s,p) debug_slice_free1(s,p) |
… | |
… | |
71 | # define g_slice_alloc0(s) calloc (1, (s)) |
55 | # define g_slice_alloc0(s) calloc (1, (s)) |
72 | # define g_slice_alloc(s) malloc ((s)) |
56 | # define g_slice_alloc(s) malloc ((s)) |
73 | # define g_slice_free1(s,p) free ((p)) |
57 | # define g_slice_free1(s,p) free ((p)) |
74 | #endif |
58 | #endif |
75 | |
59 | |
76 | // use C0X decltype for auto declarations until ISO C++ sanctifies them (if ever) |
|
|
77 | #define auto(var,expr) decltype(expr) var = (expr) |
|
|
78 | |
|
|
79 | // very ugly macro that basically declares and initialises a variable |
60 | // very ugly macro that basically declares and initialises a variable |
80 | // that is in scope for the next statement only |
61 | // that is in scope for the next statement only |
81 | // works only for stuff that can be assigned 0 and converts to false |
62 | // works only for stuff that can be assigned 0 and converts to false |
82 | // (note: works great for pointers) |
63 | // (note: works great for pointers) |
83 | // most ugly macro I ever wrote |
64 | // most ugly macro I ever wrote |
… | |
… | |
89 | |
70 | |
90 | // in range excluding end |
71 | // in range excluding end |
91 | #define IN_RANGE_EXC(val,beg,end) \ |
72 | #define IN_RANGE_EXC(val,beg,end) \ |
92 | ((unsigned int)(val) - (unsigned int)(beg) < (unsigned int)(end) - (unsigned int)(beg)) |
73 | ((unsigned int)(val) - (unsigned int)(beg) < (unsigned int)(end) - (unsigned int)(beg)) |
93 | |
74 | |
94 | void cleanup (const char *cause, bool make_core = false); |
75 | ecb_cold void cleanup (const char *cause, bool make_core = false); |
95 | void fork_abort (const char *msg); |
76 | ecb_cold void fork_abort (const char *msg); |
96 | |
77 | |
97 | // rationale for using (U) not (T) is to reduce signed/unsigned issues, |
78 | // rationale for using (U) not (T) is to reduce signed/unsigned issues, |
98 | // as a is often a constant while b is the variable. it is still a bug, though. |
79 | // as a is often a constant while b is the variable. it is still a bug, though. |
99 | template<typename T, typename U> static inline T min (T a, U b) { return (U)a < b ? (U)a : b; } |
80 | template<typename T, typename U> static inline T min (T a, U b) { return a < (T)b ? a : (T)b; } |
100 | template<typename T, typename U> static inline T max (T a, U b) { return (U)a > b ? (U)a : b; } |
81 | template<typename T, typename U> static inline T max (T a, U b) { return a > (T)b ? a : (T)b; } |
101 | 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; } |
82 | 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; } |
102 | |
83 | |
103 | template<typename T, typename U> static inline void min_it (T &v, U m) { v = min (v, (T)m); } |
84 | template<typename T, typename U> static inline void min_it (T &v, U m) { v = min (v, (T)m); } |
104 | template<typename T, typename U> static inline void max_it (T &v, U m) { v = max (v, (T)m); } |
85 | template<typename T, typename U> static inline void max_it (T &v, U m) { v = max (v, (T)m); } |
105 | 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); } |
86 | 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); } |
… | |
… | |
112 | // sign returns -1 or +1 |
93 | // sign returns -1 or +1 |
113 | template<typename T> |
94 | template<typename T> |
114 | static inline T sign (T v) { return v < 0 ? -1 : +1; } |
95 | static inline T sign (T v) { return v < 0 ? -1 : +1; } |
115 | // relies on 2c representation |
96 | // relies on 2c representation |
116 | template<> |
97 | template<> |
117 | inline sint8 sign (sint8 v) { return 1 - (sint8 (uint8 (v) >> 7) * 2); } |
98 | inline sint8 sign (sint8 v) { return 1 - (sint8 (uint8 (v) >> 7) * 2); } |
|
|
99 | template<> |
|
|
100 | inline sint16 sign (sint16 v) { return 1 - (sint16 (uint16 (v) >> 15) * 2); } |
|
|
101 | template<> |
|
|
102 | inline sint32 sign (sint32 v) { return 1 - (sint32 (uint32 (v) >> 31) * 2); } |
118 | |
103 | |
119 | // sign0 returns -1, 0 or +1 |
104 | // sign0 returns -1, 0 or +1 |
120 | template<typename T> |
105 | template<typename T> |
121 | static inline T sign0 (T v) { return v ? sign (v) : 0; } |
106 | static inline T sign0 (T v) { return v ? sign (v) : 0; } |
|
|
107 | |
|
|
108 | //clashes with C++0x |
|
|
109 | template<typename T, typename U> |
|
|
110 | static inline T copysign (T a, U b) { return a > 0 ? b : -b; } |
122 | |
111 | |
123 | // div* only work correctly for div > 0 |
112 | // div* only work correctly for div > 0 |
124 | // div, with correct rounding (< 0.5 downwards, >=0.5 upwards) |
113 | // div, with correct rounding (< 0.5 downwards, >=0.5 upwards) |
125 | template<typename T> static inline T div (T val, T div) |
114 | template<typename T> static inline T div (T val, T div) |
126 | { |
115 | { |
127 | return expect_false (val < 0) ? - ((-val + (div - 1) / 2) / div) : (val + div / 2) / div; |
116 | return ecb_expect_false (val < 0) ? - ((-val + (div - 1) / 2) / div) : (val + div / 2) / div; |
128 | } |
117 | } |
|
|
118 | |
|
|
119 | template<> inline float div (float val, float div) { return val / div; } |
|
|
120 | template<> inline double div (double val, double div) { return val / div; } |
|
|
121 | |
129 | // div, round-up |
122 | // div, round-up |
130 | template<typename T> static inline T div_ru (T val, T div) |
123 | template<typename T> static inline T div_ru (T val, T div) |
131 | { |
124 | { |
132 | return expect_false (val < 0) ? - ((-val ) / div) : (val + div - 1) / div; |
125 | return ecb_expect_false (val < 0) ? - ((-val ) / div) : (val + div - 1) / div; |
133 | } |
126 | } |
134 | // div, round-down |
127 | // div, round-down |
135 | template<typename T> static inline T div_rd (T val, T div) |
128 | template<typename T> static inline T div_rd (T val, T div) |
136 | { |
129 | { |
137 | return expect_false (val < 0) ? - ((-val + (div - 1) ) / div) : (val ) / div; |
130 | return ecb_expect_false (val < 0) ? - ((-val + (div - 1) ) / div) : (val ) / div; |
138 | } |
131 | } |
139 | |
132 | |
140 | // lerp* only work correctly for min_in < max_in |
133 | // lerp* only work correctly for min_in < max_in |
141 | // Linear intERPolate, scales val from min_in..max_in to min_out..max_out |
134 | // Linear intERPolate, scales val from min_in..max_in to min_out..max_out |
142 | template<typename T> |
135 | template<typename T> |
… | |
… | |
205 | int32_t d = b - a; |
198 | int32_t d = b - a; |
206 | d &= d >> 31; |
199 | d &= d >> 31; |
207 | return b - d; |
200 | return b - d; |
208 | } |
201 | } |
209 | |
202 | |
210 | // this is much faster than crossfires original algorithm |
203 | // this is much faster than crossfire's original algorithm |
211 | // on modern cpus |
204 | // on modern cpus |
212 | inline int |
205 | inline int |
213 | isqrt (int n) |
206 | isqrt (int n) |
214 | { |
207 | { |
215 | return (int)sqrtf ((float)n); |
208 | return (int)sqrtf ((float)n); |
|
|
209 | } |
|
|
210 | |
|
|
211 | // this is kind of like the ^^ operator, if it would exist, without sequence point. |
|
|
212 | // more handy than it looks like, due to the implicit !! done on its arguments |
|
|
213 | inline bool |
|
|
214 | logical_xor (bool a, bool b) |
|
|
215 | { |
|
|
216 | return a != b; |
|
|
217 | } |
|
|
218 | |
|
|
219 | inline bool |
|
|
220 | logical_implies (bool a, bool b) |
|
|
221 | { |
|
|
222 | return a <= b; |
216 | } |
223 | } |
217 | |
224 | |
218 | // this is only twice as fast as naive sqrtf (dx*dy+dy*dy) |
225 | // this is only twice as fast as naive sqrtf (dx*dy+dy*dy) |
219 | #if 0 |
226 | #if 0 |
220 | // and has a max. error of 6 in the range -100..+100. |
227 | // and has a max. error of 6 in the range -100..+100. |
221 | #else |
228 | #else |
222 | // and has a max. error of 9 in the range -100..+100. |
229 | // and has a max. error of 9 in the range -100..+100. |
223 | #endif |
230 | #endif |
224 | inline int |
231 | inline int |
225 | idistance (int dx, int dy) |
232 | idistance (int dx, int dy) |
226 | { |
233 | { |
227 | unsigned int dx_ = abs (dx); |
234 | unsigned int dx_ = abs (dx); |
228 | unsigned int dy_ = abs (dy); |
235 | unsigned int dy_ = abs (dy); |
229 | |
236 | |
230 | #if 0 |
237 | #if 0 |
231 | return dx_ > dy_ |
238 | return dx_ > dy_ |
… | |
… | |
234 | #else |
241 | #else |
235 | return dx_ + dy_ - min (dx_, dy_) * 5 / 8; |
242 | return dx_ + dy_ - min (dx_, dy_) * 5 / 8; |
236 | #endif |
243 | #endif |
237 | } |
244 | } |
238 | |
245 | |
|
|
246 | // can be substantially faster than floor, if your value range allows for it |
|
|
247 | template<typename T> |
|
|
248 | inline T |
|
|
249 | fastfloor (T x) |
|
|
250 | { |
|
|
251 | return std::floor (x); |
|
|
252 | } |
|
|
253 | |
|
|
254 | inline float |
|
|
255 | fastfloor (float x) |
|
|
256 | { |
|
|
257 | return sint32(x) - (x < 0); |
|
|
258 | } |
|
|
259 | |
|
|
260 | inline double |
|
|
261 | fastfloor (double x) |
|
|
262 | { |
|
|
263 | return sint64(x) - (x < 0); |
|
|
264 | } |
|
|
265 | |
239 | /* |
266 | /* |
240 | * absdir(int): Returns a number between 1 and 8, which represent |
267 | * absdir(int): Returns a number between 1 and 8, which represent |
241 | * the "absolute" direction of a number (it actually takes care of |
268 | * the "absolute" direction of a number (it actually takes care of |
242 | * "overflow" in previous calculations of a direction). |
269 | * "overflow" in previous calculations of a direction). |
243 | */ |
270 | */ |
… | |
… | |
245 | absdir (int d) |
272 | absdir (int d) |
246 | { |
273 | { |
247 | return ((d - 1) & 7) + 1; |
274 | return ((d - 1) & 7) + 1; |
248 | } |
275 | } |
249 | |
276 | |
|
|
277 | #define for_all_bits_sparse_32(mask, idxvar) \ |
|
|
278 | for (uint32_t idxvar, mask_ = mask; \ |
|
|
279 | mask_ && ((idxvar = ecb_ctz32 (mask_)), mask_ &= ~(1 << idxvar), 1);) |
|
|
280 | |
250 | extern ssize_t slice_alloc; // statistics |
281 | extern ssize_t slice_alloc; // statistics |
251 | |
282 | |
252 | void *salloc_ (int n) throw (std::bad_alloc); |
283 | void *salloc_ (int n) noexcept; |
253 | void *salloc_ (int n, void *src) throw (std::bad_alloc); |
284 | void *salloc_ (int n, void *src) noexcept; |
254 | |
285 | |
255 | // strictly the same as g_slice_alloc, but never returns 0 |
286 | // strictly the same as g_slice_alloc, but never returns 0 |
256 | template<typename T> |
287 | template<typename T> |
257 | inline T *salloc (int n = 1) throw (std::bad_alloc) { return (T *)salloc_ (n * sizeof (T)); } |
288 | inline T *salloc (int n = 1) { return (T *)salloc_ (n * sizeof (T)); } |
258 | |
289 | |
259 | // also copies src into the new area, like "memdup" |
290 | // also copies src into the new area, like "memdup" |
260 | // if src is 0, clears the memory |
291 | // if src is 0, clears the memory |
261 | template<typename T> |
292 | template<typename T> |
262 | inline T *salloc (int n, T *src) throw (std::bad_alloc) { return (T *)salloc_ (n * sizeof (T), (void *)src); } |
293 | inline T *salloc (int n, T *src) { return (T *)salloc_ (n * sizeof (T), (void *)src); } |
263 | |
294 | |
264 | // clears the memory |
295 | // clears the memory |
265 | template<typename T> |
296 | template<typename T> |
266 | inline T *salloc0(int n = 1) throw (std::bad_alloc) { return (T *)salloc_ (n * sizeof (T), 0); } |
297 | inline T *salloc0(int n = 1) { return (T *)salloc_ (n * sizeof (T), 0); } |
267 | |
298 | |
268 | // for symmetry |
299 | // for symmetry |
269 | template<typename T> |
300 | template<typename T> |
270 | inline void sfree (T *ptr, int n = 1) throw () |
301 | inline void sfree (T *ptr, int n = 1) noexcept |
271 | { |
302 | { |
272 | if (expect_true (ptr)) |
303 | if (ecb_expect_true (ptr)) |
273 | { |
304 | { |
274 | slice_alloc -= n * sizeof (T); |
305 | slice_alloc -= n * sizeof (T); |
275 | if (DEBUG_POISON) memset (ptr, DEBUG_POISON, n * sizeof (T)); |
306 | if (DEBUG_POISON) memset (ptr, DEBUG_POISON, n * sizeof (T)); |
276 | g_slice_free1 (n * sizeof (T), (void *)ptr); |
307 | g_slice_free1 (n * sizeof (T), (void *)ptr); |
277 | assert (slice_alloc >= 0);//D |
|
|
278 | } |
308 | } |
279 | } |
309 | } |
280 | |
310 | |
281 | // nulls the pointer |
311 | // nulls the pointer |
282 | template<typename T> |
312 | template<typename T> |
283 | inline void sfree0 (T *&ptr, int n = 1) throw () |
313 | inline void sfree0 (T *&ptr, int n = 1) noexcept |
284 | { |
314 | { |
285 | sfree<T> (ptr, n); |
315 | sfree<T> (ptr, n); |
286 | ptr = 0; |
316 | ptr = 0; |
287 | } |
317 | } |
288 | |
318 | |
… | |
… | |
344 | sfree ((char *)p, s); |
374 | sfree ((char *)p, s); |
345 | } |
375 | } |
346 | }; |
376 | }; |
347 | |
377 | |
348 | // a STL-compatible allocator that uses g_slice |
378 | // a STL-compatible allocator that uses g_slice |
349 | // boy, this is verbose |
379 | // boy, this is much less verbose in newer C++ versions |
350 | template<typename Tp> |
380 | template<typename Tp> |
351 | struct slice_allocator |
381 | struct slice_allocator |
352 | { |
382 | { |
353 | typedef size_t size_type; |
383 | using value_type = Tp; |
354 | typedef ptrdiff_t difference_type; |
|
|
355 | typedef Tp *pointer; |
|
|
356 | typedef const Tp *const_pointer; |
|
|
357 | typedef Tp &reference; |
|
|
358 | typedef const Tp &const_reference; |
|
|
359 | typedef Tp value_type; |
|
|
360 | |
384 | |
361 | template <class U> |
385 | slice_allocator () noexcept { } |
362 | struct rebind |
386 | template<class U> slice_allocator (const slice_allocator<U> &) noexcept {} |
|
|
387 | |
|
|
388 | value_type *allocate (std::size_t n) |
363 | { |
389 | { |
364 | typedef slice_allocator<U> other; |
390 | return salloc<Tp> (n); |
|
|
391 | } |
|
|
392 | |
|
|
393 | void deallocate (value_type *p, std::size_t n) |
|
|
394 | { |
|
|
395 | sfree<Tp> (p, n); |
|
|
396 | } |
|
|
397 | }; |
|
|
398 | |
|
|
399 | template<class T, class U> |
|
|
400 | bool operator == (const slice_allocator<T> &, const slice_allocator<U> &) noexcept |
|
|
401 | { |
|
|
402 | return true; |
|
|
403 | } |
|
|
404 | |
|
|
405 | template<class T, class U> |
|
|
406 | bool operator != (const slice_allocator<T> &x, const slice_allocator<U> &y) noexcept |
|
|
407 | { |
|
|
408 | return !(x == y); |
|
|
409 | } |
|
|
410 | |
|
|
411 | // basically a memory area, but refcounted |
|
|
412 | struct refcnt_buf |
|
|
413 | { |
|
|
414 | char *data; |
|
|
415 | |
|
|
416 | refcnt_buf (size_t size = 0); |
|
|
417 | refcnt_buf (void *data, size_t size); |
|
|
418 | |
|
|
419 | refcnt_buf (const refcnt_buf &src) |
|
|
420 | { |
|
|
421 | data = src.data; |
|
|
422 | inc (); |
|
|
423 | } |
|
|
424 | |
|
|
425 | ~refcnt_buf (); |
|
|
426 | |
|
|
427 | refcnt_buf &operator =(const refcnt_buf &src); |
|
|
428 | |
|
|
429 | operator char *() |
|
|
430 | { |
|
|
431 | return data; |
|
|
432 | } |
|
|
433 | |
|
|
434 | size_t size () const |
|
|
435 | { |
|
|
436 | return _size (); |
|
|
437 | } |
|
|
438 | |
|
|
439 | protected: |
|
|
440 | enum { |
|
|
441 | overhead = sizeof (uint32_t) * 2 |
365 | }; |
442 | }; |
366 | |
443 | |
367 | slice_allocator () throw () { } |
444 | uint32_t &_size () const |
368 | slice_allocator (const slice_allocator &) throw () { } |
|
|
369 | template<typename Tp2> |
|
|
370 | slice_allocator (const slice_allocator<Tp2> &) throw () { } |
|
|
371 | |
|
|
372 | ~slice_allocator () { } |
|
|
373 | |
|
|
374 | pointer address (reference x) const { return &x; } |
|
|
375 | const_pointer address (const_reference x) const { return &x; } |
|
|
376 | |
|
|
377 | pointer allocate (size_type n, const_pointer = 0) |
|
|
378 | { |
445 | { |
379 | return salloc<Tp> (n); |
446 | return ((unsigned int *)data)[-2]; |
380 | } |
447 | } |
381 | |
448 | |
382 | void deallocate (pointer p, size_type n) |
449 | uint32_t &_refcnt () const |
383 | { |
450 | { |
384 | sfree<Tp> (p, n); |
451 | return ((unsigned int *)data)[-1]; |
385 | } |
452 | } |
386 | |
453 | |
387 | size_type max_size () const throw () |
|
|
388 | { |
|
|
389 | return size_t (-1) / sizeof (Tp); |
|
|
390 | } |
|
|
391 | |
|
|
392 | void construct (pointer p, const Tp &val) |
|
|
393 | { |
|
|
394 | ::new (p) Tp (val); |
|
|
395 | } |
|
|
396 | |
|
|
397 | void destroy (pointer p) |
|
|
398 | { |
|
|
399 | p->~Tp (); |
|
|
400 | } |
|
|
401 | }; |
|
|
402 | |
|
|
403 | // P. L'Ecuyer, “Maximally Equidistributed Combined Tausworthe Generators”, Mathematics of Computation, 65, 213 (1996), 203–213. |
|
|
404 | // http://www.iro.umontreal.ca/~lecuyer/myftp/papers/tausme.ps |
|
|
405 | // http://www.iro.umontreal.ca/~lecuyer/myftp/papers/tausme2.ps |
|
|
406 | struct tausworthe_random_generator |
|
|
407 | { |
|
|
408 | uint32_t state [4]; |
|
|
409 | |
|
|
410 | void operator =(const tausworthe_random_generator &src) |
|
|
411 | { |
|
|
412 | state [0] = src.state [0]; |
|
|
413 | state [1] = src.state [1]; |
|
|
414 | state [2] = src.state [2]; |
|
|
415 | state [3] = src.state [3]; |
|
|
416 | } |
|
|
417 | |
|
|
418 | void seed (uint32_t seed); |
|
|
419 | uint32_t next (); |
|
|
420 | }; |
|
|
421 | |
|
|
422 | // Xorshift RNGs, George Marsaglia |
|
|
423 | // http://www.jstatsoft.org/v08/i14/paper |
|
|
424 | // this one is about 40% faster than the tausworthe one above (i.e. not much), |
|
|
425 | // despite the inlining, and has the issue of only creating 2**32-1 numbers. |
|
|
426 | // see also http://www.iro.umontreal.ca/~lecuyer/myftp/papers/xorshift.pdf |
|
|
427 | struct xorshift_random_generator |
|
|
428 | { |
|
|
429 | uint32_t x, y; |
|
|
430 | |
|
|
431 | void operator =(const xorshift_random_generator &src) |
|
|
432 | { |
|
|
433 | x = src.x; |
|
|
434 | y = src.y; |
|
|
435 | } |
|
|
436 | |
|
|
437 | void seed (uint32_t seed) |
454 | void _alloc (uint32_t size) |
438 | { |
455 | { |
439 | x = seed; |
456 | data = ((char *)salloc<char> (size + overhead)) + overhead; |
440 | y = seed * 69069U; |
457 | _size () = size; |
|
|
458 | _refcnt () = 1; |
441 | } |
459 | } |
442 | |
460 | |
443 | uint32_t next () |
461 | void _dealloc (); |
444 | { |
|
|
445 | uint32_t t = x ^ (x << 10); |
|
|
446 | x = y; |
|
|
447 | y = y ^ (y >> 13) ^ t ^ (t >> 10); |
|
|
448 | return y; |
|
|
449 | } |
|
|
450 | }; |
|
|
451 | |
462 | |
452 | template<class generator> |
463 | void inc () |
453 | struct random_number_generator : generator |
|
|
454 | { |
|
|
455 | // uniform distribution, 0 .. max (0, num - 1) |
|
|
456 | uint32_t operator ()(uint32_t num) |
|
|
457 | { |
464 | { |
458 | return !is_constant (num) ? get_range (num) // non-constant |
465 | ++_refcnt (); |
459 | : num & (num - 1) ? (this->next () * (uint64_t)num) >> 32U // constant, non-power-of-two |
|
|
460 | : this->next () & (num - 1); // constant, power-of-two |
|
|
461 | } |
466 | } |
462 | |
467 | |
463 | // return a number within (min .. max) |
468 | void dec () |
464 | int operator () (int r_min, int r_max) |
|
|
465 | { |
469 | { |
466 | return is_constant (r_min) && is_constant (r_max) && r_min <= r_max |
470 | if (!--_refcnt ()) |
467 | ? r_min + operator ()(r_max - r_min + 1) |
471 | _dealloc (); |
468 | : get_range (r_min, r_max); |
|
|
469 | } |
472 | } |
470 | |
|
|
471 | double operator ()() |
|
|
472 | { |
|
|
473 | return this->next () / (double)0xFFFFFFFFU; |
|
|
474 | } |
|
|
475 | |
|
|
476 | protected: |
|
|
477 | uint32_t get_range (uint32_t r_max); |
|
|
478 | int get_range (int r_min, int r_max); |
|
|
479 | }; |
473 | }; |
480 | |
|
|
481 | typedef random_number_generator<tausworthe_random_generator> rand_gen; |
|
|
482 | |
|
|
483 | extern rand_gen rndm, rmg_rndm; |
|
|
484 | |
474 | |
485 | INTERFACE_CLASS (attachable) |
475 | INTERFACE_CLASS (attachable) |
486 | struct refcnt_base |
476 | struct refcnt_base |
487 | { |
477 | { |
488 | typedef int refcnt_t; |
478 | typedef int refcnt_t; |
… | |
… | |
503 | // p if not null |
493 | // p if not null |
504 | refcnt_base::refcnt_t *refcnt_ref () { return p ? &p->refcnt : &refcnt_dummy; } |
494 | refcnt_base::refcnt_t *refcnt_ref () { return p ? &p->refcnt : &refcnt_dummy; } |
505 | |
495 | |
506 | void refcnt_dec () |
496 | void refcnt_dec () |
507 | { |
497 | { |
508 | if (!is_constant (p)) |
498 | if (!ecb_is_constant (p)) |
509 | --*refcnt_ref (); |
499 | --*refcnt_ref (); |
510 | else if (p) |
500 | else if (p) |
511 | --p->refcnt; |
501 | --p->refcnt; |
512 | } |
502 | } |
513 | |
503 | |
514 | void refcnt_inc () |
504 | void refcnt_inc () |
515 | { |
505 | { |
516 | if (!is_constant (p)) |
506 | if (!ecb_is_constant (p)) |
517 | ++*refcnt_ref (); |
507 | ++*refcnt_ref (); |
518 | else if (p) |
508 | else if (p) |
519 | ++p->refcnt; |
509 | ++p->refcnt; |
520 | } |
510 | } |
521 | |
511 | |
… | |
… | |
550 | typedef refptr<maptile> maptile_ptr; |
540 | typedef refptr<maptile> maptile_ptr; |
551 | typedef refptr<object> object_ptr; |
541 | typedef refptr<object> object_ptr; |
552 | typedef refptr<archetype> arch_ptr; |
542 | typedef refptr<archetype> arch_ptr; |
553 | typedef refptr<client> client_ptr; |
543 | typedef refptr<client> client_ptr; |
554 | typedef refptr<player> player_ptr; |
544 | typedef refptr<player> player_ptr; |
|
|
545 | typedef refptr<region> region_ptr; |
|
|
546 | |
|
|
547 | #define STRHSH_NULL 2166136261 |
|
|
548 | |
|
|
549 | static inline uint32_t |
|
|
550 | strhsh (const char *s) |
|
|
551 | { |
|
|
552 | // use FNV-1a hash (http://isthe.com/chongo/tech/comp/fnv/) |
|
|
553 | // it is about twice as fast as the one-at-a-time one, |
|
|
554 | // with good distribution. |
|
|
555 | // FNV-1a is faster on many cpus because the multiplication |
|
|
556 | // runs concurrently with the looping logic. |
|
|
557 | // we modify the hash a bit to improve its distribution |
|
|
558 | uint32_t hash = STRHSH_NULL; |
|
|
559 | |
|
|
560 | while (*s) |
|
|
561 | hash = (hash ^ *s++) * 16777619U; |
|
|
562 | |
|
|
563 | return hash ^ (hash >> 16); |
|
|
564 | } |
|
|
565 | |
|
|
566 | static inline uint32_t |
|
|
567 | memhsh (const char *s, size_t len) |
|
|
568 | { |
|
|
569 | uint32_t hash = STRHSH_NULL; |
|
|
570 | |
|
|
571 | while (len--) |
|
|
572 | hash = (hash ^ *s++) * 16777619U; |
|
|
573 | |
|
|
574 | return hash; |
|
|
575 | } |
555 | |
576 | |
556 | struct str_hash |
577 | struct str_hash |
557 | { |
578 | { |
558 | std::size_t operator ()(const char *s) const |
579 | std::size_t operator ()(const char *s) const |
559 | { |
580 | { |
560 | #if 0 |
|
|
561 | uint32_t hash = 0; |
|
|
562 | |
|
|
563 | /* use the one-at-a-time hash function, which supposedly is |
|
|
564 | * better than the djb2-like one used by perl5.005, but |
|
|
565 | * certainly is better then the bug used here before. |
|
|
566 | * see http://burtleburtle.net/bob/hash/doobs.html |
|
|
567 | */ |
|
|
568 | while (*s) |
|
|
569 | { |
|
|
570 | hash += *s++; |
|
|
571 | hash += hash << 10; |
|
|
572 | hash ^= hash >> 6; |
|
|
573 | } |
|
|
574 | |
|
|
575 | hash += hash << 3; |
|
|
576 | hash ^= hash >> 11; |
|
|
577 | hash += hash << 15; |
|
|
578 | #else |
|
|
579 | // use FNV-1a hash (http://isthe.com/chongo/tech/comp/fnv/) |
|
|
580 | // it is about twice as fast as the one-at-a-time one, |
|
|
581 | // with good distribution. |
|
|
582 | // FNV-1a is faster on many cpus because the multiplication |
|
|
583 | // runs concurrent with the looping logic. |
|
|
584 | uint32_t hash = 2166136261; |
|
|
585 | |
|
|
586 | while (*s) |
|
|
587 | hash = (hash ^ *s++) * 16777619; |
|
|
588 | #endif |
|
|
589 | |
|
|
590 | return hash; |
581 | return strhsh (s); |
591 | } |
582 | } |
|
|
583 | |
|
|
584 | std::size_t operator ()(const shstr &s) const |
|
|
585 | { |
|
|
586 | return strhsh (s); |
|
|
587 | } |
|
|
588 | |
|
|
589 | typedef ska::power_of_two_hash_policy hash_policy; |
592 | }; |
590 | }; |
593 | |
591 | |
594 | struct str_equal |
592 | struct str_equal |
595 | { |
593 | { |
596 | bool operator ()(const char *a, const char *b) const |
594 | bool operator ()(const char *a, const char *b) const |
… | |
… | |
622 | } |
620 | } |
623 | }; |
621 | }; |
624 | |
622 | |
625 | // This container blends advantages of linked lists |
623 | // This container blends advantages of linked lists |
626 | // (efficiency) with vectors (random access) by |
624 | // (efficiency) with vectors (random access) by |
627 | // by using an unordered vector and storing the vector |
625 | // using an unordered vector and storing the vector |
628 | // index inside the object. |
626 | // index inside the object. |
629 | // |
627 | // |
630 | // + memory-efficient on most 64 bit archs |
628 | // + memory-efficient on most 64 bit archs |
631 | // + O(1) insert/remove |
629 | // + O(1) insert/remove |
632 | // + free unique (but varying) id for inserted objects |
630 | // + free unique (but varying) id for inserted objects |
… | |
… | |
669 | insert (&obj); |
667 | insert (&obj); |
670 | } |
668 | } |
671 | |
669 | |
672 | void erase (T *obj) |
670 | void erase (T *obj) |
673 | { |
671 | { |
674 | unsigned int pos = obj->*indexmember; |
672 | object_vector_index pos = obj->*indexmember; |
675 | obj->*indexmember = 0; |
673 | obj->*indexmember = 0; |
676 | |
674 | |
677 | if (pos < this->size ()) |
675 | if (pos < this->size ()) |
678 | { |
676 | { |
679 | (*this)[pos - 1] = (*this)[this->size () - 1]; |
677 | (*this)[pos - 1] = (*this)[this->size () - 1]; |
… | |
… | |
687 | { |
685 | { |
688 | erase (&obj); |
686 | erase (&obj); |
689 | } |
687 | } |
690 | }; |
688 | }; |
691 | |
689 | |
|
|
690 | ///////////////////////////////////////////////////////////////////////////// |
|
|
691 | |
|
|
692 | // something like a vector or stack, but without |
|
|
693 | // out of bounds checking |
|
|
694 | template<typename T> |
|
|
695 | struct fixed_stack |
|
|
696 | { |
|
|
697 | T *data; |
|
|
698 | int size; |
|
|
699 | int max; |
|
|
700 | |
|
|
701 | fixed_stack () |
|
|
702 | : size (0), data (0) |
|
|
703 | { |
|
|
704 | } |
|
|
705 | |
|
|
706 | fixed_stack (int max) |
|
|
707 | : size (0), max (max) |
|
|
708 | { |
|
|
709 | data = salloc<T> (max); |
|
|
710 | } |
|
|
711 | |
|
|
712 | void reset (int new_max) |
|
|
713 | { |
|
|
714 | sfree (data, max); |
|
|
715 | size = 0; |
|
|
716 | max = new_max; |
|
|
717 | data = salloc<T> (max); |
|
|
718 | } |
|
|
719 | |
|
|
720 | void free () |
|
|
721 | { |
|
|
722 | sfree (data, max); |
|
|
723 | data = 0; |
|
|
724 | } |
|
|
725 | |
|
|
726 | ~fixed_stack () |
|
|
727 | { |
|
|
728 | sfree (data, max); |
|
|
729 | } |
|
|
730 | |
|
|
731 | T &operator[](int idx) |
|
|
732 | { |
|
|
733 | return data [idx]; |
|
|
734 | } |
|
|
735 | |
|
|
736 | void push (T v) |
|
|
737 | { |
|
|
738 | data [size++] = v; |
|
|
739 | } |
|
|
740 | |
|
|
741 | T &pop () |
|
|
742 | { |
|
|
743 | return data [--size]; |
|
|
744 | } |
|
|
745 | |
|
|
746 | T remove (int idx) |
|
|
747 | { |
|
|
748 | T v = data [idx]; |
|
|
749 | |
|
|
750 | data [idx] = data [--size]; |
|
|
751 | |
|
|
752 | return v; |
|
|
753 | } |
|
|
754 | }; |
|
|
755 | |
|
|
756 | ///////////////////////////////////////////////////////////////////////////// |
|
|
757 | |
692 | // basically does what strncpy should do, but appends "..." to strings exceeding length |
758 | // basically does what strncpy should do, but appends "..." to strings exceeding length |
693 | // returns the number of bytes actually used (including \0) |
759 | // returns the number of bytes actually used (including \0) |
694 | int assign (char *dst, const char *src, int maxsize); |
760 | int assign (char *dst, const char *src, int maxsize); |
695 | |
761 | |
696 | // type-safe version of assign |
762 | // type-safe version of assign |
… | |
… | |
705 | // return current time as timestamp |
771 | // return current time as timestamp |
706 | tstamp now (); |
772 | tstamp now (); |
707 | |
773 | |
708 | int similar_direction (int a, int b); |
774 | int similar_direction (int a, int b); |
709 | |
775 | |
710 | // like sprintf, but returns a "static" buffer |
776 | // like v?sprintf, but returns a "static" buffer |
711 | const char *format (const char *format, ...); |
777 | char *vformat (const char *format, va_list ap); |
|
|
778 | char *format (const char *format, ...) ecb_attribute ((format (printf, 1, 2))); |
|
|
779 | |
|
|
780 | // safety-check player input which will become object->msg |
|
|
781 | bool msg_is_safe (const char *msg); |
712 | |
782 | |
713 | ///////////////////////////////////////////////////////////////////////////// |
783 | ///////////////////////////////////////////////////////////////////////////// |
714 | // threads, very very thin wrappers around pthreads |
784 | // threads, very very thin wrappers around pthreads |
715 | |
785 | |
716 | struct thread |
786 | struct thread |