xref: /freebsd/lib/msun/src/e_lgammaf_r.c (revision f39bffc62c1395bde25d152c7f68fdf7cbaab414)
1 /* e_lgammaf_r.c -- float version of e_lgamma_r.c.
2  * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com.
3  * Conversion to float fixed By Steven G. Kargl.
4  */
5 
6 /*
7  * ====================================================
8  * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
9  *
10  * Developed at SunPro, a Sun Microsystems, Inc. business.
11  * Permission to use, copy, modify, and distribute this
12  * software is freely granted, provided that this notice
13  * is preserved.
14  * ====================================================
15  */
16 
17 #include <sys/cdefs.h>
18 __FBSDID("$FreeBSD$");
19 
20 #include "math.h"
21 #include "math_private.h"
22 
23 static const volatile float vzero = 0;
24 
25 static const float
26 zero=  0,
27 half=  0.5,
28 one =  1,
29 pi  =  3.1415927410e+00, /* 0x40490fdb */
30 /*
31  * Domain y in [0x1p-27, 0.27], range ~[-3.4599e-10, 3.4590e-10]:
32  * |(lgamma(2 - y) + 0.5 * y) / y - a(y)| < 2**-31.4
33  */
34 a0  =  7.72156641e-02, /* 0x3d9e233f */
35 a1  =  3.22467119e-01, /* 0x3ea51a69 */
36 a2  =  6.73484802e-02, /* 0x3d89ee00 */
37 a3  =  2.06395667e-02, /* 0x3ca9144f */
38 a4  =  6.98275631e-03, /* 0x3be4cf9b */
39 a5  =  4.11768444e-03, /* 0x3b86eda4 */
40 /*
41  * Domain x in [tc-0.24, tc+0.28], range ~[-5.6577e-10, 5.5677e-10]:
42  * |(lgamma(x) - tf) - t(x - tc)| < 2**-30.8.
43  */
44 tc  =  1.46163213e+00, /* 0x3fbb16c3 */
45 tf  = -1.21486291e-01, /* 0xbdf8cdce */
46 t0  = -2.94064460e-11, /* 0xae0154b7 */
47 t1  = -2.35939837e-08, /* 0xb2caabb8 */
48 t2  =  4.83836412e-01, /* 0x3ef7b968 */
49 t3  = -1.47586212e-01, /* 0xbe1720d7 */
50 t4  =  6.46013096e-02, /* 0x3d844db1 */
51 t5  = -3.28450352e-02, /* 0xbd068884 */
52 t6  =  1.86483748e-02, /* 0x3c98c47a */
53 t7  = -9.89206228e-03, /* 0xbc221251 */
54 /*
55  * Domain y in [-0.1, 0.232], range ~[-8.4931e-10, 8.7794e-10]:
56  * |(lgamma(1 + y) + 0.5 * y) / y - u(y) / v(y)| < 2**-31.2
57  */
58 u0  = -7.72156641e-02, /* 0xbd9e233f */
59 u1  =  7.36789703e-01, /* 0x3f3c9e40 */
60 u2  =  4.95649040e-01, /* 0x3efdc5b6 */
61 v1  =  1.10958421e+00, /* 0x3f8e06db */
62 v2  =  2.10598111e-01, /* 0x3e57a708 */
63 v3  = -1.02995494e-02, /* 0xbc28bf71 */
64 /*
65  * Domain x in (2, 3], range ~[-5.5189e-11, 5.2317e-11]:
66  * |(lgamma(y+2) - 0.5 * y) / y - s(y)/r(y)| < 2**-35.0
67  * with y = x - 2.
68  */
69 s0 = -7.72156641e-02, /* 0xbd9e233f */
70 s1 =  2.69987404e-01, /* 0x3e8a3bca */
71 s2 =  1.42851010e-01, /* 0x3e124789 */
72 s3 =  1.19389519e-02, /* 0x3c439b98 */
73 r1 =  6.79650068e-01, /* 0x3f2dfd8c */
74 r2 =  1.16058730e-01, /* 0x3dedb033 */
75 r3 =  3.75673687e-03, /* 0x3b763396 */
76 /*
77  * Domain z in [8, 0x1p24], range ~[-1.2640e-09, 1.2640e-09]:
78  * |lgamma(x) - (x - 0.5) * (log(x) - 1) - w(1/x)| < 2**-29.6.
79  */
80 w0 =  4.18938547e-01, /* 0x3ed67f1d */
81 w1 =  8.33332464e-02, /* 0x3daaaa9f */
82 w2 = -2.76129087e-03; /* 0xbb34f6c6 */
83 
84 static float
85 sin_pif(float x)
86 {
87 	volatile float vz;
88 	float y,z;
89 	int n;
90 
91 	y = -x;
92 
93 	vz = y+0x1p23F;			/* depend on 0 <= y < 0x1p23 */
94 	z = vz-0x1p23F;			/* rintf(y) for the above range */
95 	if (z == y)
96 	    return zero;
97 
98 	vz = y+0x1p21F;
99 	GET_FLOAT_WORD(n,vz);		/* bits for rounded y (units 0.25) */
100 	z = vz-0x1p21F;			/* y rounded to a multiple of 0.25 */
101 	if (z > y) {
102 	    z -= 0.25F;			/* adjust to round down */
103 	    n--;
104 	}
105 	n &= 7;				/* octant of y mod 2 */
106 	y = y - z + n * 0.25F;		/* y mod 2 */
107 
108 	switch (n) {
109 	    case 0:   y =  __kernel_sindf(pi*y); break;
110 	    case 1:
111 	    case 2:   y =  __kernel_cosdf(pi*((float)0.5-y)); break;
112 	    case 3:
113 	    case 4:   y =  __kernel_sindf(pi*(one-y)); break;
114 	    case 5:
115 	    case 6:   y = -__kernel_cosdf(pi*(y-(float)1.5)); break;
116 	    default:  y =  __kernel_sindf(pi*(y-(float)2.0)); break;
117 	    }
118 	return -y;
119 }
120 
121 
122 float
123 __ieee754_lgammaf_r(float x, int *signgamp)
124 {
125 	float nadj,p,p1,p2,q,r,t,w,y,z;
126 	int32_t hx;
127 	int i,ix;
128 
129 	GET_FLOAT_WORD(hx,x);
130 
131     /* purge +-Inf and NaNs */
132 	*signgamp = 1;
133 	ix = hx&0x7fffffff;
134 	if(ix>=0x7f800000) return x*x;
135 
136     /* purge +-0 and tiny arguments */
137 	*signgamp = 1-2*((uint32_t)hx>>31);
138 	if(ix<0x32000000) {		/* |x|<2**-27, return -log(|x|) */
139 	    if(ix==0)
140 	        return one/vzero;
141 	    return -__ieee754_logf(fabsf(x));
142 	}
143 
144     /* purge negative integers and start evaluation for other x < 0 */
145 	if(hx<0) {
146 	    *signgamp = 1;
147 	    if(ix>=0x4b000000) 		/* |x|>=2**23, must be -integer */
148 		return one/vzero;
149 	    t = sin_pif(x);
150 	    if(t==zero) return one/vzero; /* -integer */
151 	    nadj = __ieee754_logf(pi/fabsf(t*x));
152 	    if(t<zero) *signgamp = -1;
153 	    x = -x;
154 	}
155 
156     /* purge 1 and 2 */
157 	if (ix==0x3f800000||ix==0x40000000) r = 0;
158     /* for x < 2.0 */
159 	else if(ix<0x40000000) {
160 	    if(ix<=0x3f666666) { 	/* lgamma(x) = lgamma(x+1)-log(x) */
161 		r = -__ieee754_logf(x);
162 		if(ix>=0x3f3b4a20) {y = one-x; i= 0;}
163 		else if(ix>=0x3e6d3308) {y= x-(tc-one); i=1;}
164 	  	else {y = x; i=2;}
165 	    } else {
166 	  	r = zero;
167 	        if(ix>=0x3fdda618) {y=2-x;i=0;} /* [1.7316,2] */
168 	        else if(ix>=0x3F9da620) {y=x-tc;i=1;} /* [1.23,1.73] */
169 		else {y=x-one;i=2;}
170 	    }
171 	    switch(i) {
172 	      case 0:
173 		z = y*y;
174 		p1 = a0+z*(a2+z*a4);
175 		p2 = z*(a1+z*(a3+z*a5));
176 		p  = y*p1+p2;
177 		r  += p-y/2; break;
178 	      case 1:
179 		p = t0+y*t1+y*y*(t2+y*(t3+y*(t4+y*(t5+y*(t6+y*t7)))));
180 		r += tf + p; break;
181 	      case 2:
182 		p1 = y*(u0+y*(u1+y*u2));
183 		p2 = one+y*(v1+y*(v2+y*v3));
184 		r += p1/p2-y/2;
185 	    }
186 	}
187     /* x < 8.0 */
188 	else if(ix<0x41000000) {
189 	    i = x;
190 	    y = x-i;
191 	    p = y*(s0+y*(s1+y*(s2+y*s3)));
192 	    q = one+y*(r1+y*(r2+y*r3));
193 	    r = y/2+p/q;
194 	    z = one;	/* lgamma(1+s) = log(s) + lgamma(s) */
195 	    switch(i) {
196 	    case 7: z *= (y+6);		/* FALLTHRU */
197 	    case 6: z *= (y+5);		/* FALLTHRU */
198 	    case 5: z *= (y+4);		/* FALLTHRU */
199 	    case 4: z *= (y+3);		/* FALLTHRU */
200 	    case 3: z *= (y+2);		/* FALLTHRU */
201 		    r += __ieee754_logf(z); break;
202 	    }
203     /* 8.0 <= x < 2**27 */
204 	} else if (ix < 0x4d000000) {
205 	    t = __ieee754_logf(x);
206 	    z = one/x;
207 	    y = z*z;
208 	    w = w0+z*(w1+y*w2);
209 	    r = (x-half)*(t-one)+w;
210 	} else
211     /* 2**27 <= x <= inf */
212 	    r =  x*(__ieee754_logf(x)-one);
213 	if(hx<0) r = nadj - r;
214 	return r;
215 }
216