RTAI API: base/math/e_lgamma

00001 /* @(#)er_lgamma.c 5.1 93/09/24 */ 00002 /* 00003 * ==================================================== 00004 * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved. 00005 * 00006 * Developed at SunPro, a Sun Microsystems, Inc. business. 00007 * Permission to use, copy, modify, and distribute this 00008 * software is freely granted, provided that this notice 00009 * is preserved. 00010 * ==================================================== 00011 */ 00012 00013 #if defined(LIBM_SCCS) && !defined(lint) 00014 static char rcsid[] = "$NetBSD: e_lgamma_r.c,v 1.7 1995/05/10 20:45:42 jtc Exp $"; 00015 #endif 00016 00017 /* __ieee754_lgamma_r(x, signgamp) 00018 * Reentrant version of the logarithm of the Gamma function 00019 * with user provide pointer for the sign of Gamma(x). 00020 * 00021 * Method: 00022 * 1. Argument Reduction for 0 < x <= 8 00023 * Since gamma(1+s)=s*gamma(s), for x in [0,8], we may 00024 * reduce x to a number in [1.5,2.5] by 00025 * lgamma(1+s) = log(s) + lgamma(s) 00026 * for example, 00027 * lgamma(7.3) = log(6.3) + lgamma(6.3) 00028 * = log(6.3*5.3) + lgamma(5.3) 00029 * = log(6.3*5.3*4.3*3.3*2.3) + lgamma(2.3) 00030 * 2. Polynomial approximation of lgamma around its 00031 * minimun ymin=1.461632144968362245 to maintain monotonicity. 00032 * On [ymin-0.23, ymin+0.27] (i.e., [1.23164,1.73163]), use 00033 * Let z = x-ymin; 00034 * lgamma(x) = -1.214862905358496078218 + z^2*poly(z) 00035 * where 00036 * poly(z) is a 14 degree polynomial. 00037 * 2. Rational approximation in the primary interval [2,3] 00038 * We use the following approximation: 00039 * s = x-2.0; 00040 * lgamma(x) = 0.5*s + s*P(s)/Q(s) 00041 * with accuracy 00042 * |P/Q - (lgamma(x)-0.5s)| < 2**-61.71 00043 * Our algorithms are based on the following observation 00044 * 00045 * zeta(2)-1 2 zeta(3)-1 3 00046 * lgamma(2+s) = s*(1-Euler) + --------- * s - --------- * s + ... 00047 * 2 3 00048 * 00049 * where Euler = 0.5771... is the Euler constant, which is very 00050 * close to 0.5. 00051 * 00052 * 3. For x>=8, we have 00053 * lgamma(x)~(x-0.5)log(x)-x+0.5*log(2pi)+1/(12x)-1/(360x**3)+.... 00054 * (better formula: 00055 * lgamma(x)~(x-0.5)*(log(x)-1)-.5*(log(2pi)-1) + ...) 00056 * Let z = 1/x, then we approximation 00057 * f(z) = lgamma(x) - (x-0.5)(log(x)-1) 00058 * by 00059 * 3 5 11 00060 * w = w0 + w1*z + w2*z + w3*z + ... + w6*z 00061 * where 00062 * |w - f(z)| < 2**-58.74 00063 * 00064 * 4. For negative x, since (G is gamma function) 00065 * -x*G(-x)*G(x) = pi/sin(pi*x), 00066 * we have 00067 * G(x) = pi/(sin(pi*x)*(-x)*G(-x)) 00068 * since G(-x) is positive, sign(G(x)) = sign(sin(pi*x)) for x<0 00069 * Hence, for x<0, signgam = sign(sin(pi*x)) and 00070 * lgamma(x) = log(|Gamma(x)|) 00071 * = log(pi/(|x*sin(pi*x)|)) - lgamma(-x); 00072 * Note: one should avoid compute pi*(-x) directly in the 00073 * computation of sin(pi*(-x)). 00074 * 00075 * 5. Special Cases 00076 * lgamma(2+s) ~ s*(1-Euler) for tiny s 00077 * lgamma(1)=lgamma(2)=0 00078 * lgamma(x) ~ -log(x) for tiny x 00079 * lgamma(0) = lgamma(inf) = inf 00080 * lgamma(-integer) = +-inf 00081 * 00082 */ 00083 00084 #include "math.h" 00085 #include "mathP.h" 00086 00087 #ifdef __STDC__ 00088 static const double 00089 #else 00090 static double 00091 #endif 00092 two52= 4.50359962737049600000e+15, /* 0x43300000, 0x00000000 */ 00093 half= 5.00000000000000000000e-01, /* 0x3FE00000, 0x00000000 */ 00094 one = 1.00000000000000000000e+00, /* 0x3FF00000, 0x00000000 */ 00095 pi = 3.14159265358979311600e+00, /* 0x400921FB, 0x54442D18 */ 00096 a0 = 7.72156649015328655494e-02, /* 0x3FB3C467, 0xE37DB0C8 */ 00097 a1 = 3.22467033424113591611e-01, /* 0x3FD4A34C, 0xC4A60FAD */ 00098 a2 = 6.73523010531292681824e-02, /* 0x3FB13E00, 0x1A5562A7 */ 00099 a3 = 2.05808084325167332806e-02, /* 0x3F951322, 0xAC92547B */ 00100 a4 = 7.38555086081402883957e-03, /* 0x3F7E404F, 0xB68FEFE8 */ 00101 a5 = 2.89051383673415629091e-03, /* 0x3F67ADD8, 0xCCB7926B */ 00102 a6 = 1.19270763183362067845e-03, /* 0x3F538A94, 0x116F3F5D */ 00103 a7 = 5.10069792153511336608e-04, /* 0x3F40B6C6, 0x89B99C00 */ 00104 a8 = 2.20862790713908385557e-04, /* 0x3F2CF2EC, 0xED10E54D */ 00105 a9 = 1.08011567247583939954e-04, /* 0x3F1C5088, 0x987DFB07 */ 00106 a10 = 2.52144565451257326939e-05, /* 0x3EFA7074, 0x428CFA52 */ 00107 a11 = 4.48640949618915160150e-05, /* 0x3F07858E, 0x90A45837 */ 00108 tc = 1.46163214496836224576e+00, /* 0x3FF762D8, 0x6356BE3F */ 00109 tf = -1.21486290535849611461e-01, /* 0xBFBF19B9, 0xBCC38A42 */ 00110 /* tt = -(tail of tf) */ 00111 tt = -3.63867699703950536541e-18, /* 0xBC50C7CA, 0xA48A971F */ 00112 t0 = 4.83836122723810047042e-01, /* 0x3FDEF72B, 0xC8EE38A2 */ 00113 t1 = -1.47587722994593911752e-01, /* 0xBFC2E427, 0x8DC6C509 */ 00114 t2 = 6.46249402391333854778e-02, /* 0x3FB08B42, 0x94D5419B */ 00115 t3 = -3.27885410759859649565e-02, /* 0xBFA0C9A8, 0xDF35B713 */ 00116 t4 = 1.79706750811820387126e-02, /* 0x3F9266E7, 0x970AF9EC */ 00117 t5 = -1.03142241298341437450e-02, /* 0xBF851F9F, 0xBA91EC6A */ 00118 t6 = 6.10053870246291332635e-03, /* 0x3F78FCE0, 0xE370E344 */ 00119 t7 = -3.68452016781138256760e-03, /* 0xBF6E2EFF, 0xB3E914D7 */ 00120 t8 = 2.25964780900612472250e-03, /* 0x3F6282D3, 0x2E15C915 */ 00121 t9 = -1.40346469989232843813e-03, /* 0xBF56FE8E, 0xBF2D1AF1 */ 00122 t10 = 8.81081882437654011382e-04, /* 0x3F4CDF0C, 0xEF61A8E9 */ 00123 t11 = -5.38595305356740546715e-04, /* 0xBF41A610, 0x9C73E0EC */ 00124 t12 = 3.15632070903625950361e-04, /* 0x3F34AF6D, 0x6C0EBBF7 */ 00125 t13 = -3.12754168375120860518e-04, /* 0xBF347F24, 0xECC38C38 */ 00126 t14 = 3.35529192635519073543e-04, /* 0x3F35FD3E, 0xE8C2D3F4 */ 00127 u0 = -7.72156649015328655494e-02, /* 0xBFB3C467, 0xE37DB0C8 */ 00128 u1 = 6.32827064025093366517e-01, /* 0x3FE4401E, 0x8B005DFF */ 00129 u2 = 1.45492250137234768737e+00, /* 0x3FF7475C, 0xD119BD6F */ 00130 u3 = 9.77717527963372745603e-01, /* 0x3FEF4976, 0x44EA8450 */ 00131 u4 = 2.28963728064692451092e-01, /* 0x3FCD4EAE, 0xF6010924 */ 00132 u5 = 1.33810918536787660377e-02, /* 0x3F8B678B, 0xBF2BAB09 */ 00133 v1 = 2.45597793713041134822e+00, /* 0x4003A5D7, 0xC2BD619C */ 00134 v2 = 2.12848976379893395361e+00, /* 0x40010725, 0xA42B18F5 */ 00135 v3 = 7.69285150456672783825e-01, /* 0x3FE89DFB, 0xE45050AF */ 00136 v4 = 1.04222645593369134254e-01, /* 0x3FBAAE55, 0xD6537C88 */ 00137 v5 = 3.21709242282423911810e-03, /* 0x3F6A5ABB, 0x57D0CF61 */ 00138 s0 = -7.72156649015328655494e-02, /* 0xBFB3C467, 0xE37DB0C8 */ 00139 s1 = 2.14982415960608852501e-01, /* 0x3FCB848B, 0x36E20878 */ 00140 s2 = 3.25778796408930981787e-01, /* 0x3FD4D98F, 0x4F139F59 */ 00141 s3 = 1.46350472652464452805e-01, /* 0x3FC2BB9C, 0xBEE5F2F7 */ 00142 s4 = 2.66422703033638609560e-02, /* 0x3F9B481C, 0x7E939961 */ 00143 s5 = 1.84028451407337715652e-03, /* 0x3F5E26B6, 0x7368F239 */ 00144 s6 = 3.19475326584100867617e-05, /* 0x3F00BFEC, 0xDD17E945 */ 00145 r1 = 1.39200533467621045958e+00, /* 0x3FF645A7, 0x62C4AB74 */ 00146 r2 = 7.21935547567138069525e-01, /* 0x3FE71A18, 0x93D3DCDC */ 00147 r3 = 1.71933865632803078993e-01, /* 0x3FC601ED, 0xCCFBDF27 */ 00148 r4 = 1.86459191715652901344e-02, /* 0x3F9317EA, 0x742ED475 */ 00149 r5 = 7.77942496381893596434e-04, /* 0x3F497DDA, 0xCA41A95B */ 00150 r6 = 7.32668430744625636189e-06, /* 0x3EDEBAF7, 0xA5B38140 */ 00151 w0 = 4.18938533204672725052e-01, /* 0x3FDACFE3, 0x90C97D69 */ 00152 w1 = 8.33333333333329678849e-02, /* 0x3FB55555, 0x5555553B */ 00153 w2 = -2.77777777728775536470e-03, /* 0xBF66C16C, 0x16B02E5C */ 00154 w3 = 7.93650558643019558500e-04, /* 0x3F4A019F, 0x98CF38B6 */ 00155 w4 = -5.95187557450339963135e-04, /* 0xBF4380CB, 0x8C0FE741 */ 00156 w5 = 8.36339918996282139126e-04, /* 0x3F4B67BA, 0x4CDAD5D1 */ 00157 w6 = -1.63092934096575273989e-03; /* 0xBF5AB89D, 0x0B9E43E4 */ 00158 00159 #ifdef __STDC__ 00160 static const double zero= 0.00000000000000000000e+00; 00161 #else 00162 static double zero= 0.00000000000000000000e+00; 00163 #endif 00164 00165 static 00166 #ifdef __GNUC__ 00167 __inline__ 00168 #endif 00169 #ifdef __STDC__ 00170 double sin_pi(double x) 00171 #else 00172 double sin_pi(x) 00173 double x; 00174 #endif 00175 { 00176 double y,z; 00177 int n,ix; 00178 00179 GET_HIGH_WORD(ix,x); 00180 ix &= 0x7fffffff; 00181 00182 if(ix<0x3fd00000) return __kernel_sin(pi*x,zero,0); 00183 y = -x; /* x is assume negative */ 00184 00185 /* 00186 * argument reduction, make sure inexact flag not raised if input 00187 * is an integer 00188 */ 00189 z = floor(y); 00190 if(z!=y) { /* inexact anyway */ 00191 y *= 0.5; 00192 y = 2.0*(y - floor(y)); /* y = |x| mod 2.0 */ 00193 n = (int) (y*4.0); 00194 } else { 00195 if(ix>=0x43400000) { 00196 y = zero; n = 0; /* y must be even */ 00197 } else { 00198 if(ix<0x43300000) z = y+two52; /* exact */ 00199 GET_LOW_WORD(n,z); 00200 n &= 1; 00201 y = n; 00202 n<<= 2; 00203 } 00204 } 00205 switch (n) { 00206 case 0: y = __kernel_sin(pi*y,zero,0); break; 00207 case 1: 00208 case 2: y = __kernel_cos(pi*(0.5-y),zero); break; 00209 case 3: 00210 case 4: y = __kernel_sin(pi*(one-y),zero,0); break; 00211 case 5: 00212 case 6: y = -__kernel_cos(pi*(y-1.5),zero); break; 00213 default: y = __kernel_sin(pi*(y-2.0),zero,0); break; 00214 } 00215 return -y; 00216 } 00217 00218 00219 #ifdef __STDC__ 00220 double __ieee754_lgamma_r(double x, int *signgamp) 00221 #else 00222 double __ieee754_lgamma_r(x,signgamp) 00223 double x; int *signgamp; 00224 #endif 00225 { 00226 double t,y,z,nadj=0.0,p,p1,p2,p3,q,r,w; 00227 int i,hx,lx,ix; 00228 00229 EXTRACT_WORDS(hx,lx,x); 00230 00231 /* purge off +-inf, NaN, +-0, and negative arguments */ 00232 *signgamp = 1; 00233 ix = hx&0x7fffffff; 00234 if(ix>=0x7ff00000) return x*x; 00235 if((ix|lx)==0) return one/zero; 00236 if(ix<0x3b900000) { /* |x|<2**-70, return -log(|x|) */ 00237 if(hx<0) { 00238 *signgamp = -1; 00239 return -__ieee754_log(-x); 00240 } else return -__ieee754_log(x); 00241 } 00242 if(hx<0) { 00243 if(ix>=0x43300000) /* |x|>=2**52, must be -integer */ 00244 return one/zero; 00245 t = sin_pi(x); 00246 if(t==zero) return one/zero; /* -integer */ 00247 nadj = __ieee754_log(pi/fabs(t*x)); 00248 if(t<zero) *signgamp = -1; 00249 x = -x; 00250 } 00251 00252 /* purge off 1 and 2 */ 00253 if((((ix-0x3ff00000)|lx)==0)||(((ix-0x40000000)|lx)==0)) r = 0; 00254 /* for x < 2.0 */ 00255 else if(ix<0x40000000) { 00256 if(ix<=0x3feccccc) { /* lgamma(x) = lgamma(x+1)-log(x) */ 00257 r = -__ieee754_log(x); 00258 if(ix>=0x3FE76944) {y = one-x; i= 0;} 00259 else if(ix>=0x3FCDA661) {y= x-(tc-one); i=1;} 00260 else {y = x; i=2;} 00261 } else { 00262 r = zero; 00263 if(ix>=0x3FFBB4C3) {y=2.0-x;i=0;} /* [1.7316,2] */ 00264 else if(ix>=0x3FF3B4C4) {y=x-tc;i=1;} /* [1.23,1.73] */ 00265 else {y=x-one;i=2;} 00266 } 00267 switch(i) { 00268 case 0: 00269 z = y*y; 00270 p1 = a0+z*(a2+z*(a4+z*(a6+z*(a8+z*a10)))); 00271 p2 = z*(a1+z*(a3+z*(a5+z*(a7+z*(a9+z*a11))))); 00272 p = y*p1+p2; 00273 r += (p-0.5*y); break; 00274 case 1: 00275 z = y*y; 00276 w = z*y; 00277 p1 = t0+w*(t3+w*(t6+w*(t9 +w*t12))); /* parallel comp */ 00278 p2 = t1+w*(t4+w*(t7+w*(t10+w*t13))); 00279 p3 = t2+w*(t5+w*(t8+w*(t11+w*t14))); 00280 p = z*p1-(tt-w*(p2+y*p3)); 00281 r += (tf + p); break; 00282 case 2: 00283 p1 = y*(u0+y*(u1+y*(u2+y*(u3+y*(u4+y*u5))))); 00284 p2 = one+y*(v1+y*(v2+y*(v3+y*(v4+y*v5)))); 00285 r += (-0.5*y + p1/p2); 00286 } 00287 } 00288 else if(ix<0x40200000) { /* x < 8.0 */ 00289 i = (int)x; 00290 t = zero; 00291 y = x-(double)i; 00292 p = y*(s0+y*(s1+y*(s2+y*(s3+y*(s4+y*(s5+y*s6)))))); 00293 q = one+y*(r1+y*(r2+y*(r3+y*(r4+y*(r5+y*r6))))); 00294 r = half*y+p/q; 00295 z = one; /* lgamma(1+s) = log(s) + lgamma(s) */ 00296 switch(i) { 00297 case 7: z *= (y+6.0); /* FALLTHRU */ 00298 case 6: z *= (y+5.0); /* FALLTHRU */ 00299 case 5: z *= (y+4.0); /* FALLTHRU */ 00300 case 4: z *= (y+3.0); /* FALLTHRU */ 00301 case 3: z *= (y+2.0); /* FALLTHRU */ 00302 r += __ieee754_log(z); break; 00303 } 00304 /* 8.0 <= x < 2**58 */ 00305 } else if (ix < 0x43900000) { 00306 t = __ieee754_log(x); 00307 z = one/x; 00308 y = z*z; 00309 w = w0+z*(w1+y*(w2+y*(w3+y*(w4+y*(w5+y*w6))))); 00310 r = (x-half)*(t-one)+w; 00311 } else 00312 /* 2**58 <= x <= inf */ 00313 r = x*(__ieee754_log(x)-one); 00314 if(hx<0) r = nadj - r; 00315 return r; 00316 }

base/math/e_lgamma_r.c