1 | // Special functions -*- C++ -*-  |
2 |   |
3 | // Copyright (C) 2006-2019 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 tr1/gamma.tcc  |
26 | * This is an internal header file, included by other library headers.  |
27 | * Do not attempt to use it directly. @headername{tr1/cmath}  |
28 | */  |
29 |   |
30 | //  |
31 | // ISO C++ 14882 TR1: 5.2 Special functions  |
32 | //  |
33 |   |
34 | // Written by Edward Smith-Rowland based on:  |
35 | // (1) Handbook of Mathematical Functions,  |
36 | // ed. Milton Abramowitz and Irene A. Stegun,  |
37 | // Dover Publications,  |
38 | // Section 6, pp. 253-266  |
39 | // (2) The Gnu Scientific Library, http://www.gnu.org/software/gsl  |
40 | // (3) Numerical Recipes in C, by W. H. Press, S. A. Teukolsky,  |
41 | // W. T. Vetterling, B. P. Flannery, Cambridge University Press (1992),  |
42 | // 2nd ed, pp. 213-216  |
43 | // (4) Gamma, Exploring Euler's Constant, Julian Havil,  |
44 | // Princeton, 2003.  |
45 |   |
46 | #ifndef _GLIBCXX_TR1_GAMMA_TCC  |
47 | #define _GLIBCXX_TR1_GAMMA_TCC 1  |
48 |   |
49 | #include <tr1/special_function_util.h>  |
50 |   |
51 | namespace std _GLIBCXX_VISIBILITY(default)  |
52 | {  |
53 | _GLIBCXX_BEGIN_NAMESPACE_VERSION  |
54 |   |
55 | #if _GLIBCXX_USE_STD_SPEC_FUNCS  |
56 | # define _GLIBCXX_MATH_NS ::std  |
57 | #elif defined(_GLIBCXX_TR1_CMATH)  |
58 | namespace tr1  |
59 | {  |
60 | # define _GLIBCXX_MATH_NS ::std::tr1  |
61 | #else  |
62 | # error do not include this header directly, use <cmath> or <tr1/cmath>  |
63 | #endif  |
64 | // Implementation-space details.  |
65 | namespace __detail  |
66 | {  |
67 | /**  |
68 | * @brief This returns Bernoulli numbers from a table or by summation  |
69 | * for larger values.  |
70 | *  |
71 | * Recursion is unstable.  |
72 | *  |
73 | * @param __n the order n of the Bernoulli number.  |
74 | * @return The Bernoulli number of order n.  |
75 | */  |
76 | template <typename _Tp>  |
77 | _Tp  |
78 | __bernoulli_series(unsigned int __n)  |
79 | {  |
80 |   |
81 | static const _Tp __num[28] = {  |
82 | _Tp(1UL), -_Tp(1UL) / _Tp(2UL),  |
83 | _Tp(1UL) / _Tp(6UL), _Tp(0UL),  |
84 | -_Tp(1UL) / _Tp(30UL), _Tp(0UL),  |
85 | _Tp(1UL) / _Tp(42UL), _Tp(0UL),  |
86 | -_Tp(1UL) / _Tp(30UL), _Tp(0UL),  |
87 | _Tp(5UL) / _Tp(66UL), _Tp(0UL),  |
88 | -_Tp(691UL) / _Tp(2730UL), _Tp(0UL),  |
89 | _Tp(7UL) / _Tp(6UL), _Tp(0UL),  |
90 | -_Tp(3617UL) / _Tp(510UL), _Tp(0UL),  |
91 | _Tp(43867UL) / _Tp(798UL), _Tp(0UL),  |
92 | -_Tp(174611) / _Tp(330UL), _Tp(0UL),  |
93 | _Tp(854513UL) / _Tp(138UL), _Tp(0UL),  |
94 | -_Tp(236364091UL) / _Tp(2730UL), _Tp(0UL),  |
95 | _Tp(8553103UL) / _Tp(6UL), _Tp(0UL)  |
96 | };  |
97 |   |
98 | if (__n == 0)  |
99 | return _Tp(1);  |
100 |   |
101 | if (__n == 1)  |
102 | return -_Tp(1) / _Tp(2);  |
103 |   |
104 | // Take care of the rest of the odd ones.  |
105 | if (__n % 2 == 1)  |
106 | return _Tp(0);  |
107 |   |
108 | // Take care of some small evens that are painful for the series.  |
109 | if (__n < 28)  |
110 | return __num[__n];  |
111 |   |
112 |   |
113 | _Tp __fact = _Tp(1);  |
114 | if ((__n / 2) % 2 == 0)  |
115 | __fact *= _Tp(-1);  |
116 | for (unsigned int __k = 1; __k <= __n; ++__k)  |
117 | __fact *= __k / (_Tp(2) * __numeric_constants<_Tp>::__pi());  |
118 | __fact *= _Tp(2);  |
119 |   |
120 | _Tp __sum = _Tp(0);  |
121 | for (unsigned int __i = 1; __i < 1000; ++__i)  |
122 | {  |
123 | _Tp __term = std::pow(_Tp(__i), -_Tp(__n));  |
124 | if (__term < std::numeric_limits<_Tp>::epsilon())  |
125 | break;  |
126 | __sum += __term;  |
127 | }  |
128 |   |
129 | return __fact * __sum;  |
130 | }  |
131 |   |
132 |   |
133 | /**  |
134 | * @brief This returns Bernoulli number \f$B_n\f$.  |
135 | *  |
136 | * @param __n the order n of the Bernoulli number.  |
137 | * @return The Bernoulli number of order n.  |
138 | */  |
139 | template<typename _Tp>  |
140 | inline _Tp  |
141 | __bernoulli(int __n)  |
142 | { return __bernoulli_series<_Tp>(__n); }  |
143 |   |
144 |   |
145 | /**  |
146 | * @brief Return \f$log(\Gamma(x))\f$ by asymptotic expansion  |
147 | * with Bernoulli number coefficients. This is like  |
148 | * Sterling's approximation.  |
149 | *  |
150 | * @param __x The argument of the log of the gamma function.  |
151 | * @return The logarithm of the gamma function.  |
152 | */  |
153 | template<typename _Tp>  |
154 | _Tp  |
155 | __log_gamma_bernoulli(_Tp __x)  |
156 | {  |
157 | _Tp __lg = (__x - _Tp(0.5L)) * std::log(__x) - __x  |
158 | + _Tp(0.5L) * std::log(_Tp(2)  |
159 | * __numeric_constants<_Tp>::__pi());  |
160 |   |
161 | const _Tp __xx = __x * __x;  |
162 | _Tp __help = _Tp(1) / __x;  |
163 | for ( unsigned int __i = 1; __i < 20; ++__i )  |
164 | {  |
165 | const _Tp __2i = _Tp(2 * __i);  |
166 | __help /= __2i * (__2i - _Tp(1)) * __xx;  |
167 | __lg += __bernoulli<_Tp>(2 * __i) * __help;  |
168 | }  |
169 |   |
170 | return __lg;  |
171 | }  |
172 |   |
173 |   |
174 | /**  |
175 | * @brief Return \f$log(\Gamma(x))\f$ by the Lanczos method.  |
176 | * This method dominates all others on the positive axis I think.  |
177 | *  |
178 | * @param __x The argument of the log of the gamma function.  |
179 | * @return The logarithm of the gamma function.  |
180 | */  |
181 | template<typename _Tp>  |
182 | _Tp  |
183 | __log_gamma_lanczos(_Tp __x)  |
184 | {  |
185 | const _Tp __xm1 = __x - _Tp(1);  |
186 |   |
187 | static const _Tp __lanczos_cheb_7[9] = {  |
188 | _Tp( 0.99999999999980993227684700473478L),  |
189 | _Tp( 676.520368121885098567009190444019L),  |
190 | _Tp(-1259.13921672240287047156078755283L),  |
191 | _Tp( 771.3234287776530788486528258894L),  |
192 | _Tp(-176.61502916214059906584551354L),  |
193 | _Tp( 12.507343278686904814458936853L),  |
194 | _Tp(-0.13857109526572011689554707L),  |
195 | _Tp( 9.984369578019570859563e-6L),  |
196 | _Tp( 1.50563273514931155834e-7L)  |
197 | };  |
198 |   |
199 | static const _Tp __LOGROOT2PI  |
200 | = _Tp(0.9189385332046727417803297364056176L);  |
201 |   |
202 | _Tp __sum = __lanczos_cheb_7[0];  |
203 | for(unsigned int __k = 1; __k < 9; ++__k)  |
204 | __sum += __lanczos_cheb_7[__k] / (__xm1 + __k);  |
205 |   |
206 | const _Tp __term1 = (__xm1 + _Tp(0.5L))  |
207 | * std::log((__xm1 + _Tp(7.5L))  |
208 | / __numeric_constants<_Tp>::__euler());  |
209 | const _Tp __term2 = __LOGROOT2PI + std::log(__sum);  |
210 | const _Tp __result = __term1 + (__term2 - _Tp(7));  |
211 |   |
212 | return __result;  |
213 | }  |
214 |   |
215 |   |
216 | /**  |
217 | * @brief Return \f$ log(|\Gamma(x)|) \f$.  |
218 | * This will return values even for \f$ x < 0 \f$.  |
219 | * To recover the sign of \f$ \Gamma(x) \f$ for  |
220 | * any argument use @a __log_gamma_sign.  |
221 | *  |
222 | * @param __x The argument of the log of the gamma function.  |
223 | * @return The logarithm of the gamma function.  |
224 | */  |
225 | template<typename _Tp>  |
226 | _Tp  |
227 | __log_gamma(_Tp __x)  |
228 | {  |
229 | if (__x > _Tp(0.5L))  |
230 | return __log_gamma_lanczos(__x);  |
231 | else  |
232 | {  |
233 | const _Tp __sin_fact  |
234 | = std::abs(std::sin(__numeric_constants<_Tp>::__pi() * __x));  |
235 | if (__sin_fact == _Tp(0))  |
236 | std::__throw_domain_error(__N("Argument is nonpositive integer "   |
237 | "in __log_gamma" ));  |
238 | return __numeric_constants<_Tp>::__lnpi()  |
239 | - std::log(__sin_fact)  |
240 | - __log_gamma_lanczos(_Tp(1) - __x);  |
241 | }  |
242 | }  |
243 |   |
244 |   |
245 | /**  |
246 | * @brief Return the sign of \f$ \Gamma(x) \f$.  |
247 | * At nonpositive integers zero is returned.  |
248 | *  |
249 | * @param __x The argument of the gamma function.  |
250 | * @return The sign of the gamma function.  |
251 | */  |
252 | template<typename _Tp>  |
253 | _Tp  |
254 | __log_gamma_sign(_Tp __x)  |
255 | {  |
256 | if (__x > _Tp(0))  |
257 | return _Tp(1);  |
258 | else  |
259 | {  |
260 | const _Tp __sin_fact  |
261 | = std::sin(__numeric_constants<_Tp>::__pi() * __x);  |
262 | if (__sin_fact > _Tp(0))  |
263 | return (1);  |
264 | else if (__sin_fact < _Tp(0))  |
265 | return -_Tp(1);  |
266 | else  |
267 | return _Tp(0);  |
268 | }  |
269 | }  |
270 |   |
271 |   |
272 | /**  |
273 | * @brief Return the logarithm of the binomial coefficient.  |
274 | * The binomial coefficient is given by:  |
275 | * @f[  |
276 | * \left( \right) = \frac{n!}{(n-k)! k!}  |
277 | * @f]  |
278 | *  |
279 | * @param __n The first argument of the binomial coefficient.  |
280 | * @param __k The second argument of the binomial coefficient.  |
281 | * @return The binomial coefficient.  |
282 | */  |
283 | template<typename _Tp>  |
284 | _Tp  |
285 | __log_bincoef(unsigned int __n, unsigned int __k)  |
286 | {  |
287 | // Max e exponent before overflow.  |
288 | static const _Tp __max_bincoeff  |
289 | = std::numeric_limits<_Tp>::max_exponent10  |
290 | * std::log(_Tp(10)) - _Tp(1);  |
291 | #if _GLIBCXX_USE_C99_MATH_TR1  |
292 | _Tp __coeff = _GLIBCXX_MATH_NS::lgamma(_Tp(1 + __n))  |
293 | - _GLIBCXX_MATH_NS::lgamma(_Tp(1 + __k))  |
294 | - _GLIBCXX_MATH_NS::lgamma(_Tp(1 + __n - __k));  |
295 | #else  |
296 | _Tp __coeff = __log_gamma(_Tp(1 + __n))  |
297 | - __log_gamma(_Tp(1 + __k))  |
298 | - __log_gamma(_Tp(1 + __n - __k));  |
299 | #endif  |
300 | }  |
301 |   |
302 |   |
303 | /**  |
304 | * @brief Return the binomial coefficient.  |
305 | * The binomial coefficient is given by:  |
306 | * @f[  |
307 | * \left( \right) = \frac{n!}{(n-k)! k!}  |
308 | * @f]  |
309 | *  |
310 | * @param __n The first argument of the binomial coefficient.  |
311 | * @param __k The second argument of the binomial coefficient.  |
312 | * @return The binomial coefficient.  |
313 | */  |
314 | template<typename _Tp>  |
315 | _Tp  |
316 | __bincoef(unsigned int __n, unsigned int __k)  |
317 | {  |
318 | // Max e exponent before overflow.  |
319 | static const _Tp __max_bincoeff  |
320 | = std::numeric_limits<_Tp>::max_exponent10  |
321 | * std::log(_Tp(10)) - _Tp(1);  |
322 |   |
323 | const _Tp __log_coeff = __log_bincoef<_Tp>(__n, __k);  |
324 | if (__log_coeff > __max_bincoeff)  |
325 | return std::numeric_limits<_Tp>::quiet_NaN();  |
326 | else  |
327 | return std::exp(__log_coeff);  |
328 | }  |
329 |   |
330 |   |
331 | /**  |
332 | * @brief Return \f$ \Gamma(x) \f$.  |
333 | *  |
334 | * @param __x The argument of the gamma function.  |
335 | * @return The gamma function.  |
336 | */  |
337 | template<typename _Tp>  |
338 | inline _Tp  |
339 | __gamma(_Tp __x)  |
340 | { return std::exp(__log_gamma(__x)); }  |
341 |   |
342 |   |
343 | /**  |
344 | * @brief Return the digamma function by series expansion.  |
345 | * The digamma or @f$ \psi(x) @f$ function is defined by  |
346 | * @f[  |
347 | * \psi(x) = \frac{\Gamma'(x)}{\Gamma(x)}  |
348 | * @f]  |
349 | *  |
350 | * The series is given by:  |
351 | * @f[  |
352 | * \psi(x) = -\gamma_E - \frac{1}{x}  |
353 | * \sum_{k=1}^{\infty} \frac{x}{k(x + k)}  |
354 | * @f]  |
355 | */  |
356 | template<typename _Tp>  |
357 | _Tp  |
358 | __psi_series(_Tp __x)  |
359 | {  |
360 | _Tp __sum = -__numeric_constants<_Tp>::__gamma_e() - _Tp(1) / __x;  |
361 | const unsigned int __max_iter = 100000;  |
362 | for (unsigned int __k = 1; __k < __max_iter; ++__k)  |
363 | {  |
364 | const _Tp __term = __x / (__k * (__k + __x));  |
365 | __sum += __term;  |
366 | if (std::abs(__term / __sum) < std::numeric_limits<_Tp>::epsilon())  |
367 | break;  |
368 | }  |
369 | return __sum;  |
370 | }  |
371 |   |
372 |   |
373 | /**  |
374 | * @brief Return the digamma function for large argument.  |
375 | * The digamma or @f$ \psi(x) @f$ function is defined by  |
376 | * @f[  |
377 | * \psi(x) = \frac{\Gamma'(x)}{\Gamma(x)}  |
378 | * @f]  |
379 | *  |
380 | * The asymptotic series is given by:  |
381 | * @f[  |
382 | * \psi(x) = \ln(x) - \frac{1}{2x}  |
383 | * - \sum_{n=1}^{\infty} \frac{B_{2n}}{2 n x^{2n}}  |
384 | * @f]  |
385 | */  |
386 | template<typename _Tp>  |
387 | _Tp  |
388 | __psi_asymp(_Tp __x)  |
389 | {  |
390 | _Tp __sum = std::log(__x) - _Tp(0.5L) / __x;  |
391 | const _Tp __xx = __x * __x;  |
392 | _Tp __xp = __xx;  |
393 | const unsigned int __max_iter = 100;  |
394 | for (unsigned int __k = 1; __k < __max_iter; ++__k)  |
395 | {  |
396 | const _Tp __term = __bernoulli<_Tp>(2 * __k) / (2 * __k * __xp);  |
397 | __sum -= __term;  |
398 | if (std::abs(__term / __sum) < std::numeric_limits<_Tp>::epsilon())  |
399 | break;  |
400 | __xp *= __xx;  |
401 | }  |
402 | return __sum;  |
403 | }  |
404 |   |
405 |   |
406 | /**  |
407 | * @brief Return the digamma function.  |
408 | * The digamma or @f$ \psi(x) @f$ function is defined by  |
409 | * @f[  |
410 | * \psi(x) = \frac{\Gamma'(x)}{\Gamma(x)}  |
411 | * @f]  |
412 | * For negative argument the reflection formula is used:  |
413 | * @f[  |
414 | * \psi(x) = \psi(1-x) - \pi \cot(\pi x)  |
415 | * @f]  |
416 | */  |
417 | template<typename _Tp>  |
418 | _Tp  |
419 | __psi(_Tp __x)  |
420 | {  |
421 | const int __n = static_cast<int>(__x + 0.5L);  |
422 | const _Tp __eps = _Tp(4) * std::numeric_limits<_Tp>::epsilon();  |
423 | if (__n <= 0 && std::abs(__x - _Tp(__n)) < __eps)  |
424 | return std::numeric_limits<_Tp>::quiet_NaN();  |
425 | else if (__x < _Tp(0))  |
426 | {  |
427 | const _Tp __pi = __numeric_constants<_Tp>::__pi();  |
428 | return __psi(_Tp(1) - __x)  |
429 | - __pi * std::cos(__pi * __x) / std::sin(__pi * __x);  |
430 | }  |
431 | else if (__x > _Tp(100))  |
432 | return __psi_asymp(__x);  |
433 | else  |
434 | return __psi_series(__x);  |
435 | }  |
436 |   |
437 |   |
438 | /**  |
439 | * @brief Return the polygamma function @f$ \psi^{(n)}(x) @f$.  |
440 | *   |
441 | * The polygamma function is related to the Hurwitz zeta function:  |
442 | * @f[  |
443 | * \psi^{(n)}(x) = (-1)^{n+1} m! \zeta(m+1,x)  |
444 | * @f]  |
445 | */  |
446 | template<typename _Tp>  |
447 | _Tp  |
448 | __psi(unsigned int __n, _Tp __x)  |
449 | {  |
450 | if (__x <= _Tp(0))  |
451 | std::__throw_domain_error(__N("Argument out of range "   |
452 | "in __psi" ));  |
453 | else if (__n == 0)  |
454 | return __psi(__x);  |
455 | else  |
456 | {  |
457 | const _Tp __hzeta = __hurwitz_zeta(_Tp(__n + 1), __x);  |
458 | #if _GLIBCXX_USE_C99_MATH_TR1  |
459 | const _Tp __ln_nfact = _GLIBCXX_MATH_NS::lgamma(_Tp(__n + 1));  |
460 | #else  |
461 | const _Tp __ln_nfact = __log_gamma(_Tp(__n + 1));  |
462 | #endif  |
463 | _Tp __result = std::exp(__ln_nfact) * __hzeta;  |
464 | if (__n % 2 == 1)  |
465 | __result = -__result;  |
466 | return __result;  |
467 | }  |
468 | }  |
469 | } // namespace __detail  |
470 | #undef _GLIBCXX_MATH_NS  |
471 | #if ! _GLIBCXX_USE_STD_SPEC_FUNCS && defined(_GLIBCXX_TR1_CMATH)  |
472 | } // namespace tr1  |
473 | #endif  |
474 |   |
475 | _GLIBCXX_END_NAMESPACE_VERSION  |
476 | } // namespace std  |
477 |   |
478 | #endif // _GLIBCXX_TR1_GAMMA_TCC  |
479 |   |
480 | |