1 |
/* -*- Mode: C; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*- |
2 |
* |
3 |
* ***** BEGIN LICENSE BLOCK ***** |
4 |
* Version: MPL 1.1/GPL 2.0/LGPL 2.1 |
5 |
* |
6 |
* The contents of this file are subject to the Mozilla Public License Version |
7 |
* 1.1 (the "License"); you may not use this file except in compliance with |
8 |
* the License. You may obtain a copy of the License at |
9 |
* http://www.mozilla.org/MPL/ |
10 |
* |
11 |
* Software distributed under the License is distributed on an "AS IS" basis, |
12 |
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License |
13 |
* for the specific language governing rights and limitations under the |
14 |
* License. |
15 |
* |
16 |
* The Original Code is Mozilla Communicator client code, released |
17 |
* March 31, 1998. |
18 |
* |
19 |
* The Initial Developer of the Original Code is |
20 |
* Sun Microsystems, Inc. |
21 |
* Portions created by the Initial Developer are Copyright (C) 1998 |
22 |
* the Initial Developer. All Rights Reserved. |
23 |
* |
24 |
* Contributor(s): |
25 |
* |
26 |
* Alternatively, the contents of this file may be used under the terms of |
27 |
* either of the GNU General Public License Version 2 or later (the "GPL"), |
28 |
* or the GNU Lesser General Public License Version 2.1 or later (the "LGPL"), |
29 |
* in which case the provisions of the GPL or the LGPL are applicable instead |
30 |
* of those above. If you wish to allow use of your version of this file only |
31 |
* under the terms of either the GPL or the LGPL, and not to allow others to |
32 |
* use your version of this file under the terms of the MPL, indicate your |
33 |
* decision by deleting the provisions above and replace them with the notice |
34 |
* and other provisions required by the GPL or the LGPL. If you do not delete |
35 |
* the provisions above, a recipient may use your version of this file under |
36 |
* the terms of any one of the MPL, the GPL or the LGPL. |
37 |
* |
38 |
* ***** END LICENSE BLOCK ***** */ |
39 |
|
40 |
/* @(#)e_lgamma_r.c 1.3 95/01/18 */ |
41 |
/* |
42 |
* ==================================================== |
43 |
* Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved. |
44 |
* |
45 |
* Developed at SunSoft, a Sun Microsystems, Inc. business. |
46 |
* Permission to use, copy, modify, and distribute this |
47 |
* software is freely granted, provided that this notice |
48 |
* is preserved. |
49 |
* ==================================================== |
50 |
* |
51 |
*/ |
52 |
|
53 |
/* __ieee754_lgamma_r(x, signgamp) |
54 |
* Reentrant version of the logarithm of the Gamma function |
55 |
* with user provide pointer for the sign of Gamma(x). |
56 |
* |
57 |
* Method: |
58 |
* 1. Argument Reduction for 0 < x <= 8 |
59 |
* Since gamma(1+s)=s*gamma(s), for x in [0,8], we may |
60 |
* reduce x to a number in [1.5,2.5] by |
61 |
* lgamma(1+s) = log(s) + lgamma(s) |
62 |
* for example, |
63 |
* lgamma(7.3) = log(6.3) + lgamma(6.3) |
64 |
* = log(6.3*5.3) + lgamma(5.3) |
65 |
* = log(6.3*5.3*4.3*3.3*2.3) + lgamma(2.3) |
66 |
* 2. Polynomial approximation of lgamma around its |
67 |
* minimun ymin=1.461632144968362245 to maintain monotonicity. |
68 |
* On [ymin-0.23, ymin+0.27] (i.e., [1.23164,1.73163]), use |
69 |
* Let z = x-ymin; |
70 |
* lgamma(x) = -1.214862905358496078218 + z^2*poly(z) |
71 |
* where |
72 |
* poly(z) is a 14 degree polynomial. |
73 |
* 2. Rational approximation in the primary interval [2,3] |
74 |
* We use the following approximation: |
75 |
* s = x-2.0; |
76 |
* lgamma(x) = 0.5*s + s*P(s)/Q(s) |
77 |
* with accuracy |
78 |
* |P/Q - (lgamma(x)-0.5s)| < 2**-61.71 |
79 |
* Our algorithms are based on the following observation |
80 |
* |
81 |
* zeta(2)-1 2 zeta(3)-1 3 |
82 |
* lgamma(2+s) = s*(1-Euler) + --------- * s - --------- * s + ... |
83 |
* 2 3 |
84 |
* |
85 |
* where Euler = 0.5771... is the Euler constant, which is very |
86 |
* close to 0.5. |
87 |
* |
88 |
* 3. For x>=8, we have |
89 |
* lgamma(x)~(x-0.5)log(x)-x+0.5*log(2pi)+1/(12x)-1/(360x**3)+.... |
90 |
* (better formula: |
91 |
* lgamma(x)~(x-0.5)*(log(x)-1)-.5*(log(2pi)-1) + ...) |
92 |
* Let z = 1/x, then we approximation |
93 |
* f(z) = lgamma(x) - (x-0.5)(log(x)-1) |
94 |
* by |
95 |
* 3 5 11 |
96 |
* w = w0 + w1*z + w2*z + w3*z + ... + w6*z |
97 |
* where |
98 |
* |w - f(z)| < 2**-58.74 |
99 |
* |
100 |
* 4. For negative x, since (G is gamma function) |
101 |
* -x*G(-x)*G(x) = pi/sin(pi*x), |
102 |
* we have |
103 |
* G(x) = pi/(sin(pi*x)*(-x)*G(-x)) |
104 |
* since G(-x) is positive, sign(G(x)) = sign(sin(pi*x)) for x<0 |
105 |
* Hence, for x<0, signgam = sign(sin(pi*x)) and |
106 |
* lgamma(x) = log(|Gamma(x)|) |
107 |
* = log(pi/(|x*sin(pi*x)|)) - lgamma(-x); |
108 |
* Note: one should avoid compute pi*(-x) directly in the |
109 |
* computation of sin(pi*(-x)). |
110 |
* |
111 |
* 5. Special Cases |
112 |
* lgamma(2+s) ~ s*(1-Euler) for tiny s |
113 |
* lgamma(1)=lgamma(2)=0 |
114 |
* lgamma(x) ~ -log(x) for tiny x |
115 |
* lgamma(0) = lgamma(inf) = inf |
116 |
* lgamma(-integer) = +-inf |
117 |
* |
118 |
*/ |
119 |
|
120 |
#include "fdlibm.h" |
121 |
|
122 |
#ifdef __STDC__ |
123 |
static const double |
124 |
#else |
125 |
static double |
126 |
#endif |
127 |
two52= 4.50359962737049600000e+15, /* 0x43300000, 0x00000000 */ |
128 |
half= 5.00000000000000000000e-01, /* 0x3FE00000, 0x00000000 */ |
129 |
one = 1.00000000000000000000e+00, /* 0x3FF00000, 0x00000000 */ |
130 |
pi = 3.14159265358979311600e+00, /* 0x400921FB, 0x54442D18 */ |
131 |
a0 = 7.72156649015328655494e-02, /* 0x3FB3C467, 0xE37DB0C8 */ |
132 |
a1 = 3.22467033424113591611e-01, /* 0x3FD4A34C, 0xC4A60FAD */ |
133 |
a2 = 6.73523010531292681824e-02, /* 0x3FB13E00, 0x1A5562A7 */ |
134 |
a3 = 2.05808084325167332806e-02, /* 0x3F951322, 0xAC92547B */ |
135 |
a4 = 7.38555086081402883957e-03, /* 0x3F7E404F, 0xB68FEFE8 */ |
136 |
a5 = 2.89051383673415629091e-03, /* 0x3F67ADD8, 0xCCB7926B */ |
137 |
a6 = 1.19270763183362067845e-03, /* 0x3F538A94, 0x116F3F5D */ |
138 |
a7 = 5.10069792153511336608e-04, /* 0x3F40B6C6, 0x89B99C00 */ |
139 |
a8 = 2.20862790713908385557e-04, /* 0x3F2CF2EC, 0xED10E54D */ |
140 |
a9 = 1.08011567247583939954e-04, /* 0x3F1C5088, 0x987DFB07 */ |
141 |
a10 = 2.52144565451257326939e-05, /* 0x3EFA7074, 0x428CFA52 */ |
142 |
a11 = 4.48640949618915160150e-05, /* 0x3F07858E, 0x90A45837 */ |
143 |
tc = 1.46163214496836224576e+00, /* 0x3FF762D8, 0x6356BE3F */ |
144 |
tf = -1.21486290535849611461e-01, /* 0xBFBF19B9, 0xBCC38A42 */ |
145 |
/* tt = -(tail of tf) */ |
146 |
tt = -3.63867699703950536541e-18, /* 0xBC50C7CA, 0xA48A971F */ |
147 |
t0 = 4.83836122723810047042e-01, /* 0x3FDEF72B, 0xC8EE38A2 */ |
148 |
t1 = -1.47587722994593911752e-01, /* 0xBFC2E427, 0x8DC6C509 */ |
149 |
t2 = 6.46249402391333854778e-02, /* 0x3FB08B42, 0x94D5419B */ |
150 |
t3 = -3.27885410759859649565e-02, /* 0xBFA0C9A8, 0xDF35B713 */ |
151 |
t4 = 1.79706750811820387126e-02, /* 0x3F9266E7, 0x970AF9EC */ |
152 |
t5 = -1.03142241298341437450e-02, /* 0xBF851F9F, 0xBA91EC6A */ |
153 |
t6 = 6.10053870246291332635e-03, /* 0x3F78FCE0, 0xE370E344 */ |
154 |
t7 = -3.68452016781138256760e-03, /* 0xBF6E2EFF, 0xB3E914D7 */ |
155 |
t8 = 2.25964780900612472250e-03, /* 0x3F6282D3, 0x2E15C915 */ |
156 |
t9 = -1.40346469989232843813e-03, /* 0xBF56FE8E, 0xBF2D1AF1 */ |
157 |
t10 = 8.81081882437654011382e-04, /* 0x3F4CDF0C, 0xEF61A8E9 */ |
158 |
t11 = -5.38595305356740546715e-04, /* 0xBF41A610, 0x9C73E0EC */ |
159 |
t12 = 3.15632070903625950361e-04, /* 0x3F34AF6D, 0x6C0EBBF7 */ |
160 |
t13 = -3.12754168375120860518e-04, /* 0xBF347F24, 0xECC38C38 */ |
161 |
t14 = 3.35529192635519073543e-04, /* 0x3F35FD3E, 0xE8C2D3F4 */ |
162 |
u0 = -7.72156649015328655494e-02, /* 0xBFB3C467, 0xE37DB0C8 */ |
163 |
u1 = 6.32827064025093366517e-01, /* 0x3FE4401E, 0x8B005DFF */ |
164 |
u2 = 1.45492250137234768737e+00, /* 0x3FF7475C, 0xD119BD6F */ |
165 |
u3 = 9.77717527963372745603e-01, /* 0x3FEF4976, 0x44EA8450 */ |
166 |
u4 = 2.28963728064692451092e-01, /* 0x3FCD4EAE, 0xF6010924 */ |
167 |
u5 = 1.33810918536787660377e-02, /* 0x3F8B678B, 0xBF2BAB09 */ |
168 |
v1 = 2.45597793713041134822e+00, /* 0x4003A5D7, 0xC2BD619C */ |
169 |
v2 = 2.12848976379893395361e+00, /* 0x40010725, 0xA42B18F5 */ |
170 |
v3 = 7.69285150456672783825e-01, /* 0x3FE89DFB, 0xE45050AF */ |
171 |
v4 = 1.04222645593369134254e-01, /* 0x3FBAAE55, 0xD6537C88 */ |
172 |
v5 = 3.21709242282423911810e-03, /* 0x3F6A5ABB, 0x57D0CF61 */ |
173 |
s0 = -7.72156649015328655494e-02, /* 0xBFB3C467, 0xE37DB0C8 */ |
174 |
s1 = 2.14982415960608852501e-01, /* 0x3FCB848B, 0x36E20878 */ |
175 |
s2 = 3.25778796408930981787e-01, /* 0x3FD4D98F, 0x4F139F59 */ |
176 |
s3 = 1.46350472652464452805e-01, /* 0x3FC2BB9C, 0xBEE5F2F7 */ |
177 |
s4 = 2.66422703033638609560e-02, /* 0x3F9B481C, 0x7E939961 */ |
178 |
s5 = 1.84028451407337715652e-03, /* 0x3F5E26B6, 0x7368F239 */ |
179 |
s6 = 3.19475326584100867617e-05, /* 0x3F00BFEC, 0xDD17E945 */ |
180 |
r1 = 1.39200533467621045958e+00, /* 0x3FF645A7, 0x62C4AB74 */ |
181 |
r2 = 7.21935547567138069525e-01, /* 0x3FE71A18, 0x93D3DCDC */ |
182 |
r3 = 1.71933865632803078993e-01, /* 0x3FC601ED, 0xCCFBDF27 */ |
183 |
r4 = 1.86459191715652901344e-02, /* 0x3F9317EA, 0x742ED475 */ |
184 |
r5 = 7.77942496381893596434e-04, /* 0x3F497DDA, 0xCA41A95B */ |
185 |
r6 = 7.32668430744625636189e-06, /* 0x3EDEBAF7, 0xA5B38140 */ |
186 |
w0 = 4.18938533204672725052e-01, /* 0x3FDACFE3, 0x90C97D69 */ |
187 |
w1 = 8.33333333333329678849e-02, /* 0x3FB55555, 0x5555553B */ |
188 |
w2 = -2.77777777728775536470e-03, /* 0xBF66C16C, 0x16B02E5C */ |
189 |
w3 = 7.93650558643019558500e-04, /* 0x3F4A019F, 0x98CF38B6 */ |
190 |
w4 = -5.95187557450339963135e-04, /* 0xBF4380CB, 0x8C0FE741 */ |
191 |
w5 = 8.36339918996282139126e-04, /* 0x3F4B67BA, 0x4CDAD5D1 */ |
192 |
w6 = -1.63092934096575273989e-03; /* 0xBF5AB89D, 0x0B9E43E4 */ |
193 |
|
194 |
static double zero= 0.00000000000000000000e+00; |
195 |
|
196 |
#ifdef __STDC__ |
197 |
static double sin_pi(double x) |
198 |
#else |
199 |
static double sin_pi(x) |
200 |
double x; |
201 |
#endif |
202 |
{ |
203 |
fd_twoints u; |
204 |
double y,z; |
205 |
int n,ix; |
206 |
|
207 |
u.d = x; |
208 |
ix = 0x7fffffff&__HI(u); |
209 |
|
210 |
if(ix<0x3fd00000) return __kernel_sin(pi*x,zero,0); |
211 |
y = -x; /* x is assume negative */ |
212 |
|
213 |
/* |
214 |
* argument reduction, make sure inexact flag not raised if input |
215 |
* is an integer |
216 |
*/ |
217 |
z = fd_floor(y); |
218 |
if(z!=y) { /* inexact anyway */ |
219 |
y *= 0.5; |
220 |
y = 2.0*(y - fd_floor(y)); /* y = |x| mod 2.0 */ |
221 |
n = (int) (y*4.0); |
222 |
} else { |
223 |
if(ix>=0x43400000) { |
224 |
y = zero; n = 0; /* y must be even */ |
225 |
} else { |
226 |
if(ix<0x43300000) z = y+two52; /* exact */ |
227 |
u.d = z; |
228 |
n = __LO(u)&1; /* lower word of z */ |
229 |
y = n; |
230 |
n<<= 2; |
231 |
} |
232 |
} |
233 |
switch (n) { |
234 |
case 0: y = __kernel_sin(pi*y,zero,0); break; |
235 |
case 1: |
236 |
case 2: y = __kernel_cos(pi*(0.5-y),zero); break; |
237 |
case 3: |
238 |
case 4: y = __kernel_sin(pi*(one-y),zero,0); break; |
239 |
case 5: |
240 |
case 6: y = -__kernel_cos(pi*(y-1.5),zero); break; |
241 |
default: y = __kernel_sin(pi*(y-2.0),zero,0); break; |
242 |
} |
243 |
return -y; |
244 |
} |
245 |
|
246 |
|
247 |
#ifdef __STDC__ |
248 |
double __ieee754_lgamma_r(double x, int *signgamp) |
249 |
#else |
250 |
double __ieee754_lgamma_r(x,signgamp) |
251 |
double x; int *signgamp; |
252 |
#endif |
253 |
{ |
254 |
fd_twoints u; |
255 |
double t,y,z,nadj,p,p1,p2,p3,q,r,w; |
256 |
int i,hx,lx,ix; |
257 |
|
258 |
u.d = x; |
259 |
hx = __HI(u); |
260 |
lx = __LO(u); |
261 |
|
262 |
/* purge off +-inf, NaN, +-0, and negative arguments */ |
263 |
*signgamp = 1; |
264 |
ix = hx&0x7fffffff; |
265 |
if(ix>=0x7ff00000) return x*x; |
266 |
if((ix|lx)==0) return one/zero; |
267 |
if(ix<0x3b900000) { /* |x|<2**-70, return -log(|x|) */ |
268 |
if(hx<0) { |
269 |
*signgamp = -1; |
270 |
return -__ieee754_log(-x); |
271 |
} else return -__ieee754_log(x); |
272 |
} |
273 |
if(hx<0) { |
274 |
if(ix>=0x43300000) /* |x|>=2**52, must be -integer */ |
275 |
return one/zero; |
276 |
t = sin_pi(x); |
277 |
if(t==zero) return one/zero; /* -integer */ |
278 |
nadj = __ieee754_log(pi/fd_fabs(t*x)); |
279 |
if(t<zero) *signgamp = -1; |
280 |
x = -x; |
281 |
} |
282 |
|
283 |
/* purge off 1 and 2 */ |
284 |
if((((ix-0x3ff00000)|lx)==0)||(((ix-0x40000000)|lx)==0)) r = 0; |
285 |
/* for x < 2.0 */ |
286 |
else if(ix<0x40000000) { |
287 |
if(ix<=0x3feccccc) { /* lgamma(x) = lgamma(x+1)-log(x) */ |
288 |
r = -__ieee754_log(x); |
289 |
if(ix>=0x3FE76944) {y = one-x; i= 0;} |
290 |
else if(ix>=0x3FCDA661) {y= x-(tc-one); i=1;} |
291 |
else {y = x; i=2;} |
292 |
} else { |
293 |
r = zero; |
294 |
if(ix>=0x3FFBB4C3) {y=2.0-x;i=0;} /* [1.7316,2] */ |
295 |
else if(ix>=0x3FF3B4C4) {y=x-tc;i=1;} /* [1.23,1.73] */ |
296 |
else {y=x-one;i=2;} |
297 |
} |
298 |
switch(i) { |
299 |
case 0: |
300 |
z = y*y; |
301 |
p1 = a0+z*(a2+z*(a4+z*(a6+z*(a8+z*a10)))); |
302 |
p2 = z*(a1+z*(a3+z*(a5+z*(a7+z*(a9+z*a11))))); |
303 |
p = y*p1+p2; |
304 |
r += (p-0.5*y); break; |
305 |
case 1: |
306 |
z = y*y; |
307 |
w = z*y; |
308 |
p1 = t0+w*(t3+w*(t6+w*(t9 +w*t12))); /* parallel comp */ |
309 |
p2 = t1+w*(t4+w*(t7+w*(t10+w*t13))); |
310 |
p3 = t2+w*(t5+w*(t8+w*(t11+w*t14))); |
311 |
p = z*p1-(tt-w*(p2+y*p3)); |
312 |
r += (tf + p); break; |
313 |
case 2: |
314 |
p1 = y*(u0+y*(u1+y*(u2+y*(u3+y*(u4+y*u5))))); |
315 |
p2 = one+y*(v1+y*(v2+y*(v3+y*(v4+y*v5)))); |
316 |
r += (-0.5*y + p1/p2); |
317 |
} |
318 |
} |
319 |
else if(ix<0x40200000) { /* x < 8.0 */ |
320 |
i = (int)x; |
321 |
t = zero; |
322 |
y = x-(double)i; |
323 |
p = y*(s0+y*(s1+y*(s2+y*(s3+y*(s4+y*(s5+y*s6)))))); |
324 |
q = one+y*(r1+y*(r2+y*(r3+y*(r4+y*(r5+y*r6))))); |
325 |
r = half*y+p/q; |
326 |
z = one; /* lgamma(1+s) = log(s) + lgamma(s) */ |
327 |
switch(i) { |
328 |
case 7: z *= (y+6.0); /* FALLTHRU */ |
329 |
case 6: z *= (y+5.0); /* FALLTHRU */ |
330 |
case 5: z *= (y+4.0); /* FALLTHRU */ |
331 |
case 4: z *= (y+3.0); /* FALLTHRU */ |
332 |
case 3: z *= (y+2.0); /* FALLTHRU */ |
333 |
r += __ieee754_log(z); break; |
334 |
} |
335 |
/* 8.0 <= x < 2**58 */ |
336 |
} else if (ix < 0x43900000) { |
337 |
t = __ieee754_log(x); |
338 |
z = one/x; |
339 |
y = z*z; |
340 |
w = w0+z*(w1+y*(w2+y*(w3+y*(w4+y*(w5+y*w6))))); |
341 |
r = (x-half)*(t-one)+w; |
342 |
} else |
343 |
/* 2**58 <= x <= inf */ |
344 |
r = x*(__ieee754_log(x)-one); |
345 |
if(hx<0) r = nadj - r; |
346 |
return r; |
347 |
} |