xref: /illumos-gate/usr/src/lib/libm/common/C/atan.c (revision 8533946bd264dca901fdf56bf3da1d81e728b423)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
24  */
25 /*
26  * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
27  * Use is subject to license terms.
28  */
29 
30 #pragma weak __atan = atan
31 
32 /* INDENT OFF */
33 /*
34  * atan(x)
35  * Accurate Table look-up algorithm with polynomial approximation in
36  * partially product form.
37  *
38  * -- K.C. Ng, October 17, 2004
39  *
40  * Algorithm
41  *
42  * (1). Purge off Inf and NaN and 0
43  * (2). Reduce x to positive by atan(x) = -atan(-x).
44  * (3). For x <= 1/8 and let z = x*x, return
45  *	(2.1) if x < 2^(-prec/2), atan(x) = x  with inexact flag raised
46  *	(2.2) if x < 2^(-prec/4-1), atan(x) = x+(x/3)(x*x)
47  *	(2.3) if x < 2^(-prec/6-2), atan(x) = x+(z-5/3)(z*x/5)
48  *	(2.4) Otherwise
49  *		atan(x) = poly1(x) = x + A * B,
50  *	where
51  *		A = (p1*x*z) * (p2+z(p3+z))
52  *		B = (p4+z)+z*z) * (p5+z(p6+z))
53  *	Note: (i) domain of poly1 is [0, 1/8], (ii) remez relative
54  *	approximation error of poly1 is bounded by
55  * 		|(atan(x)-poly1(x))/x| <= 2^-57.61
56  * (4). For x >= 8 then
57  *	(3.1) if x >= 2^prec,     atan(x) = atan(inf) - pio2lo
58  *	(3.2) if x >= 2^(prec/3), atan(x) = atan(inf) - 1/x
59  *	(3.3) if x <= 65,	  atan(x) = atan(inf) - poly1(1/x)
60  *	(3.4) otherwise           atan(x) = atan(inf) - poly2(1/x)
61  *	where
62  *		poly2(r) = (q1*r) * (q2+z(q3+z)) * (q4+z),
63  *	its domain is [0, 0.0154]; and its remez absolute
64  *	approximation error is bounded by
65  *		|atan(x)-poly2(x)|<= 2^-59.45
66  *
67  * (5). Now x is in (0.125, 8).
68  *	Recall identity
69  *		atan(x) = atan(y) + atan((x-y)/(1+x*y)).
70  *	Let j = (ix - 0x3fc00000) >> 16, 0 <= j < 96, where ix is the high
71  *	part of x in IEEE double format. Then
72  *		atan(x) = atan(y[j]) + poly2((x-y[j])/(1+x*y[j]))
73  *	where y[j] are carefully chosen so that it matches x to around 4.5
74  *	bits and at the same time atan(y[j]) is very close to an IEEE double
75  *	floating point number. Calculation indicates that
76  *		max|(x-y[j])/(1+x*y[j])| < 0.0154
77  *		j,x
78  *
79  * Accuracy: Maximum error observed is bounded by 0.6 ulp after testing
80  * more than 10 million random arguments
81  */
82 /* INDENT ON */
83 
84 #include "libm.h"
85 #include "libm_protos.h"
86 
87 extern const double _TBL_atan[];
88 static const double g[] = {
89 /* one	= */  1.0,
90 /* p1	= */  8.02176624254765935351230154992663301527500152588e-0002,
91 /* p2	= */  1.27223421700559402580665846471674740314483642578e+0000,
92 /* p3	= */ -1.20606901800503640842521235754247754812240600586e+0000,
93 /* p4	= */ -2.36088967922325565496066701598465442657470703125e+0000,
94 /* p5	= */  1.38345799501389166152875986881554126739501953125e+0000,
95 /* p6	= */  1.06742368078953453469637224770849570631980895996e+0000,
96 /* q1   = */ -1.42796626333911796935538518482644576579332351685e-0001,
97 /* q2   = */  3.51427110447873227059810477159863497078605962912e+0000,
98 /* q3   = */  5.92129112708164262457444237952586263418197631836e-0001,
99 /* q4   = */ -1.99272234785683144409063061175402253866195678711e+0000,
100 /* pio2hi */  1.570796326794896558e+00,
101 /* pio2lo */  6.123233995736765886e-17,
102 /* t1   = */ -0.333333333333333333333333333333333,
103 /* t2   = */  0.2,
104 /* t3   = */ -1.666666666666666666666666666666666,
105 };
106 
107 #define	one g[0]
108 #define	p1 g[1]
109 #define	p2 g[2]
110 #define	p3 g[3]
111 #define	p4 g[4]
112 #define	p5 g[5]
113 #define	p6 g[6]
114 #define	q1 g[7]
115 #define	q2 g[8]
116 #define	q3 g[9]
117 #define	q4 g[10]
118 #define	pio2hi g[11]
119 #define	pio2lo g[12]
120 #define	t1 g[13]
121 #define	t2 g[14]
122 #define	t3 g[15]
123 
124 
125 double
126 atan(double x) {
127 	double y, z, r, p, s;
128 	int ix, lx, hx, j;
129 
130 	hx = ((int *) &x)[HIWORD];
131 	lx = ((int *) &x)[LOWORD];
132 	ix = hx & ~0x80000000;
133 	j = ix >> 20;
134 
135 	/* for |x| < 1/8 */
136 	if (j < 0x3fc) {
137 		if (j < 0x3f5) {	/* when |x| < 2**(-prec/6-2) */
138 			if (j < 0x3e3) {	/* if |x| < 2**(-prec/2-2) */
139 				return ((int) x == 0 ? x : one);
140 			}
141 			if (j < 0x3f1) {	/* if |x| < 2**(-prec/4-1) */
142 				return (x + (x * t1) * (x * x));
143 			} else {	/* if |x| < 2**(-prec/6-2) */
144 				z = x * x;
145 				s = t2 * x;
146 				return (x + (t3 + z) * (s * z));
147 			}
148 		}
149 		z = x * x; s = p1 * x;
150 		return (x + ((s * z) * (p2 + z * (p3 + z))) *
151 				(((p4 + z) + z * z) * (p5 + z * (p6 + z))));
152 	}
153 
154 	/* for |x| >= 8.0 */
155 	if (j >= 0x402) {
156 		if (j < 0x436) {
157 			r = one / x;
158 			if (hx >= 0) {
159 				y =  pio2hi; p =  pio2lo;
160 			} else {
161 				y = -pio2hi; p = -pio2lo;
162 			}
163 			if (ix < 0x40504000) {	/* x <  65 */
164 				z = r * r;
165 				s = p1 * r;
166 				return (y + ((p - r) - ((s * z) *
167 					(p2 + z * (p3 + z))) *
168 					(((p4 + z) + z * z) *
169 					(p5 + z * (p6 + z)))));
170 			} else if (j < 0x412) {
171 				z = r * r;
172 				return (y + (p - ((q1 * r) * (q4 + z)) *
173 					(q2 + z * (q3 + z))));
174 			} else
175 				return (y + (p - r));
176 		} else {
177 			if (j >= 0x7ff) /* x is inf or NaN */
178 				if (((ix - 0x7ff00000) | lx) != 0)
179 #if defined(FPADD_TRAPS_INCOMPLETE_ON_NAN)
180 					return (ix >= 0x7ff80000 ? x : x - x);
181 					/* assumes sparc-like QNaN */
182 #else
183 					return (x - x);
184 #endif
185 			y = -pio2lo;
186 			return (hx >= 0 ? pio2hi - y : y - pio2hi);
187 		}
188 	} else {	/* now x is between 1/8 and 8 */
189 		double *w, w0, w1, s, z;
190 		w = (double *) _TBL_atan + (((ix - 0x3fc00000) >> 16) << 1);
191 		w0 = (hx >= 0)? w[0] : -w[0];
192 		s = (x - w0) / (one + x * w0);
193 		w1 = (hx >= 0)? w[1] : -w[1];
194 		z = s * s;
195 		return (((q1 * s) * (q4 + z)) * (q2 + z * (q3 + z)) + w1);
196 	}
197 }
198