1 | /* erf_lgamma.c -- float version of er_lgamma.c. |
---|
2 | * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. |
---|
3 | */ |
---|
4 | |
---|
5 | /* |
---|
6 | * ==================================================== |
---|
7 | * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved. |
---|
8 | * |
---|
9 | * Developed at SunPro, a Sun Microsystems, Inc. business. |
---|
10 | * Permission to use, copy, modify, and distribute this |
---|
11 | * software is freely granted, provided that this notice |
---|
12 | * is preserved. |
---|
13 | * ==================================================== |
---|
14 | * |
---|
15 | */ |
---|
16 | |
---|
17 | #include "fdlibm.h" |
---|
18 | |
---|
19 | #ifdef __STDC__ |
---|
20 | static const float |
---|
21 | #else |
---|
22 | static float |
---|
23 | #endif |
---|
24 | two23= 8.3886080000e+06, /* 0x4b000000 */ |
---|
25 | half= 5.0000000000e-01, /* 0x3f000000 */ |
---|
26 | one = 1.0000000000e+00, /* 0x3f800000 */ |
---|
27 | pi = 3.1415927410e+00, /* 0x40490fdb */ |
---|
28 | a0 = 7.7215664089e-02, /* 0x3d9e233f */ |
---|
29 | a1 = 3.2246702909e-01, /* 0x3ea51a66 */ |
---|
30 | a2 = 6.7352302372e-02, /* 0x3d89f001 */ |
---|
31 | a3 = 2.0580807701e-02, /* 0x3ca89915 */ |
---|
32 | a4 = 7.3855509982e-03, /* 0x3bf2027e */ |
---|
33 | a5 = 2.8905137442e-03, /* 0x3b3d6ec6 */ |
---|
34 | a6 = 1.1927076848e-03, /* 0x3a9c54a1 */ |
---|
35 | a7 = 5.1006977446e-04, /* 0x3a05b634 */ |
---|
36 | a8 = 2.2086278477e-04, /* 0x39679767 */ |
---|
37 | a9 = 1.0801156895e-04, /* 0x38e28445 */ |
---|
38 | a10 = 2.5214456400e-05, /* 0x37d383a2 */ |
---|
39 | a11 = 4.4864096708e-05, /* 0x383c2c75 */ |
---|
40 | tc = 1.4616321325e+00, /* 0x3fbb16c3 */ |
---|
41 | tf = -1.2148628384e-01, /* 0xbdf8cdcd */ |
---|
42 | /* tt = -(tail of tf) */ |
---|
43 | tt = 6.6971006518e-09, /* 0x31e61c52 */ |
---|
44 | t0 = 4.8383611441e-01, /* 0x3ef7b95e */ |
---|
45 | t1 = -1.4758771658e-01, /* 0xbe17213c */ |
---|
46 | t2 = 6.4624942839e-02, /* 0x3d845a15 */ |
---|
47 | t3 = -3.2788541168e-02, /* 0xbd064d47 */ |
---|
48 | t4 = 1.7970675603e-02, /* 0x3c93373d */ |
---|
49 | t5 = -1.0314224288e-02, /* 0xbc28fcfe */ |
---|
50 | t6 = 6.1005386524e-03, /* 0x3bc7e707 */ |
---|
51 | t7 = -3.6845202558e-03, /* 0xbb7177fe */ |
---|
52 | t8 = 2.2596477065e-03, /* 0x3b141699 */ |
---|
53 | t9 = -1.4034647029e-03, /* 0xbab7f476 */ |
---|
54 | t10 = 8.8108185446e-04, /* 0x3a66f867 */ |
---|
55 | t11 = -5.3859531181e-04, /* 0xba0d3085 */ |
---|
56 | t12 = 3.1563205994e-04, /* 0x39a57b6b */ |
---|
57 | t13 = -3.1275415677e-04, /* 0xb9a3f927 */ |
---|
58 | t14 = 3.3552918467e-04, /* 0x39afe9f7 */ |
---|
59 | u0 = -7.7215664089e-02, /* 0xbd9e233f */ |
---|
60 | u1 = 6.3282704353e-01, /* 0x3f2200f4 */ |
---|
61 | u2 = 1.4549225569e+00, /* 0x3fba3ae7 */ |
---|
62 | u3 = 9.7771751881e-01, /* 0x3f7a4bb2 */ |
---|
63 | u4 = 2.2896373272e-01, /* 0x3e6a7578 */ |
---|
64 | u5 = 1.3381091878e-02, /* 0x3c5b3c5e */ |
---|
65 | v1 = 2.4559779167e+00, /* 0x401d2ebe */ |
---|
66 | v2 = 2.1284897327e+00, /* 0x4008392d */ |
---|
67 | v3 = 7.6928514242e-01, /* 0x3f44efdf */ |
---|
68 | v4 = 1.0422264785e-01, /* 0x3dd572af */ |
---|
69 | v5 = 3.2170924824e-03, /* 0x3b52d5db */ |
---|
70 | s0 = -7.7215664089e-02, /* 0xbd9e233f */ |
---|
71 | s1 = 2.1498242021e-01, /* 0x3e5c245a */ |
---|
72 | s2 = 3.2577878237e-01, /* 0x3ea6cc7a */ |
---|
73 | s3 = 1.4635047317e-01, /* 0x3e15dce6 */ |
---|
74 | s4 = 2.6642270386e-02, /* 0x3cda40e4 */ |
---|
75 | s5 = 1.8402845599e-03, /* 0x3af135b4 */ |
---|
76 | s6 = 3.1947532989e-05, /* 0x3805ff67 */ |
---|
77 | r1 = 1.3920053244e+00, /* 0x3fb22d3b */ |
---|
78 | r2 = 7.2193557024e-01, /* 0x3f38d0c5 */ |
---|
79 | r3 = 1.7193385959e-01, /* 0x3e300f6e */ |
---|
80 | r4 = 1.8645919859e-02, /* 0x3c98bf54 */ |
---|
81 | r5 = 7.7794247773e-04, /* 0x3a4beed6 */ |
---|
82 | r6 = 7.3266842264e-06, /* 0x36f5d7bd */ |
---|
83 | w0 = 4.1893854737e-01, /* 0x3ed67f1d */ |
---|
84 | w1 = 8.3333335817e-02, /* 0x3daaaaab */ |
---|
85 | w2 = -2.7777778450e-03, /* 0xbb360b61 */ |
---|
86 | w3 = 7.9365057172e-04, /* 0x3a500cfd */ |
---|
87 | w4 = -5.9518753551e-04, /* 0xba1c065c */ |
---|
88 | w5 = 8.3633989561e-04, /* 0x3a5b3dd2 */ |
---|
89 | w6 = -1.6309292987e-03; /* 0xbad5c4e8 */ |
---|
90 | |
---|
91 | #ifdef __STDC__ |
---|
92 | static const float zero= 0.0000000000e+00; |
---|
93 | #else |
---|
94 | static float zero= 0.0000000000e+00; |
---|
95 | #endif |
---|
96 | |
---|
97 | #ifdef __STDC__ |
---|
98 | static float sin_pif(float x) |
---|
99 | #else |
---|
100 | static float sin_pif(x) |
---|
101 | float x; |
---|
102 | #endif |
---|
103 | { |
---|
104 | float y,z; |
---|
105 | __int32_t n,ix; |
---|
106 | |
---|
107 | GET_FLOAT_WORD(ix,x); |
---|
108 | ix &= 0x7fffffff; |
---|
109 | |
---|
110 | if(ix<0x3e800000) return __kernel_sinf(pi*x,zero,0); |
---|
111 | y = -x; /* x is assume negative */ |
---|
112 | |
---|
113 | /* |
---|
114 | * argument reduction, make sure inexact flag not raised if input |
---|
115 | * is an integer |
---|
116 | */ |
---|
117 | z = floorf(y); |
---|
118 | if(z!=y) { /* inexact anyway */ |
---|
119 | y *= (float)0.5; |
---|
120 | y = (float)2.0*(y - floorf(y)); /* y = |x| mod 2.0 */ |
---|
121 | n = (__int32_t) (y*(float)4.0); |
---|
122 | } else { |
---|
123 | if(ix>=0x4b800000) { |
---|
124 | y = zero; n = 0; /* y must be even */ |
---|
125 | } else { |
---|
126 | if(ix<0x4b000000) z = y+two23; /* exact */ |
---|
127 | GET_FLOAT_WORD(n,z); |
---|
128 | n &= 1; |
---|
129 | y = n; |
---|
130 | n<<= 2; |
---|
131 | } |
---|
132 | } |
---|
133 | switch (n) { |
---|
134 | case 0: y = __kernel_sinf(pi*y,zero,0); break; |
---|
135 | case 1: |
---|
136 | case 2: y = __kernel_cosf(pi*((float)0.5-y),zero); break; |
---|
137 | case 3: |
---|
138 | case 4: y = __kernel_sinf(pi*(one-y),zero,0); break; |
---|
139 | case 5: |
---|
140 | case 6: y = -__kernel_cosf(pi*(y-(float)1.5),zero); break; |
---|
141 | default: y = __kernel_sinf(pi*(y-(float)2.0),zero,0); break; |
---|
142 | } |
---|
143 | return -y; |
---|
144 | } |
---|
145 | |
---|
146 | |
---|
147 | #ifdef __STDC__ |
---|
148 | float __ieee754_lgammaf_r(float x, int *signgamp) |
---|
149 | #else |
---|
150 | float __ieee754_lgammaf_r(x,signgamp) |
---|
151 | float x; int *signgamp; |
---|
152 | #endif |
---|
153 | { |
---|
154 | float t,y,z,nadj = 0.0,p,p1,p2,p3,q,r,w; |
---|
155 | __int32_t i,hx,ix; |
---|
156 | |
---|
157 | GET_FLOAT_WORD(hx,x); |
---|
158 | |
---|
159 | /* purge off +-inf, NaN, +-0, and negative arguments */ |
---|
160 | *signgamp = 1; |
---|
161 | ix = hx&0x7fffffff; |
---|
162 | if(ix>=0x7f800000) return x*x; |
---|
163 | if(ix==0) return one/zero; |
---|
164 | if(ix<0x1c800000) { /* |x|<2**-70, return -log(|x|) */ |
---|
165 | if(hx<0) { |
---|
166 | *signgamp = -1; |
---|
167 | return -__ieee754_logf(-x); |
---|
168 | } else return -__ieee754_logf(x); |
---|
169 | } |
---|
170 | if(hx<0) { |
---|
171 | if(ix>=0x4b000000) /* |x|>=2**23, must be -integer */ |
---|
172 | return one/zero; |
---|
173 | t = sin_pif(x); |
---|
174 | if(t==zero) return one/zero; /* -integer */ |
---|
175 | nadj = __ieee754_logf(pi/fabsf(t*x)); |
---|
176 | if(t<zero) *signgamp = -1; |
---|
177 | x = -x; |
---|
178 | } |
---|
179 | |
---|
180 | /* purge off 1 and 2 */ |
---|
181 | if (ix==0x3f800000||ix==0x40000000) r = 0; |
---|
182 | /* for x < 2.0 */ |
---|
183 | else if(ix<0x40000000) { |
---|
184 | if(ix<=0x3f666666) { /* lgamma(x) = lgamma(x+1)-log(x) */ |
---|
185 | r = -__ieee754_logf(x); |
---|
186 | if(ix>=0x3f3b4a20) {y = one-x; i= 0;} |
---|
187 | else if(ix>=0x3e6d3308) {y= x-(tc-one); i=1;} |
---|
188 | else {y = x; i=2;} |
---|
189 | } else { |
---|
190 | r = zero; |
---|
191 | if(ix>=0x3fdda618) {y=(float)2.0-x;i=0;} /* [1.7316,2] */ |
---|
192 | else if(ix>=0x3F9da620) {y=x-tc;i=1;} /* [1.23,1.73] */ |
---|
193 | else {y=x-one;i=2;} |
---|
194 | } |
---|
195 | switch(i) { |
---|
196 | case 0: |
---|
197 | z = y*y; |
---|
198 | p1 = a0+z*(a2+z*(a4+z*(a6+z*(a8+z*a10)))); |
---|
199 | p2 = z*(a1+z*(a3+z*(a5+z*(a7+z*(a9+z*a11))))); |
---|
200 | p = y*p1+p2; |
---|
201 | r += (p-(float)0.5*y); break; |
---|
202 | case 1: |
---|
203 | z = y*y; |
---|
204 | w = z*y; |
---|
205 | p1 = t0+w*(t3+w*(t6+w*(t9 +w*t12))); /* parallel comp */ |
---|
206 | p2 = t1+w*(t4+w*(t7+w*(t10+w*t13))); |
---|
207 | p3 = t2+w*(t5+w*(t8+w*(t11+w*t14))); |
---|
208 | p = z*p1-(tt-w*(p2+y*p3)); |
---|
209 | r += (tf + p); break; |
---|
210 | case 2: |
---|
211 | p1 = y*(u0+y*(u1+y*(u2+y*(u3+y*(u4+y*u5))))); |
---|
212 | p2 = one+y*(v1+y*(v2+y*(v3+y*(v4+y*v5)))); |
---|
213 | r += (-(float)0.5*y + p1/p2); |
---|
214 | } |
---|
215 | } |
---|
216 | else if(ix<0x41000000) { /* x < 8.0 */ |
---|
217 | i = (__int32_t)x; |
---|
218 | t = zero; |
---|
219 | y = x-(float)i; |
---|
220 | p = y*(s0+y*(s1+y*(s2+y*(s3+y*(s4+y*(s5+y*s6)))))); |
---|
221 | q = one+y*(r1+y*(r2+y*(r3+y*(r4+y*(r5+y*r6))))); |
---|
222 | r = half*y+p/q; |
---|
223 | z = one; /* lgamma(1+s) = log(s) + lgamma(s) */ |
---|
224 | switch(i) { |
---|
225 | case 7: z *= (y+(float)6.0); /* FALLTHRU */ |
---|
226 | case 6: z *= (y+(float)5.0); /* FALLTHRU */ |
---|
227 | case 5: z *= (y+(float)4.0); /* FALLTHRU */ |
---|
228 | case 4: z *= (y+(float)3.0); /* FALLTHRU */ |
---|
229 | case 3: z *= (y+(float)2.0); /* FALLTHRU */ |
---|
230 | r += __ieee754_logf(z); break; |
---|
231 | } |
---|
232 | /* 8.0 <= x < 2**58 */ |
---|
233 | } else if (ix < 0x5c800000) { |
---|
234 | t = __ieee754_logf(x); |
---|
235 | z = one/x; |
---|
236 | y = z*z; |
---|
237 | w = w0+z*(w1+y*(w2+y*(w3+y*(w4+y*(w5+y*w6))))); |
---|
238 | r = (x-half)*(t-one)+w; |
---|
239 | } else |
---|
240 | /* 2**58 <= x <= inf */ |
---|
241 | r = x*(__ieee754_logf(x)-one); |
---|
242 | if(hx<0) r = nadj - r; |
---|
243 | return r; |
---|
244 | } |
---|