1 | // random number generation (out of line) -*- C++ -*- |
2 | |
3 | // Copyright (C) 2009-2022 Free Software Foundation, Inc. |
4 | // |
5 | // This file is part of the GNU ISO C++ Library. This library is free |
6 | // software; you can redistribute it and/or modify it under the |
7 | // terms of the GNU General Public License as published by the |
8 | // Free Software Foundation; either version 3, or (at your option) |
9 | // any later version. |
10 | |
11 | // This library 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 | // Under Section 7 of GPL version 3, you are granted additional |
17 | // permissions described in the GCC Runtime Library Exception, version |
18 | // 3.1, as published by the Free Software Foundation. |
19 | |
20 | // You should have received a copy of the GNU General Public License and |
21 | // a copy of the GCC Runtime Library Exception along with this program; |
22 | // see the files COPYING3 and COPYING.RUNTIME respectively. If not, see |
23 | // <http://www.gnu.org/licenses/>. |
24 | |
25 | /** @file bits/random.tcc |
26 | * This is an internal header file, included by other library headers. |
27 | * Do not attempt to use it directly. @headername{random} |
28 | */ |
29 | |
30 | #ifndef _RANDOM_TCC |
31 | #define _RANDOM_TCC 1 |
32 | |
33 | #include <numeric> // std::accumulate and std::partial_sum |
34 | |
35 | namespace std _GLIBCXX_VISIBILITY(default) |
36 | { |
37 | _GLIBCXX_BEGIN_NAMESPACE_VERSION |
38 | |
39 | /// @cond undocumented |
40 | // (Further) implementation-space details. |
41 | namespace __detail |
42 | { |
43 | // General case for x = (ax + c) mod m -- use Schrage's algorithm |
44 | // to avoid integer overflow. |
45 | // |
46 | // Preconditions: a > 0, m > 0. |
47 | // |
48 | // Note: only works correctly for __m % __a < __m / __a. |
49 | template<typename _Tp, _Tp __m, _Tp __a, _Tp __c> |
50 | _Tp |
51 | _Mod<_Tp, __m, __a, __c, false, true>:: |
52 | __calc(_Tp __x) |
53 | { |
54 | if (__a == 1) |
55 | __x %= __m; |
56 | else |
57 | { |
58 | static const _Tp __q = __m / __a; |
59 | static const _Tp __r = __m % __a; |
60 | |
61 | _Tp __t1 = __a * (__x % __q); |
62 | _Tp __t2 = __r * (__x / __q); |
63 | if (__t1 >= __t2) |
64 | __x = __t1 - __t2; |
65 | else |
66 | __x = __m - __t2 + __t1; |
67 | } |
68 | |
69 | if (__c != 0) |
70 | { |
71 | const _Tp __d = __m - __x; |
72 | if (__d > __c) |
73 | __x += __c; |
74 | else |
75 | __x = __c - __d; |
76 | } |
77 | return __x; |
78 | } |
79 | |
80 | template<typename _InputIterator, typename _OutputIterator, |
81 | typename _Tp> |
82 | _OutputIterator |
83 | __normalize(_InputIterator __first, _InputIterator __last, |
84 | _OutputIterator __result, const _Tp& __factor) |
85 | { |
86 | for (; __first != __last; ++__first, ++__result) |
87 | *__result = *__first / __factor; |
88 | return __result; |
89 | } |
90 | |
91 | } // namespace __detail |
92 | /// @endcond |
93 | |
94 | #if ! __cpp_inline_variables |
95 | template<typename _UIntType, _UIntType __a, _UIntType __c, _UIntType __m> |
96 | constexpr _UIntType |
97 | linear_congruential_engine<_UIntType, __a, __c, __m>::multiplier; |
98 | |
99 | template<typename _UIntType, _UIntType __a, _UIntType __c, _UIntType __m> |
100 | constexpr _UIntType |
101 | linear_congruential_engine<_UIntType, __a, __c, __m>::increment; |
102 | |
103 | template<typename _UIntType, _UIntType __a, _UIntType __c, _UIntType __m> |
104 | constexpr _UIntType |
105 | linear_congruential_engine<_UIntType, __a, __c, __m>::modulus; |
106 | |
107 | template<typename _UIntType, _UIntType __a, _UIntType __c, _UIntType __m> |
108 | constexpr _UIntType |
109 | linear_congruential_engine<_UIntType, __a, __c, __m>::default_seed; |
110 | #endif |
111 | |
112 | /** |
113 | * Seeds the LCR with integral value @p __s, adjusted so that the |
114 | * ring identity is never a member of the convergence set. |
115 | */ |
116 | template<typename _UIntType, _UIntType __a, _UIntType __c, _UIntType __m> |
117 | void |
118 | linear_congruential_engine<_UIntType, __a, __c, __m>:: |
119 | seed(result_type __s) |
120 | { |
121 | if ((__detail::__mod<_UIntType, __m>(__c) == 0) |
122 | && (__detail::__mod<_UIntType, __m>(__s) == 0)) |
123 | _M_x = 1; |
124 | else |
125 | _M_x = __detail::__mod<_UIntType, __m>(__s); |
126 | } |
127 | |
128 | /** |
129 | * Seeds the LCR engine with a value generated by @p __q. |
130 | */ |
131 | template<typename _UIntType, _UIntType __a, _UIntType __c, _UIntType __m> |
132 | template<typename _Sseq> |
133 | auto |
134 | linear_congruential_engine<_UIntType, __a, __c, __m>:: |
135 | seed(_Sseq& __q) |
136 | -> _If_seed_seq<_Sseq> |
137 | { |
138 | const _UIntType __k0 = __m == 0 ? std::numeric_limits<_UIntType>::digits |
139 | : std::__lg(__m); |
140 | const _UIntType __k = (__k0 + 31) / 32; |
141 | uint_least32_t __arr[__k + 3]; |
142 | __q.generate(__arr + 0, __arr + __k + 3); |
143 | _UIntType __factor = 1u; |
144 | _UIntType __sum = 0u; |
145 | for (size_t __j = 0; __j < __k; ++__j) |
146 | { |
147 | __sum += __arr[__j + 3] * __factor; |
148 | __factor *= __detail::_Shift<_UIntType, 32>::__value; |
149 | } |
150 | seed(__sum); |
151 | } |
152 | |
153 | template<typename _UIntType, _UIntType __a, _UIntType __c, _UIntType __m, |
154 | typename _CharT, typename _Traits> |
155 | std::basic_ostream<_CharT, _Traits>& |
156 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
157 | const linear_congruential_engine<_UIntType, |
158 | __a, __c, __m>& __lcr) |
159 | { |
160 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
161 | |
162 | const typename __ios_base::fmtflags __flags = __os.flags(); |
163 | const _CharT __fill = __os.fill(); |
164 | __os.flags(__ios_base::dec | __ios_base::fixed | __ios_base::left); |
165 | __os.fill(__os.widen(' ')); |
166 | |
167 | __os << __lcr._M_x; |
168 | |
169 | __os.flags(__flags); |
170 | __os.fill(__fill); |
171 | return __os; |
172 | } |
173 | |
174 | template<typename _UIntType, _UIntType __a, _UIntType __c, _UIntType __m, |
175 | typename _CharT, typename _Traits> |
176 | std::basic_istream<_CharT, _Traits>& |
177 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
178 | linear_congruential_engine<_UIntType, __a, __c, __m>& __lcr) |
179 | { |
180 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
181 | |
182 | const typename __ios_base::fmtflags __flags = __is.flags(); |
183 | __is.flags(__ios_base::dec); |
184 | |
185 | __is >> __lcr._M_x; |
186 | |
187 | __is.flags(__flags); |
188 | return __is; |
189 | } |
190 | |
191 | #if ! __cpp_inline_variables |
192 | template<typename _UIntType, |
193 | size_t __w, size_t __n, size_t __m, size_t __r, |
194 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
195 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
196 | _UIntType __f> |
197 | constexpr size_t |
198 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
199 | __s, __b, __t, __c, __l, __f>::word_size; |
200 | |
201 | template<typename _UIntType, |
202 | size_t __w, size_t __n, size_t __m, size_t __r, |
203 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
204 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
205 | _UIntType __f> |
206 | constexpr size_t |
207 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
208 | __s, __b, __t, __c, __l, __f>::state_size; |
209 | |
210 | template<typename _UIntType, |
211 | size_t __w, size_t __n, size_t __m, size_t __r, |
212 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
213 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
214 | _UIntType __f> |
215 | constexpr size_t |
216 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
217 | __s, __b, __t, __c, __l, __f>::shift_size; |
218 | |
219 | template<typename _UIntType, |
220 | size_t __w, size_t __n, size_t __m, size_t __r, |
221 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
222 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
223 | _UIntType __f> |
224 | constexpr size_t |
225 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
226 | __s, __b, __t, __c, __l, __f>::mask_bits; |
227 | |
228 | template<typename _UIntType, |
229 | size_t __w, size_t __n, size_t __m, size_t __r, |
230 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
231 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
232 | _UIntType __f> |
233 | constexpr _UIntType |
234 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
235 | __s, __b, __t, __c, __l, __f>::xor_mask; |
236 | |
237 | template<typename _UIntType, |
238 | size_t __w, size_t __n, size_t __m, size_t __r, |
239 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
240 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
241 | _UIntType __f> |
242 | constexpr size_t |
243 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
244 | __s, __b, __t, __c, __l, __f>::tempering_u; |
245 | |
246 | template<typename _UIntType, |
247 | size_t __w, size_t __n, size_t __m, size_t __r, |
248 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
249 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
250 | _UIntType __f> |
251 | constexpr _UIntType |
252 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
253 | __s, __b, __t, __c, __l, __f>::tempering_d; |
254 | |
255 | template<typename _UIntType, |
256 | size_t __w, size_t __n, size_t __m, size_t __r, |
257 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
258 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
259 | _UIntType __f> |
260 | constexpr size_t |
261 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
262 | __s, __b, __t, __c, __l, __f>::tempering_s; |
263 | |
264 | template<typename _UIntType, |
265 | size_t __w, size_t __n, size_t __m, size_t __r, |
266 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
267 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
268 | _UIntType __f> |
269 | constexpr _UIntType |
270 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
271 | __s, __b, __t, __c, __l, __f>::tempering_b; |
272 | |
273 | template<typename _UIntType, |
274 | size_t __w, size_t __n, size_t __m, size_t __r, |
275 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
276 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
277 | _UIntType __f> |
278 | constexpr size_t |
279 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
280 | __s, __b, __t, __c, __l, __f>::tempering_t; |
281 | |
282 | template<typename _UIntType, |
283 | size_t __w, size_t __n, size_t __m, size_t __r, |
284 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
285 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
286 | _UIntType __f> |
287 | constexpr _UIntType |
288 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
289 | __s, __b, __t, __c, __l, __f>::tempering_c; |
290 | |
291 | template<typename _UIntType, |
292 | size_t __w, size_t __n, size_t __m, size_t __r, |
293 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
294 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
295 | _UIntType __f> |
296 | constexpr size_t |
297 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
298 | __s, __b, __t, __c, __l, __f>::tempering_l; |
299 | |
300 | template<typename _UIntType, |
301 | size_t __w, size_t __n, size_t __m, size_t __r, |
302 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
303 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
304 | _UIntType __f> |
305 | constexpr _UIntType |
306 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
307 | __s, __b, __t, __c, __l, __f>:: |
308 | initialization_multiplier; |
309 | |
310 | template<typename _UIntType, |
311 | size_t __w, size_t __n, size_t __m, size_t __r, |
312 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
313 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
314 | _UIntType __f> |
315 | constexpr _UIntType |
316 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
317 | __s, __b, __t, __c, __l, __f>::default_seed; |
318 | #endif |
319 | |
320 | template<typename _UIntType, |
321 | size_t __w, size_t __n, size_t __m, size_t __r, |
322 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
323 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
324 | _UIntType __f> |
325 | void |
326 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
327 | __s, __b, __t, __c, __l, __f>:: |
328 | seed(result_type __sd) |
329 | { |
330 | _M_x[0] = __detail::__mod<_UIntType, |
331 | __detail::_Shift<_UIntType, __w>::__value>(__sd); |
332 | |
333 | for (size_t __i = 1; __i < state_size; ++__i) |
334 | { |
335 | _UIntType __x = _M_x[__i - 1]; |
336 | __x ^= __x >> (__w - 2); |
337 | __x *= __f; |
338 | __x += __detail::__mod<_UIntType, __n>(__i); |
339 | _M_x[__i] = __detail::__mod<_UIntType, |
340 | __detail::_Shift<_UIntType, __w>::__value>(__x); |
341 | } |
342 | _M_p = state_size; |
343 | } |
344 | |
345 | template<typename _UIntType, |
346 | size_t __w, size_t __n, size_t __m, size_t __r, |
347 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
348 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
349 | _UIntType __f> |
350 | template<typename _Sseq> |
351 | auto |
352 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
353 | __s, __b, __t, __c, __l, __f>:: |
354 | seed(_Sseq& __q) |
355 | -> _If_seed_seq<_Sseq> |
356 | { |
357 | const _UIntType __upper_mask = (~_UIntType()) << __r; |
358 | const size_t __k = (__w + 31) / 32; |
359 | uint_least32_t __arr[__n * __k]; |
360 | __q.generate(__arr + 0, __arr + __n * __k); |
361 | |
362 | bool __zero = true; |
363 | for (size_t __i = 0; __i < state_size; ++__i) |
364 | { |
365 | _UIntType __factor = 1u; |
366 | _UIntType __sum = 0u; |
367 | for (size_t __j = 0; __j < __k; ++__j) |
368 | { |
369 | __sum += __arr[__k * __i + __j] * __factor; |
370 | __factor *= __detail::_Shift<_UIntType, 32>::__value; |
371 | } |
372 | _M_x[__i] = __detail::__mod<_UIntType, |
373 | __detail::_Shift<_UIntType, __w>::__value>(__sum); |
374 | |
375 | if (__zero) |
376 | { |
377 | if (__i == 0) |
378 | { |
379 | if ((_M_x[0] & __upper_mask) != 0u) |
380 | __zero = false; |
381 | } |
382 | else if (_M_x[__i] != 0u) |
383 | __zero = false; |
384 | } |
385 | } |
386 | if (__zero) |
387 | _M_x[0] = __detail::_Shift<_UIntType, __w - 1>::__value; |
388 | _M_p = state_size; |
389 | } |
390 | |
391 | template<typename _UIntType, size_t __w, |
392 | size_t __n, size_t __m, size_t __r, |
393 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
394 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
395 | _UIntType __f> |
396 | void |
397 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
398 | __s, __b, __t, __c, __l, __f>:: |
399 | _M_gen_rand(void) |
400 | { |
401 | const _UIntType __upper_mask = (~_UIntType()) << __r; |
402 | const _UIntType __lower_mask = ~__upper_mask; |
403 | |
404 | for (size_t __k = 0; __k < (__n - __m); ++__k) |
405 | { |
406 | _UIntType __y = ((_M_x[__k] & __upper_mask) |
407 | | (_M_x[__k + 1] & __lower_mask)); |
408 | _M_x[__k] = (_M_x[__k + __m] ^ (__y >> 1) |
409 | ^ ((__y & 0x01) ? __a : 0)); |
410 | } |
411 | |
412 | for (size_t __k = (__n - __m); __k < (__n - 1); ++__k) |
413 | { |
414 | _UIntType __y = ((_M_x[__k] & __upper_mask) |
415 | | (_M_x[__k + 1] & __lower_mask)); |
416 | _M_x[__k] = (_M_x[__k + (__m - __n)] ^ (__y >> 1) |
417 | ^ ((__y & 0x01) ? __a : 0)); |
418 | } |
419 | |
420 | _UIntType __y = ((_M_x[__n - 1] & __upper_mask) |
421 | | (_M_x[0] & __lower_mask)); |
422 | _M_x[__n - 1] = (_M_x[__m - 1] ^ (__y >> 1) |
423 | ^ ((__y & 0x01) ? __a : 0)); |
424 | _M_p = 0; |
425 | } |
426 | |
427 | template<typename _UIntType, size_t __w, |
428 | size_t __n, size_t __m, size_t __r, |
429 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
430 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
431 | _UIntType __f> |
432 | void |
433 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
434 | __s, __b, __t, __c, __l, __f>:: |
435 | discard(unsigned long long __z) |
436 | { |
437 | while (__z > state_size - _M_p) |
438 | { |
439 | __z -= state_size - _M_p; |
440 | _M_gen_rand(); |
441 | } |
442 | _M_p += __z; |
443 | } |
444 | |
445 | template<typename _UIntType, size_t __w, |
446 | size_t __n, size_t __m, size_t __r, |
447 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
448 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
449 | _UIntType __f> |
450 | typename |
451 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
452 | __s, __b, __t, __c, __l, __f>::result_type |
453 | mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, |
454 | __s, __b, __t, __c, __l, __f>:: |
455 | operator()() |
456 | { |
457 | // Reload the vector - cost is O(n) amortized over n calls. |
458 | if (_M_p >= state_size) |
459 | _M_gen_rand(); |
460 | |
461 | // Calculate o(x(i)). |
462 | result_type __z = _M_x[_M_p++]; |
463 | __z ^= (__z >> __u) & __d; |
464 | __z ^= (__z << __s) & __b; |
465 | __z ^= (__z << __t) & __c; |
466 | __z ^= (__z >> __l); |
467 | |
468 | return __z; |
469 | } |
470 | |
471 | template<typename _UIntType, size_t __w, |
472 | size_t __n, size_t __m, size_t __r, |
473 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
474 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
475 | _UIntType __f, typename _CharT, typename _Traits> |
476 | std::basic_ostream<_CharT, _Traits>& |
477 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
478 | const mersenne_twister_engine<_UIntType, __w, __n, __m, |
479 | __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>& __x) |
480 | { |
481 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
482 | |
483 | const typename __ios_base::fmtflags __flags = __os.flags(); |
484 | const _CharT __fill = __os.fill(); |
485 | const _CharT __space = __os.widen(' '); |
486 | __os.flags(__ios_base::dec | __ios_base::fixed | __ios_base::left); |
487 | __os.fill(__space); |
488 | |
489 | for (size_t __i = 0; __i < __n; ++__i) |
490 | __os << __x._M_x[__i] << __space; |
491 | __os << __x._M_p; |
492 | |
493 | __os.flags(__flags); |
494 | __os.fill(__fill); |
495 | return __os; |
496 | } |
497 | |
498 | template<typename _UIntType, size_t __w, |
499 | size_t __n, size_t __m, size_t __r, |
500 | _UIntType __a, size_t __u, _UIntType __d, size_t __s, |
501 | _UIntType __b, size_t __t, _UIntType __c, size_t __l, |
502 | _UIntType __f, typename _CharT, typename _Traits> |
503 | std::basic_istream<_CharT, _Traits>& |
504 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
505 | mersenne_twister_engine<_UIntType, __w, __n, __m, |
506 | __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>& __x) |
507 | { |
508 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
509 | |
510 | const typename __ios_base::fmtflags __flags = __is.flags(); |
511 | __is.flags(__ios_base::dec | __ios_base::skipws); |
512 | |
513 | for (size_t __i = 0; __i < __n; ++__i) |
514 | __is >> __x._M_x[__i]; |
515 | __is >> __x._M_p; |
516 | |
517 | __is.flags(__flags); |
518 | return __is; |
519 | } |
520 | |
521 | #if ! __cpp_inline_variables |
522 | template<typename _UIntType, size_t __w, size_t __s, size_t __r> |
523 | constexpr size_t |
524 | subtract_with_carry_engine<_UIntType, __w, __s, __r>::word_size; |
525 | |
526 | template<typename _UIntType, size_t __w, size_t __s, size_t __r> |
527 | constexpr size_t |
528 | subtract_with_carry_engine<_UIntType, __w, __s, __r>::short_lag; |
529 | |
530 | template<typename _UIntType, size_t __w, size_t __s, size_t __r> |
531 | constexpr size_t |
532 | subtract_with_carry_engine<_UIntType, __w, __s, __r>::long_lag; |
533 | |
534 | template<typename _UIntType, size_t __w, size_t __s, size_t __r> |
535 | constexpr uint_least32_t |
536 | subtract_with_carry_engine<_UIntType, __w, __s, __r>::default_seed; |
537 | #endif |
538 | |
539 | template<typename _UIntType, size_t __w, size_t __s, size_t __r> |
540 | void |
541 | subtract_with_carry_engine<_UIntType, __w, __s, __r>:: |
542 | seed(result_type __value) |
543 | { |
544 | std::linear_congruential_engine<uint_least32_t, 40014u, 0u, 2147483563u> |
545 | __lcg(__value == 0u ? default_seed : __value); |
546 | |
547 | const size_t __n = (__w + 31) / 32; |
548 | |
549 | for (size_t __i = 0; __i < long_lag; ++__i) |
550 | { |
551 | _UIntType __sum = 0u; |
552 | _UIntType __factor = 1u; |
553 | for (size_t __j = 0; __j < __n; ++__j) |
554 | { |
555 | __sum += __detail::__mod<uint_least32_t, |
556 | __detail::_Shift<uint_least32_t, 32>::__value> |
557 | (x: __lcg()) * __factor; |
558 | __factor *= __detail::_Shift<_UIntType, 32>::__value; |
559 | } |
560 | _M_x[__i] = __detail::__mod<_UIntType, |
561 | __detail::_Shift<_UIntType, __w>::__value>(__sum); |
562 | } |
563 | _M_carry = (_M_x[long_lag - 1] == 0) ? 1 : 0; |
564 | _M_p = 0; |
565 | } |
566 | |
567 | template<typename _UIntType, size_t __w, size_t __s, size_t __r> |
568 | template<typename _Sseq> |
569 | auto |
570 | subtract_with_carry_engine<_UIntType, __w, __s, __r>:: |
571 | seed(_Sseq& __q) |
572 | -> _If_seed_seq<_Sseq> |
573 | { |
574 | const size_t __k = (__w + 31) / 32; |
575 | uint_least32_t __arr[__r * __k]; |
576 | __q.generate(__arr + 0, __arr + __r * __k); |
577 | |
578 | for (size_t __i = 0; __i < long_lag; ++__i) |
579 | { |
580 | _UIntType __sum = 0u; |
581 | _UIntType __factor = 1u; |
582 | for (size_t __j = 0; __j < __k; ++__j) |
583 | { |
584 | __sum += __arr[__k * __i + __j] * __factor; |
585 | __factor *= __detail::_Shift<_UIntType, 32>::__value; |
586 | } |
587 | _M_x[__i] = __detail::__mod<_UIntType, |
588 | __detail::_Shift<_UIntType, __w>::__value>(__sum); |
589 | } |
590 | _M_carry = (_M_x[long_lag - 1] == 0) ? 1 : 0; |
591 | _M_p = 0; |
592 | } |
593 | |
594 | template<typename _UIntType, size_t __w, size_t __s, size_t __r> |
595 | typename subtract_with_carry_engine<_UIntType, __w, __s, __r>:: |
596 | result_type |
597 | subtract_with_carry_engine<_UIntType, __w, __s, __r>:: |
598 | operator()() |
599 | { |
600 | // Derive short lag index from current index. |
601 | long __ps = _M_p - short_lag; |
602 | if (__ps < 0) |
603 | __ps += long_lag; |
604 | |
605 | // Calculate new x(i) without overflow or division. |
606 | // NB: Thanks to the requirements for _UIntType, _M_x[_M_p] + _M_carry |
607 | // cannot overflow. |
608 | _UIntType __xi; |
609 | if (_M_x[__ps] >= _M_x[_M_p] + _M_carry) |
610 | { |
611 | __xi = _M_x[__ps] - _M_x[_M_p] - _M_carry; |
612 | _M_carry = 0; |
613 | } |
614 | else |
615 | { |
616 | __xi = (__detail::_Shift<_UIntType, __w>::__value |
617 | - _M_x[_M_p] - _M_carry + _M_x[__ps]); |
618 | _M_carry = 1; |
619 | } |
620 | _M_x[_M_p] = __xi; |
621 | |
622 | // Adjust current index to loop around in ring buffer. |
623 | if (++_M_p >= long_lag) |
624 | _M_p = 0; |
625 | |
626 | return __xi; |
627 | } |
628 | |
629 | template<typename _UIntType, size_t __w, size_t __s, size_t __r, |
630 | typename _CharT, typename _Traits> |
631 | std::basic_ostream<_CharT, _Traits>& |
632 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
633 | const subtract_with_carry_engine<_UIntType, |
634 | __w, __s, __r>& __x) |
635 | { |
636 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
637 | |
638 | const typename __ios_base::fmtflags __flags = __os.flags(); |
639 | const _CharT __fill = __os.fill(); |
640 | const _CharT __space = __os.widen(' '); |
641 | __os.flags(__ios_base::dec | __ios_base::fixed | __ios_base::left); |
642 | __os.fill(__space); |
643 | |
644 | for (size_t __i = 0; __i < __r; ++__i) |
645 | __os << __x._M_x[__i] << __space; |
646 | __os << __x._M_carry << __space << __x._M_p; |
647 | |
648 | __os.flags(__flags); |
649 | __os.fill(__fill); |
650 | return __os; |
651 | } |
652 | |
653 | template<typename _UIntType, size_t __w, size_t __s, size_t __r, |
654 | typename _CharT, typename _Traits> |
655 | std::basic_istream<_CharT, _Traits>& |
656 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
657 | subtract_with_carry_engine<_UIntType, __w, __s, __r>& __x) |
658 | { |
659 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
660 | |
661 | const typename __ios_base::fmtflags __flags = __is.flags(); |
662 | __is.flags(__ios_base::dec | __ios_base::skipws); |
663 | |
664 | for (size_t __i = 0; __i < __r; ++__i) |
665 | __is >> __x._M_x[__i]; |
666 | __is >> __x._M_carry; |
667 | __is >> __x._M_p; |
668 | |
669 | __is.flags(__flags); |
670 | return __is; |
671 | } |
672 | |
673 | #if ! __cpp_inline_variables |
674 | template<typename _RandomNumberEngine, size_t __p, size_t __r> |
675 | constexpr size_t |
676 | discard_block_engine<_RandomNumberEngine, __p, __r>::block_size; |
677 | |
678 | template<typename _RandomNumberEngine, size_t __p, size_t __r> |
679 | constexpr size_t |
680 | discard_block_engine<_RandomNumberEngine, __p, __r>::used_block; |
681 | #endif |
682 | |
683 | template<typename _RandomNumberEngine, size_t __p, size_t __r> |
684 | typename discard_block_engine<_RandomNumberEngine, |
685 | __p, __r>::result_type |
686 | discard_block_engine<_RandomNumberEngine, __p, __r>:: |
687 | operator()() |
688 | { |
689 | if (_M_n >= used_block) |
690 | { |
691 | _M_b.discard(block_size - _M_n); |
692 | _M_n = 0; |
693 | } |
694 | ++_M_n; |
695 | return _M_b(); |
696 | } |
697 | |
698 | template<typename _RandomNumberEngine, size_t __p, size_t __r, |
699 | typename _CharT, typename _Traits> |
700 | std::basic_ostream<_CharT, _Traits>& |
701 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
702 | const discard_block_engine<_RandomNumberEngine, |
703 | __p, __r>& __x) |
704 | { |
705 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
706 | |
707 | const typename __ios_base::fmtflags __flags = __os.flags(); |
708 | const _CharT __fill = __os.fill(); |
709 | const _CharT __space = __os.widen(' '); |
710 | __os.flags(__ios_base::dec | __ios_base::fixed | __ios_base::left); |
711 | __os.fill(__space); |
712 | |
713 | __os << __x.base() << __space << __x._M_n; |
714 | |
715 | __os.flags(__flags); |
716 | __os.fill(__fill); |
717 | return __os; |
718 | } |
719 | |
720 | template<typename _RandomNumberEngine, size_t __p, size_t __r, |
721 | typename _CharT, typename _Traits> |
722 | std::basic_istream<_CharT, _Traits>& |
723 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
724 | discard_block_engine<_RandomNumberEngine, __p, __r>& __x) |
725 | { |
726 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
727 | |
728 | const typename __ios_base::fmtflags __flags = __is.flags(); |
729 | __is.flags(__ios_base::dec | __ios_base::skipws); |
730 | |
731 | __is >> __x._M_b >> __x._M_n; |
732 | |
733 | __is.flags(__flags); |
734 | return __is; |
735 | } |
736 | |
737 | |
738 | template<typename _RandomNumberEngine, size_t __w, typename _UIntType> |
739 | typename independent_bits_engine<_RandomNumberEngine, __w, _UIntType>:: |
740 | result_type |
741 | independent_bits_engine<_RandomNumberEngine, __w, _UIntType>:: |
742 | operator()() |
743 | { |
744 | typedef typename _RandomNumberEngine::result_type _Eresult_type; |
745 | const _Eresult_type __r |
746 | = (_M_b.max() - _M_b.min() < std::numeric_limits<_Eresult_type>::max() |
747 | ? _M_b.max() - _M_b.min() + 1 : 0); |
748 | const unsigned __edig = std::numeric_limits<_Eresult_type>::digits; |
749 | const unsigned __m = __r ? std::__lg(__r) : __edig; |
750 | |
751 | typedef typename std::common_type<_Eresult_type, result_type>::type |
752 | __ctype; |
753 | const unsigned __cdig = std::numeric_limits<__ctype>::digits; |
754 | |
755 | unsigned __n, __n0; |
756 | __ctype __s0, __s1, __y0, __y1; |
757 | |
758 | for (size_t __i = 0; __i < 2; ++__i) |
759 | { |
760 | __n = (__w + __m - 1) / __m + __i; |
761 | __n0 = __n - __w % __n; |
762 | const unsigned __w0 = __w / __n; // __w0 <= __m |
763 | |
764 | __s0 = 0; |
765 | __s1 = 0; |
766 | if (__w0 < __cdig) |
767 | { |
768 | __s0 = __ctype(1) << __w0; |
769 | __s1 = __s0 << 1; |
770 | } |
771 | |
772 | __y0 = 0; |
773 | __y1 = 0; |
774 | if (__r) |
775 | { |
776 | __y0 = __s0 * (__r / __s0); |
777 | if (__s1) |
778 | __y1 = __s1 * (__r / __s1); |
779 | |
780 | if (__r - __y0 <= __y0 / __n) |
781 | break; |
782 | } |
783 | else |
784 | break; |
785 | } |
786 | |
787 | result_type __sum = 0; |
788 | for (size_t __k = 0; __k < __n0; ++__k) |
789 | { |
790 | __ctype __u; |
791 | do |
792 | __u = _M_b() - _M_b.min(); |
793 | while (__y0 && __u >= __y0); |
794 | __sum = __s0 * __sum + (__s0 ? __u % __s0 : __u); |
795 | } |
796 | for (size_t __k = __n0; __k < __n; ++__k) |
797 | { |
798 | __ctype __u; |
799 | do |
800 | __u = _M_b() - _M_b.min(); |
801 | while (__y1 && __u >= __y1); |
802 | __sum = __s1 * __sum + (__s1 ? __u % __s1 : __u); |
803 | } |
804 | return __sum; |
805 | } |
806 | |
807 | #if ! __cpp_inline_variables |
808 | template<typename _RandomNumberEngine, size_t __k> |
809 | constexpr size_t |
810 | shuffle_order_engine<_RandomNumberEngine, __k>::table_size; |
811 | #endif |
812 | |
813 | namespace __detail |
814 | { |
815 | // Determine whether an integer is representable as double. |
816 | template<typename _Tp> |
817 | constexpr bool |
818 | __representable_as_double(_Tp __x) noexcept |
819 | { |
820 | static_assert(numeric_limits<_Tp>::is_integer, "" ); |
821 | static_assert(!numeric_limits<_Tp>::is_signed, "" ); |
822 | // All integers <= 2^53 are representable. |
823 | return (__x <= (1ull << __DBL_MANT_DIG__)) |
824 | // Between 2^53 and 2^54 only even numbers are representable. |
825 | || (!(__x & 1) && __detail::__representable_as_double(__x >> 1)); |
826 | } |
827 | |
828 | // Determine whether x+1 is representable as double. |
829 | template<typename _Tp> |
830 | constexpr bool |
831 | __p1_representable_as_double(_Tp __x) noexcept |
832 | { |
833 | static_assert(numeric_limits<_Tp>::is_integer, "" ); |
834 | static_assert(!numeric_limits<_Tp>::is_signed, "" ); |
835 | return numeric_limits<_Tp>::digits < __DBL_MANT_DIG__ |
836 | || (bool(__x + 1u) // return false if x+1 wraps around to zero |
837 | && __detail::__representable_as_double(__x + 1u)); |
838 | } |
839 | } |
840 | |
841 | template<typename _RandomNumberEngine, size_t __k> |
842 | typename shuffle_order_engine<_RandomNumberEngine, __k>::result_type |
843 | shuffle_order_engine<_RandomNumberEngine, __k>:: |
844 | operator()() |
845 | { |
846 | constexpr result_type __range = max() - min(); |
847 | size_t __j = __k; |
848 | const result_type __y = _M_y - min(); |
849 | // Avoid using slower long double arithmetic if possible. |
850 | if _GLIBCXX17_CONSTEXPR (__detail::__p1_representable_as_double(__range)) |
851 | __j *= __y / (__range + 1.0); |
852 | else |
853 | __j *= __y / (__range + 1.0L); |
854 | _M_y = _M_v[__j]; |
855 | _M_v[__j] = _M_b(); |
856 | |
857 | return _M_y; |
858 | } |
859 | |
860 | template<typename _RandomNumberEngine, size_t __k, |
861 | typename _CharT, typename _Traits> |
862 | std::basic_ostream<_CharT, _Traits>& |
863 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
864 | const shuffle_order_engine<_RandomNumberEngine, __k>& __x) |
865 | { |
866 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
867 | |
868 | const typename __ios_base::fmtflags __flags = __os.flags(); |
869 | const _CharT __fill = __os.fill(); |
870 | const _CharT __space = __os.widen(' '); |
871 | __os.flags(__ios_base::dec | __ios_base::fixed | __ios_base::left); |
872 | __os.fill(__space); |
873 | |
874 | __os << __x.base(); |
875 | for (size_t __i = 0; __i < __k; ++__i) |
876 | __os << __space << __x._M_v[__i]; |
877 | __os << __space << __x._M_y; |
878 | |
879 | __os.flags(__flags); |
880 | __os.fill(__fill); |
881 | return __os; |
882 | } |
883 | |
884 | template<typename _RandomNumberEngine, size_t __k, |
885 | typename _CharT, typename _Traits> |
886 | std::basic_istream<_CharT, _Traits>& |
887 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
888 | shuffle_order_engine<_RandomNumberEngine, __k>& __x) |
889 | { |
890 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
891 | |
892 | const typename __ios_base::fmtflags __flags = __is.flags(); |
893 | __is.flags(__ios_base::dec | __ios_base::skipws); |
894 | |
895 | __is >> __x._M_b; |
896 | for (size_t __i = 0; __i < __k; ++__i) |
897 | __is >> __x._M_v[__i]; |
898 | __is >> __x._M_y; |
899 | |
900 | __is.flags(__flags); |
901 | return __is; |
902 | } |
903 | |
904 | |
905 | template<typename _IntType, typename _CharT, typename _Traits> |
906 | std::basic_ostream<_CharT, _Traits>& |
907 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
908 | const uniform_int_distribution<_IntType>& __x) |
909 | { |
910 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
911 | |
912 | const typename __ios_base::fmtflags __flags = __os.flags(); |
913 | const _CharT __fill = __os.fill(); |
914 | const _CharT __space = __os.widen(' '); |
915 | __os.flags(__ios_base::scientific | __ios_base::left); |
916 | __os.fill(__space); |
917 | |
918 | __os << __x.a() << __space << __x.b(); |
919 | |
920 | __os.flags(__flags); |
921 | __os.fill(__fill); |
922 | return __os; |
923 | } |
924 | |
925 | template<typename _IntType, typename _CharT, typename _Traits> |
926 | std::basic_istream<_CharT, _Traits>& |
927 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
928 | uniform_int_distribution<_IntType>& __x) |
929 | { |
930 | using param_type |
931 | = typename uniform_int_distribution<_IntType>::param_type; |
932 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
933 | |
934 | const typename __ios_base::fmtflags __flags = __is.flags(); |
935 | __is.flags(__ios_base::dec | __ios_base::skipws); |
936 | |
937 | _IntType __a, __b; |
938 | if (__is >> __a >> __b) |
939 | __x.param(param_type(__a, __b)); |
940 | |
941 | __is.flags(__flags); |
942 | return __is; |
943 | } |
944 | |
945 | |
946 | template<typename _RealType> |
947 | template<typename _ForwardIterator, |
948 | typename _UniformRandomNumberGenerator> |
949 | void |
950 | uniform_real_distribution<_RealType>:: |
951 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
952 | _UniformRandomNumberGenerator& __urng, |
953 | const param_type& __p) |
954 | { |
955 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
956 | __detail::_Adaptor<_UniformRandomNumberGenerator, result_type> |
957 | __aurng(__urng); |
958 | auto __range = __p.b() - __p.a(); |
959 | while (__f != __t) |
960 | *__f++ = __aurng() * __range + __p.a(); |
961 | } |
962 | |
963 | template<typename _RealType, typename _CharT, typename _Traits> |
964 | std::basic_ostream<_CharT, _Traits>& |
965 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
966 | const uniform_real_distribution<_RealType>& __x) |
967 | { |
968 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
969 | |
970 | const typename __ios_base::fmtflags __flags = __os.flags(); |
971 | const _CharT __fill = __os.fill(); |
972 | const std::streamsize __precision = __os.precision(); |
973 | const _CharT __space = __os.widen(' '); |
974 | __os.flags(__ios_base::scientific | __ios_base::left); |
975 | __os.fill(__space); |
976 | __os.precision(std::numeric_limits<_RealType>::max_digits10); |
977 | |
978 | __os << __x.a() << __space << __x.b(); |
979 | |
980 | __os.flags(__flags); |
981 | __os.fill(__fill); |
982 | __os.precision(__precision); |
983 | return __os; |
984 | } |
985 | |
986 | template<typename _RealType, typename _CharT, typename _Traits> |
987 | std::basic_istream<_CharT, _Traits>& |
988 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
989 | uniform_real_distribution<_RealType>& __x) |
990 | { |
991 | using param_type |
992 | = typename uniform_real_distribution<_RealType>::param_type; |
993 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
994 | |
995 | const typename __ios_base::fmtflags __flags = __is.flags(); |
996 | __is.flags(__ios_base::skipws); |
997 | |
998 | _RealType __a, __b; |
999 | if (__is >> __a >> __b) |
1000 | __x.param(param_type(__a, __b)); |
1001 | |
1002 | __is.flags(__flags); |
1003 | return __is; |
1004 | } |
1005 | |
1006 | |
1007 | template<typename _ForwardIterator, |
1008 | typename _UniformRandomNumberGenerator> |
1009 | void |
1010 | std::bernoulli_distribution:: |
1011 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
1012 | _UniformRandomNumberGenerator& __urng, |
1013 | const param_type& __p) |
1014 | { |
1015 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
1016 | __detail::_Adaptor<_UniformRandomNumberGenerator, double> |
1017 | __aurng(__urng); |
1018 | auto __limit = __p.p() * (__aurng.max() - __aurng.min()); |
1019 | |
1020 | while (__f != __t) |
1021 | *__f++ = (__aurng() - __aurng.min()) < __limit; |
1022 | } |
1023 | |
1024 | template<typename _CharT, typename _Traits> |
1025 | std::basic_ostream<_CharT, _Traits>& |
1026 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
1027 | const bernoulli_distribution& __x) |
1028 | { |
1029 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
1030 | |
1031 | const typename __ios_base::fmtflags __flags = __os.flags(); |
1032 | const _CharT __fill = __os.fill(); |
1033 | const std::streamsize __precision = __os.precision(); |
1034 | __os.flags(__ios_base::scientific | __ios_base::left); |
1035 | __os.fill(__os.widen(' ')); |
1036 | __os.precision(std::numeric_limits<double>::max_digits10); |
1037 | |
1038 | __os << __x.p(); |
1039 | |
1040 | __os.flags(__flags); |
1041 | __os.fill(__fill); |
1042 | __os.precision(__precision); |
1043 | return __os; |
1044 | } |
1045 | |
1046 | |
1047 | template<typename _IntType> |
1048 | template<typename _UniformRandomNumberGenerator> |
1049 | typename geometric_distribution<_IntType>::result_type |
1050 | geometric_distribution<_IntType>:: |
1051 | operator()(_UniformRandomNumberGenerator& __urng, |
1052 | const param_type& __param) |
1053 | { |
1054 | // About the epsilon thing see this thread: |
1055 | // http://gcc.gnu.org/ml/gcc-patches/2006-10/msg00971.html |
1056 | const double __naf = |
1057 | (1 - std::numeric_limits<double>::epsilon()) / 2; |
1058 | // The largest _RealType convertible to _IntType. |
1059 | const double __thr = |
1060 | std::numeric_limits<_IntType>::max() + __naf; |
1061 | __detail::_Adaptor<_UniformRandomNumberGenerator, double> |
1062 | __aurng(__urng); |
1063 | |
1064 | double __cand; |
1065 | do |
1066 | __cand = std::floor(std::log(1.0 - __aurng()) / __param._M_log_1_p); |
1067 | while (__cand >= __thr); |
1068 | |
1069 | return result_type(__cand + __naf); |
1070 | } |
1071 | |
1072 | template<typename _IntType> |
1073 | template<typename _ForwardIterator, |
1074 | typename _UniformRandomNumberGenerator> |
1075 | void |
1076 | geometric_distribution<_IntType>:: |
1077 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
1078 | _UniformRandomNumberGenerator& __urng, |
1079 | const param_type& __param) |
1080 | { |
1081 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
1082 | // About the epsilon thing see this thread: |
1083 | // http://gcc.gnu.org/ml/gcc-patches/2006-10/msg00971.html |
1084 | const double __naf = |
1085 | (1 - std::numeric_limits<double>::epsilon()) / 2; |
1086 | // The largest _RealType convertible to _IntType. |
1087 | const double __thr = |
1088 | std::numeric_limits<_IntType>::max() + __naf; |
1089 | __detail::_Adaptor<_UniformRandomNumberGenerator, double> |
1090 | __aurng(__urng); |
1091 | |
1092 | while (__f != __t) |
1093 | { |
1094 | double __cand; |
1095 | do |
1096 | __cand = std::floor(std::log(1.0 - __aurng()) |
1097 | / __param._M_log_1_p); |
1098 | while (__cand >= __thr); |
1099 | |
1100 | *__f++ = __cand + __naf; |
1101 | } |
1102 | } |
1103 | |
1104 | template<typename _IntType, |
1105 | typename _CharT, typename _Traits> |
1106 | std::basic_ostream<_CharT, _Traits>& |
1107 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
1108 | const geometric_distribution<_IntType>& __x) |
1109 | { |
1110 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
1111 | |
1112 | const typename __ios_base::fmtflags __flags = __os.flags(); |
1113 | const _CharT __fill = __os.fill(); |
1114 | const std::streamsize __precision = __os.precision(); |
1115 | __os.flags(__ios_base::scientific | __ios_base::left); |
1116 | __os.fill(__os.widen(' ')); |
1117 | __os.precision(std::numeric_limits<double>::max_digits10); |
1118 | |
1119 | __os << __x.p(); |
1120 | |
1121 | __os.flags(__flags); |
1122 | __os.fill(__fill); |
1123 | __os.precision(__precision); |
1124 | return __os; |
1125 | } |
1126 | |
1127 | template<typename _IntType, |
1128 | typename _CharT, typename _Traits> |
1129 | std::basic_istream<_CharT, _Traits>& |
1130 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
1131 | geometric_distribution<_IntType>& __x) |
1132 | { |
1133 | using param_type = typename geometric_distribution<_IntType>::param_type; |
1134 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
1135 | |
1136 | const typename __ios_base::fmtflags __flags = __is.flags(); |
1137 | __is.flags(__ios_base::skipws); |
1138 | |
1139 | double __p; |
1140 | if (__is >> __p) |
1141 | __x.param(param_type(__p)); |
1142 | |
1143 | __is.flags(__flags); |
1144 | return __is; |
1145 | } |
1146 | |
1147 | // This is Leger's algorithm, also in Devroye, Ch. X, Example 1.5. |
1148 | template<typename _IntType> |
1149 | template<typename _UniformRandomNumberGenerator> |
1150 | typename negative_binomial_distribution<_IntType>::result_type |
1151 | negative_binomial_distribution<_IntType>:: |
1152 | operator()(_UniformRandomNumberGenerator& __urng) |
1153 | { |
1154 | const double __y = _M_gd(__urng); |
1155 | |
1156 | // XXX Is the constructor too slow? |
1157 | std::poisson_distribution<result_type> __poisson(__y); |
1158 | return __poisson(__urng); |
1159 | } |
1160 | |
1161 | template<typename _IntType> |
1162 | template<typename _UniformRandomNumberGenerator> |
1163 | typename negative_binomial_distribution<_IntType>::result_type |
1164 | negative_binomial_distribution<_IntType>:: |
1165 | operator()(_UniformRandomNumberGenerator& __urng, |
1166 | const param_type& __p) |
1167 | { |
1168 | typedef typename std::gamma_distribution<double>::param_type |
1169 | param_type; |
1170 | |
1171 | const double __y = |
1172 | _M_gd(__urng, param_type(__p.k(), (1.0 - __p.p()) / __p.p())); |
1173 | |
1174 | std::poisson_distribution<result_type> __poisson(__y); |
1175 | return __poisson(__urng); |
1176 | } |
1177 | |
1178 | template<typename _IntType> |
1179 | template<typename _ForwardIterator, |
1180 | typename _UniformRandomNumberGenerator> |
1181 | void |
1182 | negative_binomial_distribution<_IntType>:: |
1183 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
1184 | _UniformRandomNumberGenerator& __urng) |
1185 | { |
1186 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
1187 | while (__f != __t) |
1188 | { |
1189 | const double __y = _M_gd(__urng); |
1190 | |
1191 | // XXX Is the constructor too slow? |
1192 | std::poisson_distribution<result_type> __poisson(__y); |
1193 | *__f++ = __poisson(__urng); |
1194 | } |
1195 | } |
1196 | |
1197 | template<typename _IntType> |
1198 | template<typename _ForwardIterator, |
1199 | typename _UniformRandomNumberGenerator> |
1200 | void |
1201 | negative_binomial_distribution<_IntType>:: |
1202 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
1203 | _UniformRandomNumberGenerator& __urng, |
1204 | const param_type& __p) |
1205 | { |
1206 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
1207 | typename std::gamma_distribution<result_type>::param_type |
1208 | __p2(__p.k(), (1.0 - __p.p()) / __p.p()); |
1209 | |
1210 | while (__f != __t) |
1211 | { |
1212 | const double __y = _M_gd(__urng, __p2); |
1213 | |
1214 | std::poisson_distribution<result_type> __poisson(__y); |
1215 | *__f++ = __poisson(__urng); |
1216 | } |
1217 | } |
1218 | |
1219 | template<typename _IntType, typename _CharT, typename _Traits> |
1220 | std::basic_ostream<_CharT, _Traits>& |
1221 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
1222 | const negative_binomial_distribution<_IntType>& __x) |
1223 | { |
1224 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
1225 | |
1226 | const typename __ios_base::fmtflags __flags = __os.flags(); |
1227 | const _CharT __fill = __os.fill(); |
1228 | const std::streamsize __precision = __os.precision(); |
1229 | const _CharT __space = __os.widen(' '); |
1230 | __os.flags(__ios_base::scientific | __ios_base::left); |
1231 | __os.fill(__os.widen(' ')); |
1232 | __os.precision(std::numeric_limits<double>::max_digits10); |
1233 | |
1234 | __os << __x.k() << __space << __x.p() |
1235 | << __space << __x._M_gd; |
1236 | |
1237 | __os.flags(__flags); |
1238 | __os.fill(__fill); |
1239 | __os.precision(__precision); |
1240 | return __os; |
1241 | } |
1242 | |
1243 | template<typename _IntType, typename _CharT, typename _Traits> |
1244 | std::basic_istream<_CharT, _Traits>& |
1245 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
1246 | negative_binomial_distribution<_IntType>& __x) |
1247 | { |
1248 | using param_type |
1249 | = typename negative_binomial_distribution<_IntType>::param_type; |
1250 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
1251 | |
1252 | const typename __ios_base::fmtflags __flags = __is.flags(); |
1253 | __is.flags(__ios_base::skipws); |
1254 | |
1255 | _IntType __k; |
1256 | double __p; |
1257 | if (__is >> __k >> __p >> __x._M_gd) |
1258 | __x.param(param_type(__k, __p)); |
1259 | |
1260 | __is.flags(__flags); |
1261 | return __is; |
1262 | } |
1263 | |
1264 | |
1265 | template<typename _IntType> |
1266 | void |
1267 | poisson_distribution<_IntType>::param_type:: |
1268 | _M_initialize() |
1269 | { |
1270 | #if _GLIBCXX_USE_C99_MATH_TR1 |
1271 | if (_M_mean >= 12) |
1272 | { |
1273 | const double __m = std::floor(x: _M_mean); |
1274 | _M_lm_thr = std::log(x: _M_mean); |
1275 | _M_lfm = std::lgamma(__m + 1); |
1276 | _M_sm = std::sqrt(x: __m); |
1277 | |
1278 | const double __pi_4 = 0.7853981633974483096156608458198757L; |
1279 | const double __dx = std::sqrt(x: 2 * __m * std::log(x: 32 * __m |
1280 | / __pi_4)); |
1281 | _M_d = std::round(x: std::max<double>(a: 6.0, b: std::min(a: __m, b: __dx))); |
1282 | const double __cx = 2 * __m + _M_d; |
1283 | _M_scx = std::sqrt(x: __cx / 2); |
1284 | _M_1cx = 1 / __cx; |
1285 | |
1286 | _M_c2b = std::sqrt(x: __pi_4 * __cx) * std::exp(x: _M_1cx); |
1287 | _M_cb = 2 * __cx * std::exp(x: -_M_d * _M_1cx * (1 + _M_d / 2)) |
1288 | / _M_d; |
1289 | } |
1290 | else |
1291 | #endif |
1292 | _M_lm_thr = std::exp(x: -_M_mean); |
1293 | } |
1294 | |
1295 | /** |
1296 | * A rejection algorithm when mean >= 12 and a simple method based |
1297 | * upon the multiplication of uniform random variates otherwise. |
1298 | * NB: The former is available only if _GLIBCXX_USE_C99_MATH_TR1 |
1299 | * is defined. |
1300 | * |
1301 | * Reference: |
1302 | * Devroye, L. Non-Uniform Random Variates Generation. Springer-Verlag, |
1303 | * New York, 1986, Ch. X, Sects. 3.3 & 3.4 (+ Errata!). |
1304 | */ |
1305 | template<typename _IntType> |
1306 | template<typename _UniformRandomNumberGenerator> |
1307 | typename poisson_distribution<_IntType>::result_type |
1308 | poisson_distribution<_IntType>:: |
1309 | operator()(_UniformRandomNumberGenerator& __urng, |
1310 | const param_type& __param) |
1311 | { |
1312 | __detail::_Adaptor<_UniformRandomNumberGenerator, double> |
1313 | __aurng(__urng); |
1314 | #if _GLIBCXX_USE_C99_MATH_TR1 |
1315 | if (__param.mean() >= 12) |
1316 | { |
1317 | double __x; |
1318 | |
1319 | // See comments above... |
1320 | const double __naf = |
1321 | (1 - std::numeric_limits<double>::epsilon()) / 2; |
1322 | const double __thr = |
1323 | std::numeric_limits<_IntType>::max() + __naf; |
1324 | |
1325 | const double __m = std::floor(__param.mean()); |
1326 | // sqrt(pi / 2) |
1327 | const double __spi_2 = 1.2533141373155002512078826424055226L; |
1328 | const double __c1 = __param._M_sm * __spi_2; |
1329 | const double __c2 = __param._M_c2b + __c1; |
1330 | const double __c3 = __c2 + 1; |
1331 | const double __c4 = __c3 + 1; |
1332 | // 1 / 78 |
1333 | const double __178 = 0.0128205128205128205128205128205128L; |
1334 | // e^(1 / 78) |
1335 | const double __e178 = 1.0129030479320018583185514777512983L; |
1336 | const double __c5 = __c4 + __e178; |
1337 | const double __c = __param._M_cb + __c5; |
1338 | const double __2cx = 2 * (2 * __m + __param._M_d); |
1339 | |
1340 | bool __reject = true; |
1341 | do |
1342 | { |
1343 | const double __u = __c * __aurng(); |
1344 | const double __e = -std::log(1.0 - __aurng()); |
1345 | |
1346 | double __w = 0.0; |
1347 | |
1348 | if (__u <= __c1) |
1349 | { |
1350 | const double __n = _M_nd(__urng); |
1351 | const double __y = -std::abs(x: __n) * __param._M_sm - 1; |
1352 | __x = std::floor(x: __y); |
1353 | __w = -__n * __n / 2; |
1354 | if (__x < -__m) |
1355 | continue; |
1356 | } |
1357 | else if (__u <= __c2) |
1358 | { |
1359 | const double __n = _M_nd(__urng); |
1360 | const double __y = 1 + std::abs(x: __n) * __param._M_scx; |
1361 | __x = std::ceil(x: __y); |
1362 | __w = __y * (2 - __y) * __param._M_1cx; |
1363 | if (__x > __param._M_d) |
1364 | continue; |
1365 | } |
1366 | else if (__u <= __c3) |
1367 | // NB: This case not in the book, nor in the Errata, |
1368 | // but should be ok... |
1369 | __x = -1; |
1370 | else if (__u <= __c4) |
1371 | __x = 0; |
1372 | else if (__u <= __c5) |
1373 | { |
1374 | __x = 1; |
1375 | // Only in the Errata, see libstdc++/83237. |
1376 | __w = __178; |
1377 | } |
1378 | else |
1379 | { |
1380 | const double __v = -std::log(1.0 - __aurng()); |
1381 | const double __y = __param._M_d |
1382 | + __v * __2cx / __param._M_d; |
1383 | __x = std::ceil(x: __y); |
1384 | __w = -__param._M_d * __param._M_1cx * (1 + __y / 2); |
1385 | } |
1386 | |
1387 | __reject = (__w - __e - __x * __param._M_lm_thr |
1388 | > __param._M_lfm - std::lgamma(__x + __m + 1)); |
1389 | |
1390 | __reject |= __x + __m >= __thr; |
1391 | |
1392 | } while (__reject); |
1393 | |
1394 | return result_type(__x + __m + __naf); |
1395 | } |
1396 | else |
1397 | #endif |
1398 | { |
1399 | _IntType __x = 0; |
1400 | double __prod = 1.0; |
1401 | |
1402 | do |
1403 | { |
1404 | __prod *= __aurng(); |
1405 | __x += 1; |
1406 | } |
1407 | while (__prod > __param._M_lm_thr); |
1408 | |
1409 | return __x - 1; |
1410 | } |
1411 | } |
1412 | |
1413 | template<typename _IntType> |
1414 | template<typename _ForwardIterator, |
1415 | typename _UniformRandomNumberGenerator> |
1416 | void |
1417 | poisson_distribution<_IntType>:: |
1418 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
1419 | _UniformRandomNumberGenerator& __urng, |
1420 | const param_type& __param) |
1421 | { |
1422 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
1423 | // We could duplicate everything from operator()... |
1424 | while (__f != __t) |
1425 | *__f++ = this->operator()(__urng, __param); |
1426 | } |
1427 | |
1428 | template<typename _IntType, |
1429 | typename _CharT, typename _Traits> |
1430 | std::basic_ostream<_CharT, _Traits>& |
1431 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
1432 | const poisson_distribution<_IntType>& __x) |
1433 | { |
1434 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
1435 | |
1436 | const typename __ios_base::fmtflags __flags = __os.flags(); |
1437 | const _CharT __fill = __os.fill(); |
1438 | const std::streamsize __precision = __os.precision(); |
1439 | const _CharT __space = __os.widen(' '); |
1440 | __os.flags(__ios_base::scientific | __ios_base::left); |
1441 | __os.fill(__space); |
1442 | __os.precision(std::numeric_limits<double>::max_digits10); |
1443 | |
1444 | __os << __x.mean() << __space << __x._M_nd; |
1445 | |
1446 | __os.flags(__flags); |
1447 | __os.fill(__fill); |
1448 | __os.precision(__precision); |
1449 | return __os; |
1450 | } |
1451 | |
1452 | template<typename _IntType, |
1453 | typename _CharT, typename _Traits> |
1454 | std::basic_istream<_CharT, _Traits>& |
1455 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
1456 | poisson_distribution<_IntType>& __x) |
1457 | { |
1458 | using param_type = typename poisson_distribution<_IntType>::param_type; |
1459 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
1460 | |
1461 | const typename __ios_base::fmtflags __flags = __is.flags(); |
1462 | __is.flags(__ios_base::skipws); |
1463 | |
1464 | double __mean; |
1465 | if (__is >> __mean >> __x._M_nd) |
1466 | __x.param(param_type(__mean)); |
1467 | |
1468 | __is.flags(__flags); |
1469 | return __is; |
1470 | } |
1471 | |
1472 | |
1473 | template<typename _IntType> |
1474 | void |
1475 | binomial_distribution<_IntType>::param_type:: |
1476 | _M_initialize() |
1477 | { |
1478 | const double __p12 = _M_p <= 0.5 ? _M_p : 1.0 - _M_p; |
1479 | |
1480 | _M_easy = true; |
1481 | |
1482 | #if _GLIBCXX_USE_C99_MATH_TR1 |
1483 | if (_M_t * __p12 >= 8) |
1484 | { |
1485 | _M_easy = false; |
1486 | const double __np = std::floor(_M_t * __p12); |
1487 | const double __pa = __np / _M_t; |
1488 | const double __1p = 1 - __pa; |
1489 | |
1490 | const double __pi_4 = 0.7853981633974483096156608458198757L; |
1491 | const double __d1x = |
1492 | std::sqrt(x: __np * __1p * std::log(x: 32 * __np |
1493 | / (81 * __pi_4 * __1p))); |
1494 | _M_d1 = std::round(x: std::max<double>(a: 1.0, b: __d1x)); |
1495 | const double __d2x = |
1496 | std::sqrt(__np * __1p * std::log(32 * _M_t * __1p |
1497 | / (__pi_4 * __pa))); |
1498 | _M_d2 = std::round(x: std::max<double>(a: 1.0, b: __d2x)); |
1499 | |
1500 | // sqrt(pi / 2) |
1501 | const double __spi_2 = 1.2533141373155002512078826424055226L; |
1502 | _M_s1 = std::sqrt(x: __np * __1p) * (1 + _M_d1 / (4 * __np)); |
1503 | _M_s2 = std::sqrt(x: __np * __1p) * (1 + _M_d2 / (4 * _M_t * __1p)); |
1504 | _M_c = 2 * _M_d1 / __np; |
1505 | _M_a1 = std::exp(x: _M_c) * _M_s1 * __spi_2; |
1506 | const double __a12 = _M_a1 + _M_s2 * __spi_2; |
1507 | const double __s1s = _M_s1 * _M_s1; |
1508 | _M_a123 = __a12 + (std::exp(_M_d1 / (_M_t * __1p)) |
1509 | * 2 * __s1s / _M_d1 |
1510 | * std::exp(x: -_M_d1 * _M_d1 / (2 * __s1s))); |
1511 | const double __s2s = _M_s2 * _M_s2; |
1512 | _M_s = (_M_a123 + 2 * __s2s / _M_d2 |
1513 | * std::exp(x: -_M_d2 * _M_d2 / (2 * __s2s))); |
1514 | _M_lf = (std::lgamma(__np + 1) |
1515 | + std::lgamma(_M_t - __np + 1)); |
1516 | _M_lp1p = std::log(x: __pa / __1p); |
1517 | |
1518 | _M_q = -std::log(x: 1 - (__p12 - __pa) / __1p); |
1519 | } |
1520 | else |
1521 | #endif |
1522 | _M_q = -std::log(x: 1 - __p12); |
1523 | } |
1524 | |
1525 | template<typename _IntType> |
1526 | template<typename _UniformRandomNumberGenerator> |
1527 | typename binomial_distribution<_IntType>::result_type |
1528 | binomial_distribution<_IntType>:: |
1529 | _M_waiting(_UniformRandomNumberGenerator& __urng, |
1530 | _IntType __t, double __q) |
1531 | { |
1532 | _IntType __x = 0; |
1533 | double __sum = 0.0; |
1534 | __detail::_Adaptor<_UniformRandomNumberGenerator, double> |
1535 | __aurng(__urng); |
1536 | |
1537 | do |
1538 | { |
1539 | if (__t == __x) |
1540 | return __x; |
1541 | const double __e = -std::log(1.0 - __aurng()); |
1542 | __sum += __e / (__t - __x); |
1543 | __x += 1; |
1544 | } |
1545 | while (__sum <= __q); |
1546 | |
1547 | return __x - 1; |
1548 | } |
1549 | |
1550 | /** |
1551 | * A rejection algorithm when t * p >= 8 and a simple waiting time |
1552 | * method - the second in the referenced book - otherwise. |
1553 | * NB: The former is available only if _GLIBCXX_USE_C99_MATH_TR1 |
1554 | * is defined. |
1555 | * |
1556 | * Reference: |
1557 | * Devroye, L. Non-Uniform Random Variates Generation. Springer-Verlag, |
1558 | * New York, 1986, Ch. X, Sect. 4 (+ Errata!). |
1559 | */ |
1560 | template<typename _IntType> |
1561 | template<typename _UniformRandomNumberGenerator> |
1562 | typename binomial_distribution<_IntType>::result_type |
1563 | binomial_distribution<_IntType>:: |
1564 | operator()(_UniformRandomNumberGenerator& __urng, |
1565 | const param_type& __param) |
1566 | { |
1567 | result_type __ret; |
1568 | const _IntType __t = __param.t(); |
1569 | const double __p = __param.p(); |
1570 | const double __p12 = __p <= 0.5 ? __p : 1.0 - __p; |
1571 | __detail::_Adaptor<_UniformRandomNumberGenerator, double> |
1572 | __aurng(__urng); |
1573 | |
1574 | #if _GLIBCXX_USE_C99_MATH_TR1 |
1575 | if (!__param._M_easy) |
1576 | { |
1577 | double __x; |
1578 | |
1579 | // See comments above... |
1580 | const double __naf = |
1581 | (1 - std::numeric_limits<double>::epsilon()) / 2; |
1582 | const double __thr = |
1583 | std::numeric_limits<_IntType>::max() + __naf; |
1584 | |
1585 | const double __np = std::floor(__t * __p12); |
1586 | |
1587 | // sqrt(pi / 2) |
1588 | const double __spi_2 = 1.2533141373155002512078826424055226L; |
1589 | const double __a1 = __param._M_a1; |
1590 | const double __a12 = __a1 + __param._M_s2 * __spi_2; |
1591 | const double __a123 = __param._M_a123; |
1592 | const double __s1s = __param._M_s1 * __param._M_s1; |
1593 | const double __s2s = __param._M_s2 * __param._M_s2; |
1594 | |
1595 | bool __reject; |
1596 | do |
1597 | { |
1598 | const double __u = __param._M_s * __aurng(); |
1599 | |
1600 | double __v; |
1601 | |
1602 | if (__u <= __a1) |
1603 | { |
1604 | const double __n = _M_nd(__urng); |
1605 | const double __y = __param._M_s1 * std::abs(x: __n); |
1606 | __reject = __y >= __param._M_d1; |
1607 | if (!__reject) |
1608 | { |
1609 | const double __e = -std::log(1.0 - __aurng()); |
1610 | __x = std::floor(x: __y); |
1611 | __v = -__e - __n * __n / 2 + __param._M_c; |
1612 | } |
1613 | } |
1614 | else if (__u <= __a12) |
1615 | { |
1616 | const double __n = _M_nd(__urng); |
1617 | const double __y = __param._M_s2 * std::abs(x: __n); |
1618 | __reject = __y >= __param._M_d2; |
1619 | if (!__reject) |
1620 | { |
1621 | const double __e = -std::log(1.0 - __aurng()); |
1622 | __x = std::floor(x: -__y); |
1623 | __v = -__e - __n * __n / 2; |
1624 | } |
1625 | } |
1626 | else if (__u <= __a123) |
1627 | { |
1628 | const double __e1 = -std::log(1.0 - __aurng()); |
1629 | const double __e2 = -std::log(1.0 - __aurng()); |
1630 | |
1631 | const double __y = __param._M_d1 |
1632 | + 2 * __s1s * __e1 / __param._M_d1; |
1633 | __x = std::floor(x: __y); |
1634 | __v = (-__e2 + __param._M_d1 * (1 / (__t - __np) |
1635 | -__y / (2 * __s1s))); |
1636 | __reject = false; |
1637 | } |
1638 | else |
1639 | { |
1640 | const double __e1 = -std::log(1.0 - __aurng()); |
1641 | const double __e2 = -std::log(1.0 - __aurng()); |
1642 | |
1643 | const double __y = __param._M_d2 |
1644 | + 2 * __s2s * __e1 / __param._M_d2; |
1645 | __x = std::floor(x: -__y); |
1646 | __v = -__e2 - __param._M_d2 * __y / (2 * __s2s); |
1647 | __reject = false; |
1648 | } |
1649 | |
1650 | __reject = __reject || __x < -__np || __x > __t - __np; |
1651 | if (!__reject) |
1652 | { |
1653 | const double __lfx = |
1654 | std::lgamma(__np + __x + 1) |
1655 | + std::lgamma(__t - (__np + __x) + 1); |
1656 | __reject = __v > __param._M_lf - __lfx |
1657 | + __x * __param._M_lp1p; |
1658 | } |
1659 | |
1660 | __reject |= __x + __np >= __thr; |
1661 | } |
1662 | while (__reject); |
1663 | |
1664 | __x += __np + __naf; |
1665 | |
1666 | const _IntType __z = _M_waiting(__urng, __t - _IntType(__x), |
1667 | __param._M_q); |
1668 | __ret = _IntType(__x) + __z; |
1669 | } |
1670 | else |
1671 | #endif |
1672 | __ret = _M_waiting(__urng, __t, __param._M_q); |
1673 | |
1674 | if (__p12 != __p) |
1675 | __ret = __t - __ret; |
1676 | return __ret; |
1677 | } |
1678 | |
1679 | template<typename _IntType> |
1680 | template<typename _ForwardIterator, |
1681 | typename _UniformRandomNumberGenerator> |
1682 | void |
1683 | binomial_distribution<_IntType>:: |
1684 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
1685 | _UniformRandomNumberGenerator& __urng, |
1686 | const param_type& __param) |
1687 | { |
1688 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
1689 | // We could duplicate everything from operator()... |
1690 | while (__f != __t) |
1691 | *__f++ = this->operator()(__urng, __param); |
1692 | } |
1693 | |
1694 | template<typename _IntType, |
1695 | typename _CharT, typename _Traits> |
1696 | std::basic_ostream<_CharT, _Traits>& |
1697 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
1698 | const binomial_distribution<_IntType>& __x) |
1699 | { |
1700 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
1701 | |
1702 | const typename __ios_base::fmtflags __flags = __os.flags(); |
1703 | const _CharT __fill = __os.fill(); |
1704 | const std::streamsize __precision = __os.precision(); |
1705 | const _CharT __space = __os.widen(' '); |
1706 | __os.flags(__ios_base::scientific | __ios_base::left); |
1707 | __os.fill(__space); |
1708 | __os.precision(std::numeric_limits<double>::max_digits10); |
1709 | |
1710 | __os << __x.t() << __space << __x.p() |
1711 | << __space << __x._M_nd; |
1712 | |
1713 | __os.flags(__flags); |
1714 | __os.fill(__fill); |
1715 | __os.precision(__precision); |
1716 | return __os; |
1717 | } |
1718 | |
1719 | template<typename _IntType, |
1720 | typename _CharT, typename _Traits> |
1721 | std::basic_istream<_CharT, _Traits>& |
1722 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
1723 | binomial_distribution<_IntType>& __x) |
1724 | { |
1725 | using param_type = typename binomial_distribution<_IntType>::param_type; |
1726 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
1727 | |
1728 | const typename __ios_base::fmtflags __flags = __is.flags(); |
1729 | __is.flags(__ios_base::dec | __ios_base::skipws); |
1730 | |
1731 | _IntType __t; |
1732 | double __p; |
1733 | if (__is >> __t >> __p >> __x._M_nd) |
1734 | __x.param(param_type(__t, __p)); |
1735 | |
1736 | __is.flags(__flags); |
1737 | return __is; |
1738 | } |
1739 | |
1740 | |
1741 | template<typename _RealType> |
1742 | template<typename _ForwardIterator, |
1743 | typename _UniformRandomNumberGenerator> |
1744 | void |
1745 | std::exponential_distribution<_RealType>:: |
1746 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
1747 | _UniformRandomNumberGenerator& __urng, |
1748 | const param_type& __p) |
1749 | { |
1750 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
1751 | __detail::_Adaptor<_UniformRandomNumberGenerator, result_type> |
1752 | __aurng(__urng); |
1753 | while (__f != __t) |
1754 | *__f++ = -std::log(result_type(1) - __aurng()) / __p.lambda(); |
1755 | } |
1756 | |
1757 | template<typename _RealType, typename _CharT, typename _Traits> |
1758 | std::basic_ostream<_CharT, _Traits>& |
1759 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
1760 | const exponential_distribution<_RealType>& __x) |
1761 | { |
1762 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
1763 | |
1764 | const typename __ios_base::fmtflags __flags = __os.flags(); |
1765 | const _CharT __fill = __os.fill(); |
1766 | const std::streamsize __precision = __os.precision(); |
1767 | __os.flags(__ios_base::scientific | __ios_base::left); |
1768 | __os.fill(__os.widen(' ')); |
1769 | __os.precision(std::numeric_limits<_RealType>::max_digits10); |
1770 | |
1771 | __os << __x.lambda(); |
1772 | |
1773 | __os.flags(__flags); |
1774 | __os.fill(__fill); |
1775 | __os.precision(__precision); |
1776 | return __os; |
1777 | } |
1778 | |
1779 | template<typename _RealType, typename _CharT, typename _Traits> |
1780 | std::basic_istream<_CharT, _Traits>& |
1781 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
1782 | exponential_distribution<_RealType>& __x) |
1783 | { |
1784 | using param_type |
1785 | = typename exponential_distribution<_RealType>::param_type; |
1786 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
1787 | |
1788 | const typename __ios_base::fmtflags __flags = __is.flags(); |
1789 | __is.flags(__ios_base::dec | __ios_base::skipws); |
1790 | |
1791 | _RealType __lambda; |
1792 | if (__is >> __lambda) |
1793 | __x.param(param_type(__lambda)); |
1794 | |
1795 | __is.flags(__flags); |
1796 | return __is; |
1797 | } |
1798 | |
1799 | |
1800 | /** |
1801 | * Polar method due to Marsaglia. |
1802 | * |
1803 | * Devroye, L. Non-Uniform Random Variates Generation. Springer-Verlag, |
1804 | * New York, 1986, Ch. V, Sect. 4.4. |
1805 | */ |
1806 | template<typename _RealType> |
1807 | template<typename _UniformRandomNumberGenerator> |
1808 | typename normal_distribution<_RealType>::result_type |
1809 | normal_distribution<_RealType>:: |
1810 | operator()(_UniformRandomNumberGenerator& __urng, |
1811 | const param_type& __param) |
1812 | { |
1813 | result_type __ret; |
1814 | __detail::_Adaptor<_UniformRandomNumberGenerator, result_type> |
1815 | __aurng(__urng); |
1816 | |
1817 | if (_M_saved_available) |
1818 | { |
1819 | _M_saved_available = false; |
1820 | __ret = _M_saved; |
1821 | } |
1822 | else |
1823 | { |
1824 | result_type __x, __y, __r2; |
1825 | do |
1826 | { |
1827 | __x = result_type(2.0) * __aurng() - 1.0; |
1828 | __y = result_type(2.0) * __aurng() - 1.0; |
1829 | __r2 = __x * __x + __y * __y; |
1830 | } |
1831 | while (__r2 > 1.0 || __r2 == 0.0); |
1832 | |
1833 | const result_type __mult = std::sqrt(-2 * std::log(__r2) / __r2); |
1834 | _M_saved = __x * __mult; |
1835 | _M_saved_available = true; |
1836 | __ret = __y * __mult; |
1837 | } |
1838 | |
1839 | __ret = __ret * __param.stddev() + __param.mean(); |
1840 | return __ret; |
1841 | } |
1842 | |
1843 | template<typename _RealType> |
1844 | template<typename _ForwardIterator, |
1845 | typename _UniformRandomNumberGenerator> |
1846 | void |
1847 | normal_distribution<_RealType>:: |
1848 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
1849 | _UniformRandomNumberGenerator& __urng, |
1850 | const param_type& __param) |
1851 | { |
1852 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
1853 | |
1854 | if (__f == __t) |
1855 | return; |
1856 | |
1857 | if (_M_saved_available) |
1858 | { |
1859 | _M_saved_available = false; |
1860 | *__f++ = _M_saved * __param.stddev() + __param.mean(); |
1861 | |
1862 | if (__f == __t) |
1863 | return; |
1864 | } |
1865 | |
1866 | __detail::_Adaptor<_UniformRandomNumberGenerator, result_type> |
1867 | __aurng(__urng); |
1868 | |
1869 | while (__f + 1 < __t) |
1870 | { |
1871 | result_type __x, __y, __r2; |
1872 | do |
1873 | { |
1874 | __x = result_type(2.0) * __aurng() - 1.0; |
1875 | __y = result_type(2.0) * __aurng() - 1.0; |
1876 | __r2 = __x * __x + __y * __y; |
1877 | } |
1878 | while (__r2 > 1.0 || __r2 == 0.0); |
1879 | |
1880 | const result_type __mult = std::sqrt(-2 * std::log(__r2) / __r2); |
1881 | *__f++ = __y * __mult * __param.stddev() + __param.mean(); |
1882 | *__f++ = __x * __mult * __param.stddev() + __param.mean(); |
1883 | } |
1884 | |
1885 | if (__f != __t) |
1886 | { |
1887 | result_type __x, __y, __r2; |
1888 | do |
1889 | { |
1890 | __x = result_type(2.0) * __aurng() - 1.0; |
1891 | __y = result_type(2.0) * __aurng() - 1.0; |
1892 | __r2 = __x * __x + __y * __y; |
1893 | } |
1894 | while (__r2 > 1.0 || __r2 == 0.0); |
1895 | |
1896 | const result_type __mult = std::sqrt(-2 * std::log(__r2) / __r2); |
1897 | _M_saved = __x * __mult; |
1898 | _M_saved_available = true; |
1899 | *__f = __y * __mult * __param.stddev() + __param.mean(); |
1900 | } |
1901 | } |
1902 | |
1903 | template<typename _RealType> |
1904 | bool |
1905 | operator==(const std::normal_distribution<_RealType>& __d1, |
1906 | const std::normal_distribution<_RealType>& __d2) |
1907 | { |
1908 | if (__d1._M_param == __d2._M_param |
1909 | && __d1._M_saved_available == __d2._M_saved_available) |
1910 | { |
1911 | if (__d1._M_saved_available |
1912 | && __d1._M_saved == __d2._M_saved) |
1913 | return true; |
1914 | else if(!__d1._M_saved_available) |
1915 | return true; |
1916 | else |
1917 | return false; |
1918 | } |
1919 | else |
1920 | return false; |
1921 | } |
1922 | |
1923 | template<typename _RealType, typename _CharT, typename _Traits> |
1924 | std::basic_ostream<_CharT, _Traits>& |
1925 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
1926 | const normal_distribution<_RealType>& __x) |
1927 | { |
1928 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
1929 | |
1930 | const typename __ios_base::fmtflags __flags = __os.flags(); |
1931 | const _CharT __fill = __os.fill(); |
1932 | const std::streamsize __precision = __os.precision(); |
1933 | const _CharT __space = __os.widen(' '); |
1934 | __os.flags(__ios_base::scientific | __ios_base::left); |
1935 | __os.fill(__space); |
1936 | __os.precision(std::numeric_limits<_RealType>::max_digits10); |
1937 | |
1938 | __os << __x.mean() << __space << __x.stddev() |
1939 | << __space << __x._M_saved_available; |
1940 | if (__x._M_saved_available) |
1941 | __os << __space << __x._M_saved; |
1942 | |
1943 | __os.flags(__flags); |
1944 | __os.fill(__fill); |
1945 | __os.precision(__precision); |
1946 | return __os; |
1947 | } |
1948 | |
1949 | template<typename _RealType, typename _CharT, typename _Traits> |
1950 | std::basic_istream<_CharT, _Traits>& |
1951 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
1952 | normal_distribution<_RealType>& __x) |
1953 | { |
1954 | using param_type = typename normal_distribution<_RealType>::param_type; |
1955 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
1956 | |
1957 | const typename __ios_base::fmtflags __flags = __is.flags(); |
1958 | __is.flags(__ios_base::dec | __ios_base::skipws); |
1959 | |
1960 | double __mean, __stddev; |
1961 | bool __saved_avail; |
1962 | if (__is >> __mean >> __stddev >> __saved_avail) |
1963 | { |
1964 | if (!__saved_avail || (__is >> __x._M_saved)) |
1965 | { |
1966 | __x._M_saved_available = __saved_avail; |
1967 | __x.param(param_type(__mean, __stddev)); |
1968 | } |
1969 | } |
1970 | |
1971 | __is.flags(__flags); |
1972 | return __is; |
1973 | } |
1974 | |
1975 | |
1976 | template<typename _RealType> |
1977 | template<typename _ForwardIterator, |
1978 | typename _UniformRandomNumberGenerator> |
1979 | void |
1980 | lognormal_distribution<_RealType>:: |
1981 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
1982 | _UniformRandomNumberGenerator& __urng, |
1983 | const param_type& __p) |
1984 | { |
1985 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
1986 | while (__f != __t) |
1987 | *__f++ = std::exp(__p.s() * _M_nd(__urng) + __p.m()); |
1988 | } |
1989 | |
1990 | template<typename _RealType, typename _CharT, typename _Traits> |
1991 | std::basic_ostream<_CharT, _Traits>& |
1992 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
1993 | const lognormal_distribution<_RealType>& __x) |
1994 | { |
1995 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
1996 | |
1997 | const typename __ios_base::fmtflags __flags = __os.flags(); |
1998 | const _CharT __fill = __os.fill(); |
1999 | const std::streamsize __precision = __os.precision(); |
2000 | const _CharT __space = __os.widen(' '); |
2001 | __os.flags(__ios_base::scientific | __ios_base::left); |
2002 | __os.fill(__space); |
2003 | __os.precision(std::numeric_limits<_RealType>::max_digits10); |
2004 | |
2005 | __os << __x.m() << __space << __x.s() |
2006 | << __space << __x._M_nd; |
2007 | |
2008 | __os.flags(__flags); |
2009 | __os.fill(__fill); |
2010 | __os.precision(__precision); |
2011 | return __os; |
2012 | } |
2013 | |
2014 | template<typename _RealType, typename _CharT, typename _Traits> |
2015 | std::basic_istream<_CharT, _Traits>& |
2016 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
2017 | lognormal_distribution<_RealType>& __x) |
2018 | { |
2019 | using param_type |
2020 | = typename lognormal_distribution<_RealType>::param_type; |
2021 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
2022 | |
2023 | const typename __ios_base::fmtflags __flags = __is.flags(); |
2024 | __is.flags(__ios_base::dec | __ios_base::skipws); |
2025 | |
2026 | _RealType __m, __s; |
2027 | if (__is >> __m >> __s >> __x._M_nd) |
2028 | __x.param(param_type(__m, __s)); |
2029 | |
2030 | __is.flags(__flags); |
2031 | return __is; |
2032 | } |
2033 | |
2034 | template<typename _RealType> |
2035 | template<typename _ForwardIterator, |
2036 | typename _UniformRandomNumberGenerator> |
2037 | void |
2038 | std::chi_squared_distribution<_RealType>:: |
2039 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
2040 | _UniformRandomNumberGenerator& __urng) |
2041 | { |
2042 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
2043 | while (__f != __t) |
2044 | *__f++ = 2 * _M_gd(__urng); |
2045 | } |
2046 | |
2047 | template<typename _RealType> |
2048 | template<typename _ForwardIterator, |
2049 | typename _UniformRandomNumberGenerator> |
2050 | void |
2051 | std::chi_squared_distribution<_RealType>:: |
2052 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
2053 | _UniformRandomNumberGenerator& __urng, |
2054 | const typename |
2055 | std::gamma_distribution<result_type>::param_type& __p) |
2056 | { |
2057 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
2058 | while (__f != __t) |
2059 | *__f++ = 2 * _M_gd(__urng, __p); |
2060 | } |
2061 | |
2062 | template<typename _RealType, typename _CharT, typename _Traits> |
2063 | std::basic_ostream<_CharT, _Traits>& |
2064 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
2065 | const chi_squared_distribution<_RealType>& __x) |
2066 | { |
2067 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
2068 | |
2069 | const typename __ios_base::fmtflags __flags = __os.flags(); |
2070 | const _CharT __fill = __os.fill(); |
2071 | const std::streamsize __precision = __os.precision(); |
2072 | const _CharT __space = __os.widen(' '); |
2073 | __os.flags(__ios_base::scientific | __ios_base::left); |
2074 | __os.fill(__space); |
2075 | __os.precision(std::numeric_limits<_RealType>::max_digits10); |
2076 | |
2077 | __os << __x.n() << __space << __x._M_gd; |
2078 | |
2079 | __os.flags(__flags); |
2080 | __os.fill(__fill); |
2081 | __os.precision(__precision); |
2082 | return __os; |
2083 | } |
2084 | |
2085 | template<typename _RealType, typename _CharT, typename _Traits> |
2086 | std::basic_istream<_CharT, _Traits>& |
2087 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
2088 | chi_squared_distribution<_RealType>& __x) |
2089 | { |
2090 | using param_type |
2091 | = typename chi_squared_distribution<_RealType>::param_type; |
2092 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
2093 | |
2094 | const typename __ios_base::fmtflags __flags = __is.flags(); |
2095 | __is.flags(__ios_base::dec | __ios_base::skipws); |
2096 | |
2097 | _RealType __n; |
2098 | if (__is >> __n >> __x._M_gd) |
2099 | __x.param(param_type(__n)); |
2100 | |
2101 | __is.flags(__flags); |
2102 | return __is; |
2103 | } |
2104 | |
2105 | |
2106 | template<typename _RealType> |
2107 | template<typename _UniformRandomNumberGenerator> |
2108 | typename cauchy_distribution<_RealType>::result_type |
2109 | cauchy_distribution<_RealType>:: |
2110 | operator()(_UniformRandomNumberGenerator& __urng, |
2111 | const param_type& __p) |
2112 | { |
2113 | __detail::_Adaptor<_UniformRandomNumberGenerator, result_type> |
2114 | __aurng(__urng); |
2115 | _RealType __u; |
2116 | do |
2117 | __u = __aurng(); |
2118 | while (__u == 0.5); |
2119 | |
2120 | const _RealType __pi = 3.1415926535897932384626433832795029L; |
2121 | return __p.a() + __p.b() * std::tan(__pi * __u); |
2122 | } |
2123 | |
2124 | template<typename _RealType> |
2125 | template<typename _ForwardIterator, |
2126 | typename _UniformRandomNumberGenerator> |
2127 | void |
2128 | cauchy_distribution<_RealType>:: |
2129 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
2130 | _UniformRandomNumberGenerator& __urng, |
2131 | const param_type& __p) |
2132 | { |
2133 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
2134 | const _RealType __pi = 3.1415926535897932384626433832795029L; |
2135 | __detail::_Adaptor<_UniformRandomNumberGenerator, result_type> |
2136 | __aurng(__urng); |
2137 | while (__f != __t) |
2138 | { |
2139 | _RealType __u; |
2140 | do |
2141 | __u = __aurng(); |
2142 | while (__u == 0.5); |
2143 | |
2144 | *__f++ = __p.a() + __p.b() * std::tan(__pi * __u); |
2145 | } |
2146 | } |
2147 | |
2148 | template<typename _RealType, typename _CharT, typename _Traits> |
2149 | std::basic_ostream<_CharT, _Traits>& |
2150 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
2151 | const cauchy_distribution<_RealType>& __x) |
2152 | { |
2153 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
2154 | |
2155 | const typename __ios_base::fmtflags __flags = __os.flags(); |
2156 | const _CharT __fill = __os.fill(); |
2157 | const std::streamsize __precision = __os.precision(); |
2158 | const _CharT __space = __os.widen(' '); |
2159 | __os.flags(__ios_base::scientific | __ios_base::left); |
2160 | __os.fill(__space); |
2161 | __os.precision(std::numeric_limits<_RealType>::max_digits10); |
2162 | |
2163 | __os << __x.a() << __space << __x.b(); |
2164 | |
2165 | __os.flags(__flags); |
2166 | __os.fill(__fill); |
2167 | __os.precision(__precision); |
2168 | return __os; |
2169 | } |
2170 | |
2171 | template<typename _RealType, typename _CharT, typename _Traits> |
2172 | std::basic_istream<_CharT, _Traits>& |
2173 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
2174 | cauchy_distribution<_RealType>& __x) |
2175 | { |
2176 | using param_type = typename cauchy_distribution<_RealType>::param_type; |
2177 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
2178 | |
2179 | const typename __ios_base::fmtflags __flags = __is.flags(); |
2180 | __is.flags(__ios_base::dec | __ios_base::skipws); |
2181 | |
2182 | _RealType __a, __b; |
2183 | if (__is >> __a >> __b) |
2184 | __x.param(param_type(__a, __b)); |
2185 | |
2186 | __is.flags(__flags); |
2187 | return __is; |
2188 | } |
2189 | |
2190 | |
2191 | template<typename _RealType> |
2192 | template<typename _ForwardIterator, |
2193 | typename _UniformRandomNumberGenerator> |
2194 | void |
2195 | std::fisher_f_distribution<_RealType>:: |
2196 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
2197 | _UniformRandomNumberGenerator& __urng) |
2198 | { |
2199 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
2200 | while (__f != __t) |
2201 | *__f++ = ((_M_gd_x(__urng) * n()) / (_M_gd_y(__urng) * m())); |
2202 | } |
2203 | |
2204 | template<typename _RealType> |
2205 | template<typename _ForwardIterator, |
2206 | typename _UniformRandomNumberGenerator> |
2207 | void |
2208 | std::fisher_f_distribution<_RealType>:: |
2209 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
2210 | _UniformRandomNumberGenerator& __urng, |
2211 | const param_type& __p) |
2212 | { |
2213 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
2214 | typedef typename std::gamma_distribution<result_type>::param_type |
2215 | param_type; |
2216 | param_type __p1(__p.m() / 2); |
2217 | param_type __p2(__p.n() / 2); |
2218 | while (__f != __t) |
2219 | *__f++ = ((_M_gd_x(__urng, __p1) * n()) |
2220 | / (_M_gd_y(__urng, __p2) * m())); |
2221 | } |
2222 | |
2223 | template<typename _RealType, typename _CharT, typename _Traits> |
2224 | std::basic_ostream<_CharT, _Traits>& |
2225 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
2226 | const fisher_f_distribution<_RealType>& __x) |
2227 | { |
2228 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
2229 | |
2230 | const typename __ios_base::fmtflags __flags = __os.flags(); |
2231 | const _CharT __fill = __os.fill(); |
2232 | const std::streamsize __precision = __os.precision(); |
2233 | const _CharT __space = __os.widen(' '); |
2234 | __os.flags(__ios_base::scientific | __ios_base::left); |
2235 | __os.fill(__space); |
2236 | __os.precision(std::numeric_limits<_RealType>::max_digits10); |
2237 | |
2238 | __os << __x.m() << __space << __x.n() |
2239 | << __space << __x._M_gd_x << __space << __x._M_gd_y; |
2240 | |
2241 | __os.flags(__flags); |
2242 | __os.fill(__fill); |
2243 | __os.precision(__precision); |
2244 | return __os; |
2245 | } |
2246 | |
2247 | template<typename _RealType, typename _CharT, typename _Traits> |
2248 | std::basic_istream<_CharT, _Traits>& |
2249 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
2250 | fisher_f_distribution<_RealType>& __x) |
2251 | { |
2252 | using param_type |
2253 | = typename fisher_f_distribution<_RealType>::param_type; |
2254 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
2255 | |
2256 | const typename __ios_base::fmtflags __flags = __is.flags(); |
2257 | __is.flags(__ios_base::dec | __ios_base::skipws); |
2258 | |
2259 | _RealType __m, __n; |
2260 | if (__is >> __m >> __n >> __x._M_gd_x >> __x._M_gd_y) |
2261 | __x.param(param_type(__m, __n)); |
2262 | |
2263 | __is.flags(__flags); |
2264 | return __is; |
2265 | } |
2266 | |
2267 | |
2268 | template<typename _RealType> |
2269 | template<typename _ForwardIterator, |
2270 | typename _UniformRandomNumberGenerator> |
2271 | void |
2272 | std::student_t_distribution<_RealType>:: |
2273 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
2274 | _UniformRandomNumberGenerator& __urng) |
2275 | { |
2276 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
2277 | while (__f != __t) |
2278 | *__f++ = _M_nd(__urng) * std::sqrt(n() / _M_gd(__urng)); |
2279 | } |
2280 | |
2281 | template<typename _RealType> |
2282 | template<typename _ForwardIterator, |
2283 | typename _UniformRandomNumberGenerator> |
2284 | void |
2285 | std::student_t_distribution<_RealType>:: |
2286 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
2287 | _UniformRandomNumberGenerator& __urng, |
2288 | const param_type& __p) |
2289 | { |
2290 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
2291 | typename std::gamma_distribution<result_type>::param_type |
2292 | __p2(__p.n() / 2, 2); |
2293 | while (__f != __t) |
2294 | *__f++ = _M_nd(__urng) * std::sqrt(__p.n() / _M_gd(__urng, __p2)); |
2295 | } |
2296 | |
2297 | template<typename _RealType, typename _CharT, typename _Traits> |
2298 | std::basic_ostream<_CharT, _Traits>& |
2299 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
2300 | const student_t_distribution<_RealType>& __x) |
2301 | { |
2302 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
2303 | |
2304 | const typename __ios_base::fmtflags __flags = __os.flags(); |
2305 | const _CharT __fill = __os.fill(); |
2306 | const std::streamsize __precision = __os.precision(); |
2307 | const _CharT __space = __os.widen(' '); |
2308 | __os.flags(__ios_base::scientific | __ios_base::left); |
2309 | __os.fill(__space); |
2310 | __os.precision(std::numeric_limits<_RealType>::max_digits10); |
2311 | |
2312 | __os << __x.n() << __space << __x._M_nd << __space << __x._M_gd; |
2313 | |
2314 | __os.flags(__flags); |
2315 | __os.fill(__fill); |
2316 | __os.precision(__precision); |
2317 | return __os; |
2318 | } |
2319 | |
2320 | template<typename _RealType, typename _CharT, typename _Traits> |
2321 | std::basic_istream<_CharT, _Traits>& |
2322 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
2323 | student_t_distribution<_RealType>& __x) |
2324 | { |
2325 | using param_type |
2326 | = typename student_t_distribution<_RealType>::param_type; |
2327 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
2328 | |
2329 | const typename __ios_base::fmtflags __flags = __is.flags(); |
2330 | __is.flags(__ios_base::dec | __ios_base::skipws); |
2331 | |
2332 | _RealType __n; |
2333 | if (__is >> __n >> __x._M_nd >> __x._M_gd) |
2334 | __x.param(param_type(__n)); |
2335 | |
2336 | __is.flags(__flags); |
2337 | return __is; |
2338 | } |
2339 | |
2340 | |
2341 | template<typename _RealType> |
2342 | void |
2343 | gamma_distribution<_RealType>::param_type:: |
2344 | _M_initialize() |
2345 | { |
2346 | _M_malpha = _M_alpha < 1.0 ? _M_alpha + _RealType(1.0) : _M_alpha; |
2347 | |
2348 | const _RealType __a1 = _M_malpha - _RealType(1.0) / _RealType(3.0); |
2349 | _M_a2 = _RealType(1.0) / std::sqrt(_RealType(9.0) * __a1); |
2350 | } |
2351 | |
2352 | /** |
2353 | * Marsaglia, G. and Tsang, W. W. |
2354 | * "A Simple Method for Generating Gamma Variables" |
2355 | * ACM Transactions on Mathematical Software, 26, 3, 363-372, 2000. |
2356 | */ |
2357 | template<typename _RealType> |
2358 | template<typename _UniformRandomNumberGenerator> |
2359 | typename gamma_distribution<_RealType>::result_type |
2360 | gamma_distribution<_RealType>:: |
2361 | operator()(_UniformRandomNumberGenerator& __urng, |
2362 | const param_type& __param) |
2363 | { |
2364 | __detail::_Adaptor<_UniformRandomNumberGenerator, result_type> |
2365 | __aurng(__urng); |
2366 | |
2367 | result_type __u, __v, __n; |
2368 | const result_type __a1 = (__param._M_malpha |
2369 | - _RealType(1.0) / _RealType(3.0)); |
2370 | |
2371 | do |
2372 | { |
2373 | do |
2374 | { |
2375 | __n = _M_nd(__urng); |
2376 | __v = result_type(1.0) + __param._M_a2 * __n; |
2377 | } |
2378 | while (__v <= 0.0); |
2379 | |
2380 | __v = __v * __v * __v; |
2381 | __u = __aurng(); |
2382 | } |
2383 | while (__u > result_type(1.0) - 0.0331 * __n * __n * __n * __n |
2384 | && (std::log(__u) > (0.5 * __n * __n + __a1 |
2385 | * (1.0 - __v + std::log(__v))))); |
2386 | |
2387 | if (__param.alpha() == __param._M_malpha) |
2388 | return __a1 * __v * __param.beta(); |
2389 | else |
2390 | { |
2391 | do |
2392 | __u = __aurng(); |
2393 | while (__u == 0.0); |
2394 | |
2395 | return (std::pow(__u, result_type(1.0) / __param.alpha()) |
2396 | * __a1 * __v * __param.beta()); |
2397 | } |
2398 | } |
2399 | |
2400 | template<typename _RealType> |
2401 | template<typename _ForwardIterator, |
2402 | typename _UniformRandomNumberGenerator> |
2403 | void |
2404 | gamma_distribution<_RealType>:: |
2405 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
2406 | _UniformRandomNumberGenerator& __urng, |
2407 | const param_type& __param) |
2408 | { |
2409 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
2410 | __detail::_Adaptor<_UniformRandomNumberGenerator, result_type> |
2411 | __aurng(__urng); |
2412 | |
2413 | result_type __u, __v, __n; |
2414 | const result_type __a1 = (__param._M_malpha |
2415 | - _RealType(1.0) / _RealType(3.0)); |
2416 | |
2417 | if (__param.alpha() == __param._M_malpha) |
2418 | while (__f != __t) |
2419 | { |
2420 | do |
2421 | { |
2422 | do |
2423 | { |
2424 | __n = _M_nd(__urng); |
2425 | __v = result_type(1.0) + __param._M_a2 * __n; |
2426 | } |
2427 | while (__v <= 0.0); |
2428 | |
2429 | __v = __v * __v * __v; |
2430 | __u = __aurng(); |
2431 | } |
2432 | while (__u > result_type(1.0) - 0.0331 * __n * __n * __n * __n |
2433 | && (std::log(__u) > (0.5 * __n * __n + __a1 |
2434 | * (1.0 - __v + std::log(__v))))); |
2435 | |
2436 | *__f++ = __a1 * __v * __param.beta(); |
2437 | } |
2438 | else |
2439 | while (__f != __t) |
2440 | { |
2441 | do |
2442 | { |
2443 | do |
2444 | { |
2445 | __n = _M_nd(__urng); |
2446 | __v = result_type(1.0) + __param._M_a2 * __n; |
2447 | } |
2448 | while (__v <= 0.0); |
2449 | |
2450 | __v = __v * __v * __v; |
2451 | __u = __aurng(); |
2452 | } |
2453 | while (__u > result_type(1.0) - 0.0331 * __n * __n * __n * __n |
2454 | && (std::log(__u) > (0.5 * __n * __n + __a1 |
2455 | * (1.0 - __v + std::log(__v))))); |
2456 | |
2457 | do |
2458 | __u = __aurng(); |
2459 | while (__u == 0.0); |
2460 | |
2461 | *__f++ = (std::pow(__u, result_type(1.0) / __param.alpha()) |
2462 | * __a1 * __v * __param.beta()); |
2463 | } |
2464 | } |
2465 | |
2466 | template<typename _RealType, typename _CharT, typename _Traits> |
2467 | std::basic_ostream<_CharT, _Traits>& |
2468 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
2469 | const gamma_distribution<_RealType>& __x) |
2470 | { |
2471 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
2472 | |
2473 | const typename __ios_base::fmtflags __flags = __os.flags(); |
2474 | const _CharT __fill = __os.fill(); |
2475 | const std::streamsize __precision = __os.precision(); |
2476 | const _CharT __space = __os.widen(' '); |
2477 | __os.flags(__ios_base::scientific | __ios_base::left); |
2478 | __os.fill(__space); |
2479 | __os.precision(std::numeric_limits<_RealType>::max_digits10); |
2480 | |
2481 | __os << __x.alpha() << __space << __x.beta() |
2482 | << __space << __x._M_nd; |
2483 | |
2484 | __os.flags(__flags); |
2485 | __os.fill(__fill); |
2486 | __os.precision(__precision); |
2487 | return __os; |
2488 | } |
2489 | |
2490 | template<typename _RealType, typename _CharT, typename _Traits> |
2491 | std::basic_istream<_CharT, _Traits>& |
2492 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
2493 | gamma_distribution<_RealType>& __x) |
2494 | { |
2495 | using param_type = typename gamma_distribution<_RealType>::param_type; |
2496 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
2497 | |
2498 | const typename __ios_base::fmtflags __flags = __is.flags(); |
2499 | __is.flags(__ios_base::dec | __ios_base::skipws); |
2500 | |
2501 | _RealType __alpha_val, __beta_val; |
2502 | if (__is >> __alpha_val >> __beta_val >> __x._M_nd) |
2503 | __x.param(param_type(__alpha_val, __beta_val)); |
2504 | |
2505 | __is.flags(__flags); |
2506 | return __is; |
2507 | } |
2508 | |
2509 | |
2510 | template<typename _RealType> |
2511 | template<typename _UniformRandomNumberGenerator> |
2512 | typename weibull_distribution<_RealType>::result_type |
2513 | weibull_distribution<_RealType>:: |
2514 | operator()(_UniformRandomNumberGenerator& __urng, |
2515 | const param_type& __p) |
2516 | { |
2517 | __detail::_Adaptor<_UniformRandomNumberGenerator, result_type> |
2518 | __aurng(__urng); |
2519 | return __p.b() * std::pow(-std::log(result_type(1) - __aurng()), |
2520 | result_type(1) / __p.a()); |
2521 | } |
2522 | |
2523 | template<typename _RealType> |
2524 | template<typename _ForwardIterator, |
2525 | typename _UniformRandomNumberGenerator> |
2526 | void |
2527 | weibull_distribution<_RealType>:: |
2528 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
2529 | _UniformRandomNumberGenerator& __urng, |
2530 | const param_type& __p) |
2531 | { |
2532 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
2533 | __detail::_Adaptor<_UniformRandomNumberGenerator, result_type> |
2534 | __aurng(__urng); |
2535 | auto __inv_a = result_type(1) / __p.a(); |
2536 | |
2537 | while (__f != __t) |
2538 | *__f++ = __p.b() * std::pow(-std::log(result_type(1) - __aurng()), |
2539 | __inv_a); |
2540 | } |
2541 | |
2542 | template<typename _RealType, typename _CharT, typename _Traits> |
2543 | std::basic_ostream<_CharT, _Traits>& |
2544 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
2545 | const weibull_distribution<_RealType>& __x) |
2546 | { |
2547 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
2548 | |
2549 | const typename __ios_base::fmtflags __flags = __os.flags(); |
2550 | const _CharT __fill = __os.fill(); |
2551 | const std::streamsize __precision = __os.precision(); |
2552 | const _CharT __space = __os.widen(' '); |
2553 | __os.flags(__ios_base::scientific | __ios_base::left); |
2554 | __os.fill(__space); |
2555 | __os.precision(std::numeric_limits<_RealType>::max_digits10); |
2556 | |
2557 | __os << __x.a() << __space << __x.b(); |
2558 | |
2559 | __os.flags(__flags); |
2560 | __os.fill(__fill); |
2561 | __os.precision(__precision); |
2562 | return __os; |
2563 | } |
2564 | |
2565 | template<typename _RealType, typename _CharT, typename _Traits> |
2566 | std::basic_istream<_CharT, _Traits>& |
2567 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
2568 | weibull_distribution<_RealType>& __x) |
2569 | { |
2570 | using param_type = typename weibull_distribution<_RealType>::param_type; |
2571 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
2572 | |
2573 | const typename __ios_base::fmtflags __flags = __is.flags(); |
2574 | __is.flags(__ios_base::dec | __ios_base::skipws); |
2575 | |
2576 | _RealType __a, __b; |
2577 | if (__is >> __a >> __b) |
2578 | __x.param(param_type(__a, __b)); |
2579 | |
2580 | __is.flags(__flags); |
2581 | return __is; |
2582 | } |
2583 | |
2584 | |
2585 | template<typename _RealType> |
2586 | template<typename _UniformRandomNumberGenerator> |
2587 | typename extreme_value_distribution<_RealType>::result_type |
2588 | extreme_value_distribution<_RealType>:: |
2589 | operator()(_UniformRandomNumberGenerator& __urng, |
2590 | const param_type& __p) |
2591 | { |
2592 | __detail::_Adaptor<_UniformRandomNumberGenerator, result_type> |
2593 | __aurng(__urng); |
2594 | return __p.a() - __p.b() * std::log(-std::log(result_type(1) |
2595 | - __aurng())); |
2596 | } |
2597 | |
2598 | template<typename _RealType> |
2599 | template<typename _ForwardIterator, |
2600 | typename _UniformRandomNumberGenerator> |
2601 | void |
2602 | extreme_value_distribution<_RealType>:: |
2603 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
2604 | _UniformRandomNumberGenerator& __urng, |
2605 | const param_type& __p) |
2606 | { |
2607 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
2608 | __detail::_Adaptor<_UniformRandomNumberGenerator, result_type> |
2609 | __aurng(__urng); |
2610 | |
2611 | while (__f != __t) |
2612 | *__f++ = __p.a() - __p.b() * std::log(-std::log(result_type(1) |
2613 | - __aurng())); |
2614 | } |
2615 | |
2616 | template<typename _RealType, typename _CharT, typename _Traits> |
2617 | std::basic_ostream<_CharT, _Traits>& |
2618 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
2619 | const extreme_value_distribution<_RealType>& __x) |
2620 | { |
2621 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
2622 | |
2623 | const typename __ios_base::fmtflags __flags = __os.flags(); |
2624 | const _CharT __fill = __os.fill(); |
2625 | const std::streamsize __precision = __os.precision(); |
2626 | const _CharT __space = __os.widen(' '); |
2627 | __os.flags(__ios_base::scientific | __ios_base::left); |
2628 | __os.fill(__space); |
2629 | __os.precision(std::numeric_limits<_RealType>::max_digits10); |
2630 | |
2631 | __os << __x.a() << __space << __x.b(); |
2632 | |
2633 | __os.flags(__flags); |
2634 | __os.fill(__fill); |
2635 | __os.precision(__precision); |
2636 | return __os; |
2637 | } |
2638 | |
2639 | template<typename _RealType, typename _CharT, typename _Traits> |
2640 | std::basic_istream<_CharT, _Traits>& |
2641 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
2642 | extreme_value_distribution<_RealType>& __x) |
2643 | { |
2644 | using param_type |
2645 | = typename extreme_value_distribution<_RealType>::param_type; |
2646 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
2647 | |
2648 | const typename __ios_base::fmtflags __flags = __is.flags(); |
2649 | __is.flags(__ios_base::dec | __ios_base::skipws); |
2650 | |
2651 | _RealType __a, __b; |
2652 | if (__is >> __a >> __b) |
2653 | __x.param(param_type(__a, __b)); |
2654 | |
2655 | __is.flags(__flags); |
2656 | return __is; |
2657 | } |
2658 | |
2659 | |
2660 | template<typename _IntType> |
2661 | void |
2662 | discrete_distribution<_IntType>::param_type:: |
2663 | _M_initialize() |
2664 | { |
2665 | if (_M_prob.size() < 2) |
2666 | { |
2667 | _M_prob.clear(); |
2668 | return; |
2669 | } |
2670 | |
2671 | const double __sum = std::accumulate(first: _M_prob.begin(), |
2672 | last: _M_prob.end(), init: 0.0); |
2673 | __glibcxx_assert(__sum > 0); |
2674 | // Now normalize the probabilites. |
2675 | __detail::__normalize(first: _M_prob.begin(), last: _M_prob.end(), result: _M_prob.begin(), |
2676 | factor: __sum); |
2677 | // Accumulate partial sums. |
2678 | _M_cp.reserve(n: _M_prob.size()); |
2679 | std::partial_sum(first: _M_prob.begin(), last: _M_prob.end(), |
2680 | result: std::back_inserter(x&: _M_cp)); |
2681 | // Make sure the last cumulative probability is one. |
2682 | _M_cp[_M_cp.size() - 1] = 1.0; |
2683 | } |
2684 | |
2685 | template<typename _IntType> |
2686 | template<typename _Func> |
2687 | discrete_distribution<_IntType>::param_type:: |
2688 | param_type(size_t __nw, double __xmin, double __xmax, _Func __fw) |
2689 | : _M_prob(), _M_cp() |
2690 | { |
2691 | const size_t __n = __nw == 0 ? 1 : __nw; |
2692 | const double __delta = (__xmax - __xmin) / __n; |
2693 | |
2694 | _M_prob.reserve(__n); |
2695 | for (size_t __k = 0; __k < __nw; ++__k) |
2696 | _M_prob.push_back(__fw(__xmin + __k * __delta + 0.5 * __delta)); |
2697 | |
2698 | _M_initialize(); |
2699 | } |
2700 | |
2701 | template<typename _IntType> |
2702 | template<typename _UniformRandomNumberGenerator> |
2703 | typename discrete_distribution<_IntType>::result_type |
2704 | discrete_distribution<_IntType>:: |
2705 | operator()(_UniformRandomNumberGenerator& __urng, |
2706 | const param_type& __param) |
2707 | { |
2708 | if (__param._M_cp.empty()) |
2709 | return result_type(0); |
2710 | |
2711 | __detail::_Adaptor<_UniformRandomNumberGenerator, double> |
2712 | __aurng(__urng); |
2713 | |
2714 | const double __p = __aurng(); |
2715 | auto __pos = std::lower_bound(__param._M_cp.begin(), |
2716 | __param._M_cp.end(), __p); |
2717 | |
2718 | return __pos - __param._M_cp.begin(); |
2719 | } |
2720 | |
2721 | template<typename _IntType> |
2722 | template<typename _ForwardIterator, |
2723 | typename _UniformRandomNumberGenerator> |
2724 | void |
2725 | discrete_distribution<_IntType>:: |
2726 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
2727 | _UniformRandomNumberGenerator& __urng, |
2728 | const param_type& __param) |
2729 | { |
2730 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
2731 | |
2732 | if (__param._M_cp.empty()) |
2733 | { |
2734 | while (__f != __t) |
2735 | *__f++ = result_type(0); |
2736 | return; |
2737 | } |
2738 | |
2739 | __detail::_Adaptor<_UniformRandomNumberGenerator, double> |
2740 | __aurng(__urng); |
2741 | |
2742 | while (__f != __t) |
2743 | { |
2744 | const double __p = __aurng(); |
2745 | auto __pos = std::lower_bound(__param._M_cp.begin(), |
2746 | __param._M_cp.end(), __p); |
2747 | |
2748 | *__f++ = __pos - __param._M_cp.begin(); |
2749 | } |
2750 | } |
2751 | |
2752 | template<typename _IntType, typename _CharT, typename _Traits> |
2753 | std::basic_ostream<_CharT, _Traits>& |
2754 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
2755 | const discrete_distribution<_IntType>& __x) |
2756 | { |
2757 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
2758 | |
2759 | const typename __ios_base::fmtflags __flags = __os.flags(); |
2760 | const _CharT __fill = __os.fill(); |
2761 | const std::streamsize __precision = __os.precision(); |
2762 | const _CharT __space = __os.widen(' '); |
2763 | __os.flags(__ios_base::scientific | __ios_base::left); |
2764 | __os.fill(__space); |
2765 | __os.precision(std::numeric_limits<double>::max_digits10); |
2766 | |
2767 | std::vector<double> __prob = __x.probabilities(); |
2768 | __os << __prob.size(); |
2769 | for (auto __dit = __prob.begin(); __dit != __prob.end(); ++__dit) |
2770 | __os << __space << *__dit; |
2771 | |
2772 | __os.flags(__flags); |
2773 | __os.fill(__fill); |
2774 | __os.precision(__precision); |
2775 | return __os; |
2776 | } |
2777 | |
2778 | namespace __detail |
2779 | { |
2780 | template<typename _ValT, typename _CharT, typename _Traits> |
2781 | basic_istream<_CharT, _Traits>& |
2782 | (basic_istream<_CharT, _Traits>& __is, |
2783 | vector<_ValT>& __vals, size_t __n) |
2784 | { |
2785 | __vals.reserve(__n); |
2786 | while (__n--) |
2787 | { |
2788 | _ValT __val; |
2789 | if (__is >> __val) |
2790 | __vals.push_back(__val); |
2791 | else |
2792 | break; |
2793 | } |
2794 | return __is; |
2795 | } |
2796 | } // namespace __detail |
2797 | |
2798 | template<typename _IntType, typename _CharT, typename _Traits> |
2799 | std::basic_istream<_CharT, _Traits>& |
2800 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
2801 | discrete_distribution<_IntType>& __x) |
2802 | { |
2803 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
2804 | |
2805 | const typename __ios_base::fmtflags __flags = __is.flags(); |
2806 | __is.flags(__ios_base::dec | __ios_base::skipws); |
2807 | |
2808 | size_t __n; |
2809 | if (__is >> __n) |
2810 | { |
2811 | std::vector<double> __prob_vec; |
2812 | if (__detail::__extract_params(__is, __prob_vec, __n)) |
2813 | __x.param({__prob_vec.begin(), __prob_vec.end()}); |
2814 | } |
2815 | |
2816 | __is.flags(__flags); |
2817 | return __is; |
2818 | } |
2819 | |
2820 | |
2821 | template<typename _RealType> |
2822 | void |
2823 | piecewise_constant_distribution<_RealType>::param_type:: |
2824 | _M_initialize() |
2825 | { |
2826 | if (_M_int.size() < 2 |
2827 | || (_M_int.size() == 2 |
2828 | && _M_int[0] == _RealType(0) |
2829 | && _M_int[1] == _RealType(1))) |
2830 | { |
2831 | _M_int.clear(); |
2832 | _M_den.clear(); |
2833 | return; |
2834 | } |
2835 | |
2836 | const double __sum = std::accumulate(first: _M_den.begin(), |
2837 | last: _M_den.end(), init: 0.0); |
2838 | __glibcxx_assert(__sum > 0); |
2839 | |
2840 | __detail::__normalize(first: _M_den.begin(), last: _M_den.end(), result: _M_den.begin(), |
2841 | factor: __sum); |
2842 | |
2843 | _M_cp.reserve(n: _M_den.size()); |
2844 | std::partial_sum(first: _M_den.begin(), last: _M_den.end(), |
2845 | result: std::back_inserter(x&: _M_cp)); |
2846 | |
2847 | // Make sure the last cumulative probability is one. |
2848 | _M_cp[_M_cp.size() - 1] = 1.0; |
2849 | |
2850 | for (size_t __k = 0; __k < _M_den.size(); ++__k) |
2851 | _M_den[__k] /= _M_int[__k + 1] - _M_int[__k]; |
2852 | } |
2853 | |
2854 | template<typename _RealType> |
2855 | template<typename _InputIteratorB, typename _InputIteratorW> |
2856 | piecewise_constant_distribution<_RealType>::param_type:: |
2857 | param_type(_InputIteratorB __bbegin, |
2858 | _InputIteratorB __bend, |
2859 | _InputIteratorW __wbegin) |
2860 | : _M_int(), _M_den(), _M_cp() |
2861 | { |
2862 | if (__bbegin != __bend) |
2863 | { |
2864 | for (;;) |
2865 | { |
2866 | _M_int.push_back(*__bbegin); |
2867 | ++__bbegin; |
2868 | if (__bbegin == __bend) |
2869 | break; |
2870 | |
2871 | _M_den.push_back(*__wbegin); |
2872 | ++__wbegin; |
2873 | } |
2874 | } |
2875 | |
2876 | _M_initialize(); |
2877 | } |
2878 | |
2879 | template<typename _RealType> |
2880 | template<typename _Func> |
2881 | piecewise_constant_distribution<_RealType>::param_type:: |
2882 | param_type(initializer_list<_RealType> __bl, _Func __fw) |
2883 | : _M_int(), _M_den(), _M_cp() |
2884 | { |
2885 | _M_int.reserve(__bl.size()); |
2886 | for (auto __biter = __bl.begin(); __biter != __bl.end(); ++__biter) |
2887 | _M_int.push_back(*__biter); |
2888 | |
2889 | _M_den.reserve(n: _M_int.size() - 1); |
2890 | for (size_t __k = 0; __k < _M_int.size() - 1; ++__k) |
2891 | _M_den.push_back(__fw(0.5 * (_M_int[__k + 1] + _M_int[__k]))); |
2892 | |
2893 | _M_initialize(); |
2894 | } |
2895 | |
2896 | template<typename _RealType> |
2897 | template<typename _Func> |
2898 | piecewise_constant_distribution<_RealType>::param_type:: |
2899 | param_type(size_t __nw, _RealType __xmin, _RealType __xmax, _Func __fw) |
2900 | : _M_int(), _M_den(), _M_cp() |
2901 | { |
2902 | const size_t __n = __nw == 0 ? 1 : __nw; |
2903 | const _RealType __delta = (__xmax - __xmin) / __n; |
2904 | |
2905 | _M_int.reserve(__n + 1); |
2906 | for (size_t __k = 0; __k <= __nw; ++__k) |
2907 | _M_int.push_back(__xmin + __k * __delta); |
2908 | |
2909 | _M_den.reserve(__n); |
2910 | for (size_t __k = 0; __k < __nw; ++__k) |
2911 | _M_den.push_back(__fw(_M_int[__k] + 0.5 * __delta)); |
2912 | |
2913 | _M_initialize(); |
2914 | } |
2915 | |
2916 | template<typename _RealType> |
2917 | template<typename _UniformRandomNumberGenerator> |
2918 | typename piecewise_constant_distribution<_RealType>::result_type |
2919 | piecewise_constant_distribution<_RealType>:: |
2920 | operator()(_UniformRandomNumberGenerator& __urng, |
2921 | const param_type& __param) |
2922 | { |
2923 | __detail::_Adaptor<_UniformRandomNumberGenerator, double> |
2924 | __aurng(__urng); |
2925 | |
2926 | const double __p = __aurng(); |
2927 | if (__param._M_cp.empty()) |
2928 | return __p; |
2929 | |
2930 | auto __pos = std::lower_bound(__param._M_cp.begin(), |
2931 | __param._M_cp.end(), __p); |
2932 | const size_t __i = __pos - __param._M_cp.begin(); |
2933 | |
2934 | const double __pref = __i > 0 ? __param._M_cp[__i - 1] : 0.0; |
2935 | |
2936 | return __param._M_int[__i] + (__p - __pref) / __param._M_den[__i]; |
2937 | } |
2938 | |
2939 | template<typename _RealType> |
2940 | template<typename _ForwardIterator, |
2941 | typename _UniformRandomNumberGenerator> |
2942 | void |
2943 | piecewise_constant_distribution<_RealType>:: |
2944 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
2945 | _UniformRandomNumberGenerator& __urng, |
2946 | const param_type& __param) |
2947 | { |
2948 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
2949 | __detail::_Adaptor<_UniformRandomNumberGenerator, double> |
2950 | __aurng(__urng); |
2951 | |
2952 | if (__param._M_cp.empty()) |
2953 | { |
2954 | while (__f != __t) |
2955 | *__f++ = __aurng(); |
2956 | return; |
2957 | } |
2958 | |
2959 | while (__f != __t) |
2960 | { |
2961 | const double __p = __aurng(); |
2962 | |
2963 | auto __pos = std::lower_bound(__param._M_cp.begin(), |
2964 | __param._M_cp.end(), __p); |
2965 | const size_t __i = __pos - __param._M_cp.begin(); |
2966 | |
2967 | const double __pref = __i > 0 ? __param._M_cp[__i - 1] : 0.0; |
2968 | |
2969 | *__f++ = (__param._M_int[__i] |
2970 | + (__p - __pref) / __param._M_den[__i]); |
2971 | } |
2972 | } |
2973 | |
2974 | template<typename _RealType, typename _CharT, typename _Traits> |
2975 | std::basic_ostream<_CharT, _Traits>& |
2976 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
2977 | const piecewise_constant_distribution<_RealType>& __x) |
2978 | { |
2979 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
2980 | |
2981 | const typename __ios_base::fmtflags __flags = __os.flags(); |
2982 | const _CharT __fill = __os.fill(); |
2983 | const std::streamsize __precision = __os.precision(); |
2984 | const _CharT __space = __os.widen(' '); |
2985 | __os.flags(__ios_base::scientific | __ios_base::left); |
2986 | __os.fill(__space); |
2987 | __os.precision(std::numeric_limits<_RealType>::max_digits10); |
2988 | |
2989 | std::vector<_RealType> __int = __x.intervals(); |
2990 | __os << __int.size() - 1; |
2991 | |
2992 | for (auto __xit = __int.begin(); __xit != __int.end(); ++__xit) |
2993 | __os << __space << *__xit; |
2994 | |
2995 | std::vector<double> __den = __x.densities(); |
2996 | for (auto __dit = __den.begin(); __dit != __den.end(); ++__dit) |
2997 | __os << __space << *__dit; |
2998 | |
2999 | __os.flags(__flags); |
3000 | __os.fill(__fill); |
3001 | __os.precision(__precision); |
3002 | return __os; |
3003 | } |
3004 | |
3005 | template<typename _RealType, typename _CharT, typename _Traits> |
3006 | std::basic_istream<_CharT, _Traits>& |
3007 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
3008 | piecewise_constant_distribution<_RealType>& __x) |
3009 | { |
3010 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
3011 | |
3012 | const typename __ios_base::fmtflags __flags = __is.flags(); |
3013 | __is.flags(__ios_base::dec | __ios_base::skipws); |
3014 | |
3015 | size_t __n; |
3016 | if (__is >> __n) |
3017 | { |
3018 | std::vector<_RealType> __int_vec; |
3019 | if (__detail::__extract_params(__is, __int_vec, __n + 1)) |
3020 | { |
3021 | std::vector<double> __den_vec; |
3022 | if (__detail::__extract_params(__is, __den_vec, __n)) |
3023 | { |
3024 | __x.param({ __int_vec.begin(), __int_vec.end(), |
3025 | __den_vec.begin() }); |
3026 | } |
3027 | } |
3028 | } |
3029 | |
3030 | __is.flags(__flags); |
3031 | return __is; |
3032 | } |
3033 | |
3034 | |
3035 | template<typename _RealType> |
3036 | void |
3037 | piecewise_linear_distribution<_RealType>::param_type:: |
3038 | _M_initialize() |
3039 | { |
3040 | if (_M_int.size() < 2 |
3041 | || (_M_int.size() == 2 |
3042 | && _M_int[0] == _RealType(0) |
3043 | && _M_int[1] == _RealType(1) |
3044 | && _M_den[0] == _M_den[1])) |
3045 | { |
3046 | _M_int.clear(); |
3047 | _M_den.clear(); |
3048 | return; |
3049 | } |
3050 | |
3051 | double __sum = 0.0; |
3052 | _M_cp.reserve(n: _M_int.size() - 1); |
3053 | _M_m.reserve(n: _M_int.size() - 1); |
3054 | for (size_t __k = 0; __k < _M_int.size() - 1; ++__k) |
3055 | { |
3056 | const _RealType __delta = _M_int[__k + 1] - _M_int[__k]; |
3057 | __sum += 0.5 * (_M_den[__k + 1] + _M_den[__k]) * __delta; |
3058 | _M_cp.push_back(x: __sum); |
3059 | _M_m.push_back((_M_den[__k + 1] - _M_den[__k]) / __delta); |
3060 | } |
3061 | __glibcxx_assert(__sum > 0); |
3062 | |
3063 | // Now normalize the densities... |
3064 | __detail::__normalize(first: _M_den.begin(), last: _M_den.end(), result: _M_den.begin(), |
3065 | factor: __sum); |
3066 | // ... and partial sums... |
3067 | __detail::__normalize(first: _M_cp.begin(), last: _M_cp.end(), result: _M_cp.begin(), factor: __sum); |
3068 | // ... and slopes. |
3069 | __detail::__normalize(first: _M_m.begin(), last: _M_m.end(), result: _M_m.begin(), factor: __sum); |
3070 | |
3071 | // Make sure the last cumulative probablility is one. |
3072 | _M_cp[_M_cp.size() - 1] = 1.0; |
3073 | } |
3074 | |
3075 | template<typename _RealType> |
3076 | template<typename _InputIteratorB, typename _InputIteratorW> |
3077 | piecewise_linear_distribution<_RealType>::param_type:: |
3078 | param_type(_InputIteratorB __bbegin, |
3079 | _InputIteratorB __bend, |
3080 | _InputIteratorW __wbegin) |
3081 | : _M_int(), _M_den(), _M_cp(), _M_m() |
3082 | { |
3083 | for (; __bbegin != __bend; ++__bbegin, ++__wbegin) |
3084 | { |
3085 | _M_int.push_back(*__bbegin); |
3086 | _M_den.push_back(*__wbegin); |
3087 | } |
3088 | |
3089 | _M_initialize(); |
3090 | } |
3091 | |
3092 | template<typename _RealType> |
3093 | template<typename _Func> |
3094 | piecewise_linear_distribution<_RealType>::param_type:: |
3095 | param_type(initializer_list<_RealType> __bl, _Func __fw) |
3096 | : _M_int(), _M_den(), _M_cp(), _M_m() |
3097 | { |
3098 | _M_int.reserve(__bl.size()); |
3099 | _M_den.reserve(n: __bl.size()); |
3100 | for (auto __biter = __bl.begin(); __biter != __bl.end(); ++__biter) |
3101 | { |
3102 | _M_int.push_back(*__biter); |
3103 | _M_den.push_back(__fw(*__biter)); |
3104 | } |
3105 | |
3106 | _M_initialize(); |
3107 | } |
3108 | |
3109 | template<typename _RealType> |
3110 | template<typename _Func> |
3111 | piecewise_linear_distribution<_RealType>::param_type:: |
3112 | param_type(size_t __nw, _RealType __xmin, _RealType __xmax, _Func __fw) |
3113 | : _M_int(), _M_den(), _M_cp(), _M_m() |
3114 | { |
3115 | const size_t __n = __nw == 0 ? 1 : __nw; |
3116 | const _RealType __delta = (__xmax - __xmin) / __n; |
3117 | |
3118 | _M_int.reserve(__n + 1); |
3119 | _M_den.reserve(n: __n + 1); |
3120 | for (size_t __k = 0; __k <= __nw; ++__k) |
3121 | { |
3122 | _M_int.push_back(__xmin + __k * __delta); |
3123 | _M_den.push_back(__fw(_M_int[__k] + __delta)); |
3124 | } |
3125 | |
3126 | _M_initialize(); |
3127 | } |
3128 | |
3129 | template<typename _RealType> |
3130 | template<typename _UniformRandomNumberGenerator> |
3131 | typename piecewise_linear_distribution<_RealType>::result_type |
3132 | piecewise_linear_distribution<_RealType>:: |
3133 | operator()(_UniformRandomNumberGenerator& __urng, |
3134 | const param_type& __param) |
3135 | { |
3136 | __detail::_Adaptor<_UniformRandomNumberGenerator, double> |
3137 | __aurng(__urng); |
3138 | |
3139 | const double __p = __aurng(); |
3140 | if (__param._M_cp.empty()) |
3141 | return __p; |
3142 | |
3143 | auto __pos = std::lower_bound(__param._M_cp.begin(), |
3144 | __param._M_cp.end(), __p); |
3145 | const size_t __i = __pos - __param._M_cp.begin(); |
3146 | |
3147 | const double __pref = __i > 0 ? __param._M_cp[__i - 1] : 0.0; |
3148 | |
3149 | const double __a = 0.5 * __param._M_m[__i]; |
3150 | const double __b = __param._M_den[__i]; |
3151 | const double __cm = __p - __pref; |
3152 | |
3153 | _RealType __x = __param._M_int[__i]; |
3154 | if (__a == 0) |
3155 | __x += __cm / __b; |
3156 | else |
3157 | { |
3158 | const double __d = __b * __b + 4.0 * __a * __cm; |
3159 | __x += 0.5 * (std::sqrt(x: __d) - __b) / __a; |
3160 | } |
3161 | |
3162 | return __x; |
3163 | } |
3164 | |
3165 | template<typename _RealType> |
3166 | template<typename _ForwardIterator, |
3167 | typename _UniformRandomNumberGenerator> |
3168 | void |
3169 | piecewise_linear_distribution<_RealType>:: |
3170 | __generate_impl(_ForwardIterator __f, _ForwardIterator __t, |
3171 | _UniformRandomNumberGenerator& __urng, |
3172 | const param_type& __param) |
3173 | { |
3174 | __glibcxx_function_requires(_ForwardIteratorConcept<_ForwardIterator>) |
3175 | // We could duplicate everything from operator()... |
3176 | while (__f != __t) |
3177 | *__f++ = this->operator()(__urng, __param); |
3178 | } |
3179 | |
3180 | template<typename _RealType, typename _CharT, typename _Traits> |
3181 | std::basic_ostream<_CharT, _Traits>& |
3182 | operator<<(std::basic_ostream<_CharT, _Traits>& __os, |
3183 | const piecewise_linear_distribution<_RealType>& __x) |
3184 | { |
3185 | using __ios_base = typename basic_ostream<_CharT, _Traits>::ios_base; |
3186 | |
3187 | const typename __ios_base::fmtflags __flags = __os.flags(); |
3188 | const _CharT __fill = __os.fill(); |
3189 | const std::streamsize __precision = __os.precision(); |
3190 | const _CharT __space = __os.widen(' '); |
3191 | __os.flags(__ios_base::scientific | __ios_base::left); |
3192 | __os.fill(__space); |
3193 | __os.precision(std::numeric_limits<_RealType>::max_digits10); |
3194 | |
3195 | std::vector<_RealType> __int = __x.intervals(); |
3196 | __os << __int.size() - 1; |
3197 | |
3198 | for (auto __xit = __int.begin(); __xit != __int.end(); ++__xit) |
3199 | __os << __space << *__xit; |
3200 | |
3201 | std::vector<double> __den = __x.densities(); |
3202 | for (auto __dit = __den.begin(); __dit != __den.end(); ++__dit) |
3203 | __os << __space << *__dit; |
3204 | |
3205 | __os.flags(__flags); |
3206 | __os.fill(__fill); |
3207 | __os.precision(__precision); |
3208 | return __os; |
3209 | } |
3210 | |
3211 | template<typename _RealType, typename _CharT, typename _Traits> |
3212 | std::basic_istream<_CharT, _Traits>& |
3213 | operator>>(std::basic_istream<_CharT, _Traits>& __is, |
3214 | piecewise_linear_distribution<_RealType>& __x) |
3215 | { |
3216 | using __ios_base = typename basic_istream<_CharT, _Traits>::ios_base; |
3217 | |
3218 | const typename __ios_base::fmtflags __flags = __is.flags(); |
3219 | __is.flags(__ios_base::dec | __ios_base::skipws); |
3220 | |
3221 | size_t __n; |
3222 | if (__is >> __n) |
3223 | { |
3224 | vector<_RealType> __int_vec; |
3225 | if (__detail::__extract_params(__is, __int_vec, __n + 1)) |
3226 | { |
3227 | vector<double> __den_vec; |
3228 | if (__detail::__extract_params(__is, __den_vec, __n + 1)) |
3229 | { |
3230 | __x.param({ __int_vec.begin(), __int_vec.end(), |
3231 | __den_vec.begin() }); |
3232 | } |
3233 | } |
3234 | } |
3235 | __is.flags(__flags); |
3236 | return __is; |
3237 | } |
3238 | |
3239 | |
3240 | template<typename _IntType, typename> |
3241 | seed_seq::seed_seq(std::initializer_list<_IntType> __il) |
3242 | { |
3243 | _M_v.reserve(n: __il.size()); |
3244 | for (auto __iter = __il.begin(); __iter != __il.end(); ++__iter) |
3245 | _M_v.push_back(__detail::__mod<result_type, |
3246 | __detail::_Shift<result_type, 32>::__value>(*__iter)); |
3247 | } |
3248 | |
3249 | template<typename _InputIterator> |
3250 | seed_seq::seed_seq(_InputIterator __begin, _InputIterator __end) |
3251 | { |
3252 | if _GLIBCXX17_CONSTEXPR (__is_random_access_iter<_InputIterator>::value) |
3253 | _M_v.reserve(n: std::distance(__begin, __end)); |
3254 | |
3255 | for (_InputIterator __iter = __begin; __iter != __end; ++__iter) |
3256 | _M_v.push_back(__detail::__mod<result_type, |
3257 | __detail::_Shift<result_type, 32>::__value>(*__iter)); |
3258 | } |
3259 | |
3260 | template<typename _RandomAccessIterator> |
3261 | void |
3262 | seed_seq::generate(_RandomAccessIterator __begin, |
3263 | _RandomAccessIterator __end) |
3264 | { |
3265 | typedef typename iterator_traits<_RandomAccessIterator>::value_type |
3266 | _Type; |
3267 | |
3268 | if (__begin == __end) |
3269 | return; |
3270 | |
3271 | std::fill(__begin, __end, _Type(0x8b8b8b8bu)); |
3272 | |
3273 | const size_t __n = __end - __begin; |
3274 | const size_t __s = _M_v.size(); |
3275 | const size_t __t = (__n >= 623) ? 11 |
3276 | : (__n >= 68) ? 7 |
3277 | : (__n >= 39) ? 5 |
3278 | : (__n >= 7) ? 3 |
3279 | : (__n - 1) / 2; |
3280 | const size_t __p = (__n - __t) / 2; |
3281 | const size_t __q = __p + __t; |
3282 | const size_t __m = std::max(a: size_t(__s + 1), b: __n); |
3283 | |
3284 | #ifndef __UINT32_TYPE__ |
3285 | struct _Up |
3286 | { |
3287 | _Up(uint_least32_t v) : _M_v(v & 0xffffffffu) { } |
3288 | |
3289 | operator uint_least32_t() const { return _M_v; } |
3290 | |
3291 | uint_least32_t _M_v; |
3292 | }; |
3293 | using uint32_t = _Up; |
3294 | #endif |
3295 | |
3296 | // k == 0, every element in [begin,end) equals 0x8b8b8b8bu |
3297 | { |
3298 | uint32_t __r1 = 1371501266u; |
3299 | uint32_t __r2 = __r1 + __s; |
3300 | __begin[__p] += __r1; |
3301 | __begin[__q] = (uint32_t)__begin[__q] + __r2; |
3302 | __begin[0] = __r2; |
3303 | } |
3304 | |
3305 | for (size_t __k = 1; __k <= __s; ++__k) |
3306 | { |
3307 | const size_t __kn = __k % __n; |
3308 | const size_t __kpn = (__k + __p) % __n; |
3309 | const size_t __kqn = (__k + __q) % __n; |
3310 | uint32_t __arg = (__begin[__kn] |
3311 | ^ __begin[__kpn] |
3312 | ^ __begin[(__k - 1) % __n]); |
3313 | uint32_t __r1 = 1664525u * (__arg ^ (__arg >> 27)); |
3314 | uint32_t __r2 = __r1 + (uint32_t)__kn + _M_v[__k - 1]; |
3315 | __begin[__kpn] = (uint32_t)__begin[__kpn] + __r1; |
3316 | __begin[__kqn] = (uint32_t)__begin[__kqn] + __r2; |
3317 | __begin[__kn] = __r2; |
3318 | } |
3319 | |
3320 | for (size_t __k = __s + 1; __k < __m; ++__k) |
3321 | { |
3322 | const size_t __kn = __k % __n; |
3323 | const size_t __kpn = (__k + __p) % __n; |
3324 | const size_t __kqn = (__k + __q) % __n; |
3325 | uint32_t __arg = (__begin[__kn] |
3326 | ^ __begin[__kpn] |
3327 | ^ __begin[(__k - 1) % __n]); |
3328 | uint32_t __r1 = 1664525u * (__arg ^ (__arg >> 27)); |
3329 | uint32_t __r2 = __r1 + (uint32_t)__kn; |
3330 | __begin[__kpn] = (uint32_t)__begin[__kpn] + __r1; |
3331 | __begin[__kqn] = (uint32_t)__begin[__kqn] + __r2; |
3332 | __begin[__kn] = __r2; |
3333 | } |
3334 | |
3335 | for (size_t __k = __m; __k < __m + __n; ++__k) |
3336 | { |
3337 | const size_t __kn = __k % __n; |
3338 | const size_t __kpn = (__k + __p) % __n; |
3339 | const size_t __kqn = (__k + __q) % __n; |
3340 | uint32_t __arg = (__begin[__kn] |
3341 | + __begin[__kpn] |
3342 | + __begin[(__k - 1) % __n]); |
3343 | uint32_t __r3 = 1566083941u * (__arg ^ (__arg >> 27)); |
3344 | uint32_t __r4 = __r3 - __kn; |
3345 | __begin[__kpn] ^= __r3; |
3346 | __begin[__kqn] ^= __r4; |
3347 | __begin[__kn] = __r4; |
3348 | } |
3349 | } |
3350 | |
3351 | template<typename _RealType, size_t __bits, |
3352 | typename _UniformRandomNumberGenerator> |
3353 | _RealType |
3354 | generate_canonical(_UniformRandomNumberGenerator& __urng) |
3355 | { |
3356 | static_assert(std::is_floating_point<_RealType>::value, |
3357 | "template argument must be a floating point type" ); |
3358 | |
3359 | const size_t __b |
3360 | = std::min(a: static_cast<size_t>(std::numeric_limits<_RealType>::digits), |
3361 | b: __bits); |
3362 | const long double __r = static_cast<long double>(__urng.max()) |
3363 | - static_cast<long double>(__urng.min()) + 1.0L; |
3364 | const size_t __log2r = std::log(x: __r) / std::log(x: 2.0L); |
3365 | const size_t __m = std::max<size_t>(a: 1UL, |
3366 | b: (__b + __log2r - 1UL) / __log2r); |
3367 | _RealType __ret; |
3368 | _RealType __sum = _RealType(0); |
3369 | _RealType __tmp = _RealType(1); |
3370 | for (size_t __k = __m; __k != 0; --__k) |
3371 | { |
3372 | __sum += _RealType(__urng() - __urng.min()) * __tmp; |
3373 | __tmp *= __r; |
3374 | } |
3375 | __ret = __sum / __tmp; |
3376 | if (__builtin_expect(__ret >= _RealType(1), 0)) |
3377 | { |
3378 | #if _GLIBCXX_USE_C99_MATH_TR1 |
3379 | __ret = std::nextafter(_RealType(1), _RealType(0)); |
3380 | #else |
3381 | __ret = _RealType(1) |
3382 | - std::numeric_limits<_RealType>::epsilon() / _RealType(2); |
3383 | #endif |
3384 | } |
3385 | return __ret; |
3386 | } |
3387 | |
3388 | _GLIBCXX_END_NAMESPACE_VERSION |
3389 | } // namespace |
3390 | |
3391 | #endif |
3392 | |