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/hypergeometric.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:  |
35 | // (1) Handbook of Mathematical Functions,  |
36 | // ed. Milton Abramowitz and Irene A. Stegun,  |
37 | // Dover Publications,  |
38 | // Section 6, pp. 555-566  |
39 | // (2) The Gnu Scientific Library, http://www.gnu.org/software/gsl  |
40 |   |
41 | #ifndef _GLIBCXX_TR1_HYPERGEOMETRIC_TCC  |
42 | #define _GLIBCXX_TR1_HYPERGEOMETRIC_TCC 1  |
43 |   |
44 | namespace std _GLIBCXX_VISIBILITY(default)  |
45 | {  |
46 | _GLIBCXX_BEGIN_NAMESPACE_VERSION  |
47 |   |
48 | #if _GLIBCXX_USE_STD_SPEC_FUNCS  |
49 | # define _GLIBCXX_MATH_NS ::std  |
50 | #elif defined(_GLIBCXX_TR1_CMATH)  |
51 | namespace tr1  |
52 | {  |
53 | # define _GLIBCXX_MATH_NS ::std::tr1  |
54 | #else  |
55 | # error do not include this header directly, use <cmath> or <tr1/cmath>  |
56 | #endif  |
57 | // [5.2] Special functions  |
58 |   |
59 | // Implementation-space details.  |
60 | namespace __detail  |
61 | {  |
62 | /**  |
63 | * @brief This routine returns the confluent hypergeometric function  |
64 | * by series expansion.  |
65 | *   |
66 | * @f[  |
67 | * _1F_1(a;c;x) = \frac{\Gamma(c)}{\Gamma(a)}  |
68 | * \sum_{n=0}^{\infty}  |
69 | * \frac{\Gamma(a+n)}{\Gamma(c+n)}  |
70 | * \frac{x^n}{n!}  |
71 | * @f]  |
72 | *   |
73 | * If a and b are integers and a < 0 and either b > 0 or b < a  |
74 | * then the series is a polynomial with a finite number of  |
75 | * terms. If b is an integer and b <= 0 the confluent  |
76 | * hypergeometric function is undefined.  |
77 | *  |
78 | * @param __a The "numerator" parameter.  |
79 | * @param __c The "denominator" parameter.  |
80 | * @param __x The argument of the confluent hypergeometric function.  |
81 | * @return The confluent hypergeometric function.  |
82 | */  |
83 | template<typename _Tp>  |
84 | _Tp  |
85 | __conf_hyperg_series(_Tp __a, _Tp __c, _Tp __x)  |
86 | {  |
87 | const _Tp __eps = std::numeric_limits<_Tp>::epsilon();  |
88 |   |
89 | _Tp __term = _Tp(1);  |
90 | _Tp __Fac = _Tp(1);  |
91 | const unsigned int __max_iter = 100000;  |
92 | unsigned int __i;  |
93 | for (__i = 0; __i < __max_iter; ++__i)  |
94 | {  |
95 | __term *= (__a + _Tp(__i)) * __x  |
96 | / ((__c + _Tp(__i)) * _Tp(1 + __i));  |
97 | if (std::abs(__term) < __eps)  |
98 | {  |
99 | break;  |
100 | }  |
101 | __Fac += __term;  |
102 | }  |
103 | if (__i == __max_iter)  |
104 | std::__throw_runtime_error(__N("Series failed to converge "   |
105 | "in __conf_hyperg_series." ));  |
106 |   |
107 | return __Fac;  |
108 | }  |
109 |   |
110 |   |
111 | /**  |
112 | * @brief Return the hypogeometric function @f$ _2F_1(a,b;c;x) @f$  |
113 | * by an iterative procedure described in  |
114 | * Luke, Algorithms for the Computation of Mathematical Functions.  |
115 | *  |
116 | * Like the case of the 2F1 rational approximations, these are   |
117 | * probably guaranteed to converge for x < 0, barring gross   |
118 | * numerical instability in the pre-asymptotic regime.   |
119 | */  |
120 | template<typename _Tp>  |
121 | _Tp  |
122 | __conf_hyperg_luke(_Tp __a, _Tp __c, _Tp __xin)  |
123 | {  |
124 | const _Tp __big = std::pow(std::numeric_limits<_Tp>::max(), _Tp(0.16L));  |
125 | const int __nmax = 20000;  |
126 | const _Tp __eps = std::numeric_limits<_Tp>::epsilon();  |
127 | const _Tp __x = -__xin;  |
128 | const _Tp __x3 = __x * __x * __x;  |
129 | const _Tp __t0 = __a / __c;  |
130 | const _Tp __t1 = (__a + _Tp(1)) / (_Tp(2) * __c);  |
131 | const _Tp __t2 = (__a + _Tp(2)) / (_Tp(2) * (__c + _Tp(1)));  |
132 | _Tp __F = _Tp(1);  |
133 | _Tp __prec;  |
134 |   |
135 | _Tp __Bnm3 = _Tp(1);  |
136 | _Tp __Bnm2 = _Tp(1) + __t1 * __x;  |
137 | _Tp __Bnm1 = _Tp(1) + __t2 * __x * (_Tp(1) + __t1 / _Tp(3) * __x);  |
138 |   |
139 | _Tp __Anm3 = _Tp(1);  |
140 | _Tp __Anm2 = __Bnm2 - __t0 * __x;  |
141 | _Tp __Anm1 = __Bnm1 - __t0 * (_Tp(1) + __t2 * __x) * __x  |
142 | + __t0 * __t1 * (__c / (__c + _Tp(1))) * __x * __x;  |
143 |   |
144 | int __n = 3;  |
145 | while(1)  |
146 | {  |
147 | _Tp __npam1 = _Tp(__n - 1) + __a;  |
148 | _Tp __npcm1 = _Tp(__n - 1) + __c;  |
149 | _Tp __npam2 = _Tp(__n - 2) + __a;  |
150 | _Tp __npcm2 = _Tp(__n - 2) + __c;  |
151 | _Tp __tnm1 = _Tp(2 * __n - 1);  |
152 | _Tp __tnm3 = _Tp(2 * __n - 3);  |
153 | _Tp __tnm5 = _Tp(2 * __n - 5);  |
154 | _Tp __F1 = (_Tp(__n - 2) - __a) / (_Tp(2) * __tnm3 * __npcm1);  |
155 | _Tp __F2 = (_Tp(__n) + __a) * __npam1  |
156 | / (_Tp(4) * __tnm1 * __tnm3 * __npcm2 * __npcm1);  |
157 | _Tp __F3 = -__npam2 * __npam1 * (_Tp(__n - 2) - __a)  |
158 | / (_Tp(8) * __tnm3 * __tnm3 * __tnm5  |
159 | * (_Tp(__n - 3) + __c) * __npcm2 * __npcm1);  |
160 | _Tp __E = -__npam1 * (_Tp(__n - 1) - __c)  |
161 | / (_Tp(2) * __tnm3 * __npcm2 * __npcm1);  |
162 |   |
163 | _Tp __An = (_Tp(1) + __F1 * __x) * __Anm1  |
164 | + (__E + __F2 * __x) * __x * __Anm2 + __F3 * __x3 * __Anm3;  |
165 | _Tp __Bn = (_Tp(1) + __F1 * __x) * __Bnm1  |
166 | + (__E + __F2 * __x) * __x * __Bnm2 + __F3 * __x3 * __Bnm3;  |
167 | _Tp __r = __An / __Bn;  |
168 |   |
169 | __prec = std::abs((__F - __r) / __F);  |
170 | __F = __r;  |
171 |   |
172 | if (__prec < __eps || __n > __nmax)  |
173 | break;  |
174 |   |
175 | if (std::abs(__An) > __big || std::abs(__Bn) > __big)  |
176 | {  |
177 | __An /= __big;  |
178 | __Bn /= __big;  |
179 | __Anm1 /= __big;  |
180 | __Bnm1 /= __big;  |
181 | __Anm2 /= __big;  |
182 | __Bnm2 /= __big;  |
183 | __Anm3 /= __big;  |
184 | __Bnm3 /= __big;  |
185 | }  |
186 | else if (std::abs(__An) < _Tp(1) / __big  |
187 | || std::abs(__Bn) < _Tp(1) / __big)  |
188 | {  |
189 | __An *= __big;  |
190 | __Bn *= __big;  |
191 | __Anm1 *= __big;  |
192 | __Bnm1 *= __big;  |
193 | __Anm2 *= __big;  |
194 | __Bnm2 *= __big;  |
195 | __Anm3 *= __big;  |
196 | __Bnm3 *= __big;  |
197 | }  |
198 |   |
199 | ++__n;  |
200 | __Bnm3 = __Bnm2;  |
201 | __Bnm2 = __Bnm1;  |
202 | __Bnm1 = __Bn;  |
203 | __Anm3 = __Anm2;  |
204 | __Anm2 = __Anm1;  |
205 | __Anm1 = __An;  |
206 | }  |
207 |   |
208 | if (__n >= __nmax)  |
209 | std::__throw_runtime_error(__N("Iteration failed to converge "   |
210 | "in __conf_hyperg_luke." ));  |
211 |   |
212 | return __F;  |
213 | }  |
214 |   |
215 |   |
216 | /**  |
217 | * @brief Return the confluent hypogeometric function  |
218 | * @f$ _1F_1(a;c;x) @f$.  |
219 | *   |
220 | * @todo Handle b == nonpositive integer blowup - return NaN.  |
221 | *  |
222 | * @param __a The @a numerator parameter.  |
223 | * @param __c The @a denominator parameter.  |
224 | * @param __x The argument of the confluent hypergeometric function.  |
225 | * @return The confluent hypergeometric function.  |
226 | */  |
227 | template<typename _Tp>  |
228 | _Tp  |
229 | __conf_hyperg(_Tp __a, _Tp __c, _Tp __x)  |
230 | {  |
231 | #if _GLIBCXX_USE_C99_MATH_TR1  |
232 | const _Tp __c_nint = _GLIBCXX_MATH_NS::nearbyint(__c);  |
233 | #else  |
234 | const _Tp __c_nint = static_cast<int>(__c + _Tp(0.5L));  |
235 | #endif  |
236 | if (__isnan(__a) || __isnan(__c) || __isnan(__x))  |
237 | return std::numeric_limits<_Tp>::quiet_NaN();  |
238 | else if (__c_nint == __c && __c_nint <= 0)  |
239 | return std::numeric_limits<_Tp>::infinity();  |
240 | else if (__a == _Tp(0))  |
241 | return _Tp(1);  |
242 | else if (__c == __a)  |
243 | return std::exp(__x);  |
244 | else if (__x < _Tp(0))  |
245 | return __conf_hyperg_luke(__a, __c, __x);  |
246 | else  |
247 | return __conf_hyperg_series(__a, __c, __x);  |
248 | }  |
249 |   |
250 |   |
251 | /**  |
252 | * @brief Return the hypogeometric function @f$ _2F_1(a,b;c;x) @f$  |
253 | * by series expansion.  |
254 | *   |
255 | * The hypogeometric function is defined by  |
256 | * @f[  |
257 | * _2F_1(a,b;c;x) = \frac{\Gamma(c)}{\Gamma(a)\Gamma(b)}  |
258 | * \sum_{n=0}^{\infty}  |
259 | * \frac{\Gamma(a+n)\Gamma(b+n)}{\Gamma(c+n)}  |
260 | * \frac{x^n}{n!}  |
261 | * @f]  |
262 | *   |
263 | * This works and it's pretty fast.  |
264 | *  |
265 | * @param __a The first @a numerator parameter.  |
266 | * @param __a The second @a numerator parameter.  |
267 | * @param __c The @a denominator parameter.  |
268 | * @param __x The argument of the confluent hypergeometric function.  |
269 | * @return The confluent hypergeometric function.  |
270 | */  |
271 | template<typename _Tp>  |
272 | _Tp  |
273 | __hyperg_series(_Tp __a, _Tp __b, _Tp __c, _Tp __x)  |
274 | {  |
275 | const _Tp __eps = std::numeric_limits<_Tp>::epsilon();  |
276 |   |
277 | _Tp __term = _Tp(1);  |
278 | _Tp __Fabc = _Tp(1);  |
279 | const unsigned int __max_iter = 100000;  |
280 | unsigned int __i;  |
281 | for (__i = 0; __i < __max_iter; ++__i)  |
282 | {  |
283 | __term *= (__a + _Tp(__i)) * (__b + _Tp(__i)) * __x  |
284 | / ((__c + _Tp(__i)) * _Tp(1 + __i));  |
285 | if (std::abs(__term) < __eps)  |
286 | {  |
287 | break;  |
288 | }  |
289 | __Fabc += __term;  |
290 | }  |
291 | if (__i == __max_iter)  |
292 | std::__throw_runtime_error(__N("Series failed to converge "   |
293 | "in __hyperg_series." ));  |
294 |   |
295 | return __Fabc;  |
296 | }  |
297 |   |
298 |   |
299 | /**  |
300 | * @brief Return the hypogeometric function @f$ _2F_1(a,b;c;x) @f$  |
301 | * by an iterative procedure described in  |
302 | * Luke, Algorithms for the Computation of Mathematical Functions.  |
303 | */  |
304 | template<typename _Tp>  |
305 | _Tp  |
306 | __hyperg_luke(_Tp __a, _Tp __b, _Tp __c, _Tp __xin)  |
307 | {  |
308 | const _Tp __big = std::pow(std::numeric_limits<_Tp>::max(), _Tp(0.16L));  |
309 | const int __nmax = 20000;  |
310 | const _Tp __eps = std::numeric_limits<_Tp>::epsilon();  |
311 | const _Tp __x = -__xin;  |
312 | const _Tp __x3 = __x * __x * __x;  |
313 | const _Tp __t0 = __a * __b / __c;  |
314 | const _Tp __t1 = (__a + _Tp(1)) * (__b + _Tp(1)) / (_Tp(2) * __c);  |
315 | const _Tp __t2 = (__a + _Tp(2)) * (__b + _Tp(2))  |
316 | / (_Tp(2) * (__c + _Tp(1)));  |
317 |   |
318 | _Tp __F = _Tp(1);  |
319 |   |
320 | _Tp __Bnm3 = _Tp(1);  |
321 | _Tp __Bnm2 = _Tp(1) + __t1 * __x;  |
322 | _Tp __Bnm1 = _Tp(1) + __t2 * __x * (_Tp(1) + __t1 / _Tp(3) * __x);  |
323 |   |
324 | _Tp __Anm3 = _Tp(1);  |
325 | _Tp __Anm2 = __Bnm2 - __t0 * __x;  |
326 | _Tp __Anm1 = __Bnm1 - __t0 * (_Tp(1) + __t2 * __x) * __x  |
327 | + __t0 * __t1 * (__c / (__c + _Tp(1))) * __x * __x;  |
328 |   |
329 | int __n = 3;  |
330 | while (1)  |
331 | {  |
332 | const _Tp __npam1 = _Tp(__n - 1) + __a;  |
333 | const _Tp __npbm1 = _Tp(__n - 1) + __b;  |
334 | const _Tp __npcm1 = _Tp(__n - 1) + __c;  |
335 | const _Tp __npam2 = _Tp(__n - 2) + __a;  |
336 | const _Tp __npbm2 = _Tp(__n - 2) + __b;  |
337 | const _Tp __npcm2 = _Tp(__n - 2) + __c;  |
338 | const _Tp __tnm1 = _Tp(2 * __n - 1);  |
339 | const _Tp __tnm3 = _Tp(2 * __n - 3);  |
340 | const _Tp __tnm5 = _Tp(2 * __n - 5);  |
341 | const _Tp __n2 = __n * __n;  |
342 | const _Tp __F1 = (_Tp(3) * __n2 + (__a + __b - _Tp(6)) * __n  |
343 | + _Tp(2) - __a * __b - _Tp(2) * (__a + __b))  |
344 | / (_Tp(2) * __tnm3 * __npcm1);  |
345 | const _Tp __F2 = -(_Tp(3) * __n2 - (__a + __b + _Tp(6)) * __n  |
346 | + _Tp(2) - __a * __b) * __npam1 * __npbm1  |
347 | / (_Tp(4) * __tnm1 * __tnm3 * __npcm2 * __npcm1);  |
348 | const _Tp __F3 = (__npam2 * __npam1 * __npbm2 * __npbm1  |
349 | * (_Tp(__n - 2) - __a) * (_Tp(__n - 2) - __b))  |
350 | / (_Tp(8) * __tnm3 * __tnm3 * __tnm5  |
351 | * (_Tp(__n - 3) + __c) * __npcm2 * __npcm1);  |
352 | const _Tp __E = -__npam1 * __npbm1 * (_Tp(__n - 1) - __c)  |
353 | / (_Tp(2) * __tnm3 * __npcm2 * __npcm1);  |
354 |   |
355 | _Tp __An = (_Tp(1) + __F1 * __x) * __Anm1  |
356 | + (__E + __F2 * __x) * __x * __Anm2 + __F3 * __x3 * __Anm3;  |
357 | _Tp __Bn = (_Tp(1) + __F1 * __x) * __Bnm1  |
358 | + (__E + __F2 * __x) * __x * __Bnm2 + __F3 * __x3 * __Bnm3;  |
359 | const _Tp __r = __An / __Bn;  |
360 |   |
361 | const _Tp __prec = std::abs((__F - __r) / __F);  |
362 | __F = __r;  |
363 |   |
364 | if (__prec < __eps || __n > __nmax)  |
365 | break;  |
366 |   |
367 | if (std::abs(__An) > __big || std::abs(__Bn) > __big)  |
368 | {  |
369 | __An /= __big;  |
370 | __Bn /= __big;  |
371 | __Anm1 /= __big;  |
372 | __Bnm1 /= __big;  |
373 | __Anm2 /= __big;  |
374 | __Bnm2 /= __big;  |
375 | __Anm3 /= __big;  |
376 | __Bnm3 /= __big;  |
377 | }  |
378 | else if (std::abs(__An) < _Tp(1) / __big  |
379 | || std::abs(__Bn) < _Tp(1) / __big)  |
380 | {  |
381 | __An *= __big;  |
382 | __Bn *= __big;  |
383 | __Anm1 *= __big;  |
384 | __Bnm1 *= __big;  |
385 | __Anm2 *= __big;  |
386 | __Bnm2 *= __big;  |
387 | __Anm3 *= __big;  |
388 | __Bnm3 *= __big;  |
389 | }  |
390 |   |
391 | ++__n;  |
392 | __Bnm3 = __Bnm2;  |
393 | __Bnm2 = __Bnm1;  |
394 | __Bnm1 = __Bn;  |
395 | __Anm3 = __Anm2;  |
396 | __Anm2 = __Anm1;  |
397 | __Anm1 = __An;  |
398 | }  |
399 |   |
400 | if (__n >= __nmax)  |
401 | std::__throw_runtime_error(__N("Iteration failed to converge "   |
402 | "in __hyperg_luke." ));  |
403 |   |
404 | return __F;  |
405 | }  |
406 |   |
407 |   |
408 | /**  |
409 | * @brief Return the hypogeometric function @f$ _2F_1(a,b;c;x) @f$   |
410 | * by the reflection formulae in Abramowitz & Stegun formula  |
411 | * 15.3.6 for d = c - a - b not integral and formula 15.3.11 for  |
412 | * d = c - a - b integral. This assumes a, b, c != negative  |
413 | * integer.  |
414 | *  |
415 | * The hypogeometric function is defined by  |
416 | * @f[  |
417 | * _2F_1(a,b;c;x) = \frac{\Gamma(c)}{\Gamma(a)\Gamma(b)}  |
418 | * \sum_{n=0}^{\infty}  |
419 | * \frac{\Gamma(a+n)\Gamma(b+n)}{\Gamma(c+n)}  |
420 | * \frac{x^n}{n!}  |
421 | * @f]  |
422 | *  |
423 | * The reflection formula for nonintegral @f$ d = c - a - b @f$ is:  |
424 | * @f[  |
425 | * _2F_1(a,b;c;x) = \frac{\Gamma(c)\Gamma(d)}{\Gamma(c-a)\Gamma(c-b)}  |
426 | * _2F_1(a,b;1-d;1-x)  |
427 | * + \frac{\Gamma(c)\Gamma(-d)}{\Gamma(a)\Gamma(b)}  |
428 | * _2F_1(c-a,c-b;1+d;1-x)  |
429 | * @f]  |
430 | *  |
431 | * The reflection formula for integral @f$ m = c - a - b @f$ is:  |
432 | * @f[  |
433 | * _2F_1(a,b;a+b+m;x) = \frac{\Gamma(m)\Gamma(a+b+m)}{\Gamma(a+m)\Gamma(b+m)}  |
434 | * \sum_{k=0}^{m-1} \frac{(m+a)_k(m+b)_k}{k!(1-m)_k}  |
435 | * -   |
436 | * @f]  |
437 | */  |
438 | template<typename _Tp>  |
439 | _Tp  |
440 | __hyperg_reflect(_Tp __a, _Tp __b, _Tp __c, _Tp __x)  |
441 | {  |
442 | const _Tp __d = __c - __a - __b;  |
443 | const int __intd = std::floor(__d + _Tp(0.5L));  |
444 | const _Tp __eps = std::numeric_limits<_Tp>::epsilon();  |
445 | const _Tp __toler = _Tp(1000) * __eps;  |
446 | const _Tp __log_max = std::log(std::numeric_limits<_Tp>::max());  |
447 | const bool __d_integer = (std::abs(__d - __intd) < __toler);  |
448 |   |
449 | if (__d_integer)  |
450 | {  |
451 | const _Tp __ln_omx = std::log(_Tp(1) - __x);  |
452 | const _Tp __ad = std::abs(__d);  |
453 | _Tp __F1, __F2;  |
454 |   |
455 | _Tp __d1, __d2;  |
456 | if (__d >= _Tp(0))  |
457 | {  |
458 | __d1 = __d;  |
459 | __d2 = _Tp(0);  |
460 | }  |
461 | else  |
462 | {  |
463 | __d1 = _Tp(0);  |
464 | __d2 = __d;  |
465 | }  |
466 |   |
467 | const _Tp __lng_c = __log_gamma(__c);  |
468 |   |
469 | // Evaluate F1.  |
470 | if (__ad < __eps)  |
471 | {  |
472 | // d = c - a - b = 0.  |
473 | __F1 = _Tp(0);  |
474 | }  |
475 | else  |
476 | {  |
477 |   |
478 | bool __ok_d1 = true;  |
479 | _Tp __lng_ad, __lng_ad1, __lng_bd1;  |
480 | __try  |
481 | {  |
482 | __lng_ad = __log_gamma(__ad);  |
483 | __lng_ad1 = __log_gamma(__a + __d1);  |
484 | __lng_bd1 = __log_gamma(__b + __d1);  |
485 | }  |
486 | __catch(...)  |
487 | {  |
488 | __ok_d1 = false;  |
489 | }  |
490 |   |
491 | if (__ok_d1)  |
492 | {  |
493 | /* Gamma functions in the denominator are ok.  |
494 | * Proceed with evaluation.  |
495 | */  |
496 | _Tp __sum1 = _Tp(1);  |
497 | _Tp __term = _Tp(1);  |
498 | _Tp __ln_pre1 = __lng_ad + __lng_c + __d2 * __ln_omx  |
499 | - __lng_ad1 - __lng_bd1;  |
500 |   |
501 | /* Do F1 sum.  |
502 | */  |
503 | for (int __i = 1; __i < __ad; ++__i)  |
504 | {  |
505 | const int __j = __i - 1;  |
506 | __term *= (__a + __d2 + __j) * (__b + __d2 + __j)  |
507 | / (_Tp(1) + __d2 + __j) / __i * (_Tp(1) - __x);  |
508 | __sum1 += __term;  |
509 | }  |
510 |   |
511 | if (__ln_pre1 > __log_max)  |
512 | std::__throw_runtime_error(__N("Overflow of gamma functions"   |
513 | " in __hyperg_luke." ));  |
514 | else  |
515 | __F1 = std::exp(__ln_pre1) * __sum1;  |
516 | }  |
517 | else  |
518 | {  |
519 | // Gamma functions in the denominator were not ok.  |
520 | // So the F1 term is zero.  |
521 | __F1 = _Tp(0);  |
522 | }  |
523 | } // end F1 evaluation  |
524 |   |
525 | // Evaluate F2.  |
526 | bool __ok_d2 = true;  |
527 | _Tp __lng_ad2, __lng_bd2;  |
528 | __try  |
529 | {  |
530 | __lng_ad2 = __log_gamma(__a + __d2);  |
531 | __lng_bd2 = __log_gamma(__b + __d2);  |
532 | }  |
533 | __catch(...)  |
534 | {  |
535 | __ok_d2 = false;  |
536 | }  |
537 |   |
538 | if (__ok_d2)  |
539 | {  |
540 | // Gamma functions in the denominator are ok.  |
541 | // Proceed with evaluation.  |
542 | const int __maxiter = 2000;  |
543 | const _Tp __psi_1 = -__numeric_constants<_Tp>::__gamma_e();  |
544 | const _Tp __psi_1pd = __psi(_Tp(1) + __ad);  |
545 | const _Tp __psi_apd1 = __psi(__a + __d1);  |
546 | const _Tp __psi_bpd1 = __psi(__b + __d1);  |
547 |   |
548 | _Tp __psi_term = __psi_1 + __psi_1pd - __psi_apd1  |
549 | - __psi_bpd1 - __ln_omx;  |
550 | _Tp __fact = _Tp(1);  |
551 | _Tp __sum2 = __psi_term;  |
552 | _Tp __ln_pre2 = __lng_c + __d1 * __ln_omx  |
553 | - __lng_ad2 - __lng_bd2;  |
554 |   |
555 | // Do F2 sum.  |
556 | int __j;  |
557 | for (__j = 1; __j < __maxiter; ++__j)  |
558 | {  |
559 | // Values for psi functions use recurrence;  |
560 | // Abramowitz & Stegun 6.3.5  |
561 | const _Tp __term1 = _Tp(1) / _Tp(__j)  |
562 | + _Tp(1) / (__ad + __j);  |
563 | const _Tp __term2 = _Tp(1) / (__a + __d1 + _Tp(__j - 1))  |
564 | + _Tp(1) / (__b + __d1 + _Tp(__j - 1));  |
565 | __psi_term += __term1 - __term2;  |
566 | __fact *= (__a + __d1 + _Tp(__j - 1))  |
567 | * (__b + __d1 + _Tp(__j - 1))  |
568 | / ((__ad + __j) * __j) * (_Tp(1) - __x);  |
569 | const _Tp __delta = __fact * __psi_term;  |
570 | __sum2 += __delta;  |
571 | if (std::abs(__delta) < __eps * std::abs(__sum2))  |
572 | break;  |
573 | }  |
574 | if (__j == __maxiter)  |
575 | std::__throw_runtime_error(__N("Sum F2 failed to converge "   |
576 | "in __hyperg_reflect" ));  |
577 |   |
578 | if (__sum2 == _Tp(0))  |
579 | __F2 = _Tp(0);  |
580 | else  |
581 | __F2 = std::exp(__ln_pre2) * __sum2;  |
582 | }  |
583 | else  |
584 | {  |
585 | // Gamma functions in the denominator not ok.  |
586 | // So the F2 term is zero.  |
587 | __F2 = _Tp(0);  |
588 | } // end F2 evaluation  |
589 |   |
590 | const _Tp __sgn_2 = (__intd % 2 == 1 ? -_Tp(1) : _Tp(1));  |
591 | const _Tp __F = __F1 + __sgn_2 * __F2;  |
592 |   |
593 | return __F;  |
594 | }  |
595 | else  |
596 | {  |
597 | // d = c - a - b not an integer.  |
598 |   |
599 | // These gamma functions appear in the denominator, so we  |
600 | // catch their harmless domain errors and set the terms to zero.  |
601 | bool __ok1 = true;  |
602 | _Tp __sgn_g1ca = _Tp(0), __ln_g1ca = _Tp(0);  |
603 | _Tp __sgn_g1cb = _Tp(0), __ln_g1cb = _Tp(0);  |
604 | __try  |
605 | {  |
606 | __sgn_g1ca = __log_gamma_sign(__c - __a);  |
607 | __ln_g1ca = __log_gamma(__c - __a);  |
608 | __sgn_g1cb = __log_gamma_sign(__c - __b);  |
609 | __ln_g1cb = __log_gamma(__c - __b);  |
610 | }  |
611 | __catch(...)  |
612 | {  |
613 | __ok1 = false;  |
614 | }  |
615 |   |
616 | bool __ok2 = true;  |
617 | _Tp __sgn_g2a = _Tp(0), __ln_g2a = _Tp(0);  |
618 | _Tp __sgn_g2b = _Tp(0), __ln_g2b = _Tp(0);  |
619 | __try  |
620 | {  |
621 | __sgn_g2a = __log_gamma_sign(__a);  |
622 | __ln_g2a = __log_gamma(__a);  |
623 | __sgn_g2b = __log_gamma_sign(__b);  |
624 | __ln_g2b = __log_gamma(__b);  |
625 | }  |
626 | __catch(...)  |
627 | {  |
628 | __ok2 = false;  |
629 | }  |
630 |   |
631 | const _Tp __sgn_gc = __log_gamma_sign(__c);  |
632 | const _Tp __ln_gc = __log_gamma(__c);  |
633 | const _Tp __sgn_gd = __log_gamma_sign(__d);  |
634 | const _Tp __ln_gd = __log_gamma(__d);  |
635 | const _Tp __sgn_gmd = __log_gamma_sign(-__d);  |
636 | const _Tp __ln_gmd = __log_gamma(-__d);  |
637 |   |
638 | const _Tp __sgn1 = __sgn_gc * __sgn_gd * __sgn_g1ca * __sgn_g1cb;  |
639 | const _Tp __sgn2 = __sgn_gc * __sgn_gmd * __sgn_g2a * __sgn_g2b;  |
640 |   |
641 | _Tp __pre1, __pre2;  |
642 | if (__ok1 && __ok2)  |
643 | {  |
644 | _Tp __ln_pre1 = __ln_gc + __ln_gd - __ln_g1ca - __ln_g1cb;  |
645 | _Tp __ln_pre2 = __ln_gc + __ln_gmd - __ln_g2a - __ln_g2b  |
646 | + __d * std::log(_Tp(1) - __x);  |
647 | if (__ln_pre1 < __log_max && __ln_pre2 < __log_max)  |
648 | {  |
649 | __pre1 = std::exp(__ln_pre1);  |
650 | __pre2 = std::exp(__ln_pre2);  |
651 | __pre1 *= __sgn1;  |
652 | __pre2 *= __sgn2;  |
653 | }  |
654 | else  |
655 | {  |
656 | std::__throw_runtime_error(__N("Overflow of gamma functions "   |
657 | "in __hyperg_reflect" ));  |
658 | }  |
659 | }  |
660 | else if (__ok1 && !__ok2)  |
661 | {  |
662 | _Tp __ln_pre1 = __ln_gc + __ln_gd - __ln_g1ca - __ln_g1cb;  |
663 | if (__ln_pre1 < __log_max)  |
664 | {  |
665 | __pre1 = std::exp(__ln_pre1);  |
666 | __pre1 *= __sgn1;  |
667 | __pre2 = _Tp(0);  |
668 | }  |
669 | else  |
670 | {  |
671 | std::__throw_runtime_error(__N("Overflow of gamma functions "   |
672 | "in __hyperg_reflect" ));  |
673 | }  |
674 | }  |
675 | else if (!__ok1 && __ok2)  |
676 | {  |
677 | _Tp __ln_pre2 = __ln_gc + __ln_gmd - __ln_g2a - __ln_g2b  |
678 | + __d * std::log(_Tp(1) - __x);  |
679 | if (__ln_pre2 < __log_max)  |
680 | {  |
681 | __pre1 = _Tp(0);  |
682 | __pre2 = std::exp(__ln_pre2);  |
683 | __pre2 *= __sgn2;  |
684 | }  |
685 | else  |
686 | {  |
687 | std::__throw_runtime_error(__N("Overflow of gamma functions "   |
688 | "in __hyperg_reflect" ));  |
689 | }  |
690 | }  |
691 | else  |
692 | {  |
693 | __pre1 = _Tp(0);  |
694 | __pre2 = _Tp(0);  |
695 | std::__throw_runtime_error(__N("Underflow of gamma functions "   |
696 | "in __hyperg_reflect" ));  |
697 | }  |
698 |   |
699 | const _Tp __F1 = __hyperg_series(__a, __b, _Tp(1) - __d,  |
700 | _Tp(1) - __x);  |
701 | const _Tp __F2 = __hyperg_series(__c - __a, __c - __b, _Tp(1) + __d,  |
702 | _Tp(1) - __x);  |
703 |   |
704 | const _Tp __F = __pre1 * __F1 + __pre2 * __F2;  |
705 |   |
706 | return __F;  |
707 | }  |
708 | }  |
709 |   |
710 |   |
711 | /**  |
712 | * @brief Return the hypogeometric function @f$ _2F_1(a,b;c;x) @f$.  |
713 | *  |
714 | * The hypogeometric function is defined by  |
715 | * @f[  |
716 | * _2F_1(a,b;c;x) = \frac{\Gamma(c)}{\Gamma(a)\Gamma(b)}  |
717 | * \sum_{n=0}^{\infty}  |
718 | * \frac{\Gamma(a+n)\Gamma(b+n)}{\Gamma(c+n)}  |
719 | * \frac{x^n}{n!}  |
720 | * @f]  |
721 | *  |
722 | * @param __a The first @a numerator parameter.  |
723 | * @param __a The second @a numerator parameter.  |
724 | * @param __c The @a denominator parameter.  |
725 | * @param __x The argument of the confluent hypergeometric function.  |
726 | * @return The confluent hypergeometric function.  |
727 | */  |
728 | template<typename _Tp>  |
729 | _Tp  |
730 | __hyperg(_Tp __a, _Tp __b, _Tp __c, _Tp __x)  |
731 | {  |
732 | #if _GLIBCXX_USE_C99_MATH_TR1  |
733 | const _Tp __a_nint = _GLIBCXX_MATH_NS::nearbyint(__a);  |
734 | const _Tp __b_nint = _GLIBCXX_MATH_NS::nearbyint(__b);  |
735 | const _Tp __c_nint = _GLIBCXX_MATH_NS::nearbyint(__c);  |
736 | #else  |
737 | const _Tp __a_nint = static_cast<int>(__a + _Tp(0.5L));  |
738 | const _Tp __b_nint = static_cast<int>(__b + _Tp(0.5L));  |
739 | const _Tp __c_nint = static_cast<int>(__c + _Tp(0.5L));  |
740 | #endif  |
741 | const _Tp __toler = _Tp(1000) * std::numeric_limits<_Tp>::epsilon();  |
742 | if (std::abs(__x) >= _Tp(1))  |
743 | std::__throw_domain_error(__N("Argument outside unit circle "   |
744 | "in __hyperg." ));  |
745 | else if (__isnan(__a) || __isnan(__b)  |
746 | || __isnan(__c) || __isnan(__x))  |
747 | return std::numeric_limits<_Tp>::quiet_NaN();  |
748 | else if (__c_nint == __c && __c_nint <= _Tp(0))  |
749 | return std::numeric_limits<_Tp>::infinity();  |
750 | else if (std::abs(__c - __b) < __toler || std::abs(__c - __a) < __toler)  |
751 | return std::pow(_Tp(1) - __x, __c - __a - __b);  |
752 | else if (__a >= _Tp(0) && __b >= _Tp(0) && __c >= _Tp(0)  |
753 | && __x >= _Tp(0) && __x < _Tp(0.995L))  |
754 | return __hyperg_series(__a, __b, __c, __x);  |
755 | else if (std::abs(__a) < _Tp(10) && std::abs(__b) < _Tp(10))  |
756 | {  |
757 | // For integer a and b the hypergeometric function is a  |
758 | // finite polynomial.  |
759 | if (__a < _Tp(0) && std::abs(__a - __a_nint) < __toler)  |
760 | return __hyperg_series(__a_nint, __b, __c, __x);  |
761 | else if (__b < _Tp(0) && std::abs(__b - __b_nint) < __toler)  |
762 | return __hyperg_series(__a, __b_nint, __c, __x);  |
763 | else if (__x < -_Tp(0.25L))  |
764 | return __hyperg_luke(__a, __b, __c, __x);  |
765 | else if (__x < _Tp(0.5L))  |
766 | return __hyperg_series(__a, __b, __c, __x);  |
767 | else  |
768 | if (std::abs(__c) > _Tp(10))  |
769 | return __hyperg_series(__a, __b, __c, __x);  |
770 | else  |
771 | return __hyperg_reflect(__a, __b, __c, __x);  |
772 | }  |
773 | else  |
774 | return __hyperg_luke(__a, __b, __c, __x);  |
775 | }  |
776 | } // namespace __detail  |
777 | #undef _GLIBCXX_MATH_NS  |
778 | #if ! _GLIBCXX_USE_STD_SPEC_FUNCS && defined(_GLIBCXX_TR1_CMATH)  |
779 | } // namespace tr1  |
780 | #endif  |
781 |   |
782 | _GLIBCXX_END_NAMESPACE_VERSION  |
783 | }  |
784 |   |
785 | #endif // _GLIBCXX_TR1_HYPERGEOMETRIC_TCC  |
786 | |