]>
Commit | Line | Data |
---|---|---|
63640cb7 | 1 | /* Double-precision floating point 2^x. |
0ac5ae23 UD |
2 | Copyright (C) 1997,1998,2000,2001,2005,2006,2011 |
3 | Free Software Foundation, Inc. | |
63640cb7 UD |
4 | This file is part of the GNU C Library. |
5 | Contributed by Geoffrey Keating <geoffk@ozemail.com.au> | |
6 | ||
7 | The GNU C Library is free software; you can redistribute it and/or | |
41bdb6e2 AJ |
8 | modify it under the terms of the GNU Lesser General Public |
9 | License as published by the Free Software Foundation; either | |
10 | version 2.1 of the License, or (at your option) any later version. | |
63640cb7 UD |
11 | |
12 | The GNU C Library is distributed in the hope that it will be useful, | |
13 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
41bdb6e2 | 15 | Lesser General Public License for more details. |
63640cb7 | 16 | |
41bdb6e2 AJ |
17 | You should have received a copy of the GNU Lesser General Public |
18 | License along with the GNU C Library; if not, write to the Free | |
19 | Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA | |
20 | 02111-1307 USA. */ | |
63640cb7 UD |
21 | |
22 | /* The basic design here is from | |
23 | Shmuel Gal and Boris Bachelis, "An Accurate Elementary Mathematical | |
24 | Library for the IEEE Floating Point Standard", ACM Trans. Math. Soft., | |
25 | 17 (1), March 1991, pp. 26-45. | |
26 | It has been slightly modified to compute 2^x instead of e^x. | |
27 | */ | |
63640cb7 UD |
28 | #include <stdlib.h> |
29 | #include <float.h> | |
30 | #include <ieee754.h> | |
31 | #include <math.h> | |
32 | #include <fenv.h> | |
33 | #include <inttypes.h> | |
34 | #include <math_private.h> | |
35 | ||
36 | #include "t_exp2.h" | |
37 | ||
d38f1dba UD |
38 | static const double TWO1023 = 8.988465674311579539e+307; |
39 | static const double TWOM1000 = 9.3326361850321887899e-302; | |
63640cb7 UD |
40 | |
41 | double | |
42 | __ieee754_exp2 (double x) | |
43 | { | |
44 | static const double himark = (double) DBL_MAX_EXP; | |
601d2942 | 45 | static const double lomark = (double) (DBL_MIN_EXP - DBL_MANT_DIG - 1); |
63640cb7 UD |
46 | |
47 | /* Check for usual case. */ | |
99ce7b04 | 48 | if (__builtin_expect (isless (x, himark), 1)) |
63640cb7 | 49 | { |
99ce7b04 UD |
50 | /* Exceptional cases: */ |
51 | if (__builtin_expect (! isgreaterequal (x, lomark), 0)) | |
52 | { | |
53 | if (__isinf (x)) | |
54 | /* e^-inf == 0, with no error. */ | |
55 | return 0; | |
56 | else | |
57 | /* Underflow */ | |
58 | return TWOM1000 * TWOM1000; | |
59 | } | |
60 | ||
63640cb7 UD |
61 | static const double THREEp42 = 13194139533312.0; |
62 | int tval, unsafe; | |
63 | double rx, x22, result; | |
64 | union ieee754_double ex2_u, scale_u; | |
65 | fenv_t oldenv; | |
66 | ||
d38f1dba | 67 | libc_feholdexcept (&oldenv); |
63640cb7 UD |
68 | #ifdef FE_TONEAREST |
69 | /* If we don't have this, it's too bad. */ | |
d38f1dba | 70 | libc_fesetround (FE_TONEAREST); |
63640cb7 UD |
71 | #endif |
72 | ||
73 | /* 1. Argument reduction. | |
74 | Choose integers ex, -256 <= t < 256, and some real | |
75 | -1/1024 <= x1 <= 1024 so that | |
76 | x = ex + t/512 + x1. | |
77 | ||
78 | First, calculate rx = ex + t/512. */ | |
79 | rx = x + THREEp42; | |
80 | rx -= THREEp42; | |
81 | x -= rx; /* Compute x=x1. */ | |
82 | /* Compute tval = (ex*512 + t)+256. | |
83 | Now, t = (tval mod 512)-256 and ex=tval/512 [that's mod, NOT %; and | |
84 | /-round-to-nearest not the usual c integer /]. */ | |
85 | tval = (int) (rx * 512.0 + 256.0); | |
86 | ||
87 | /* 2. Adjust for accurate table entry. | |
88 | Find e so that | |
89 | x = ex + t/512 + e + x2 | |
90 | where -1e6 < e < 1e6, and | |
91 | (double)(2^(t/512+e)) | |
92 | is accurate to one part in 2^-64. */ | |
93 | ||
94 | /* 'tval & 511' is the same as 'tval%512' except that it's always | |
95 | positive. | |
96 | Compute x = x2. */ | |
97 | x -= exp2_deltatable[tval & 511]; | |
98 | ||
99 | /* 3. Compute ex2 = 2^(t/512+e+ex). */ | |
100 | ex2_u.d = exp2_accuratetable[tval & 511]; | |
101 | tval >>= 9; | |
102 | unsafe = abs(tval) >= -DBL_MIN_EXP - 1; | |
103 | ex2_u.ieee.exponent += tval >> unsafe; | |
104 | scale_u.d = 1.0; | |
105 | scale_u.ieee.exponent += tval - (tval >> unsafe); | |
106 | ||
107 | /* 4. Approximate 2^x2 - 1, using a fourth-degree polynomial, | |
108 | with maximum error in [-2^-10-2^-30,2^-10+2^-30] | |
109 | less than 10^-19. */ | |
110 | ||
111 | x22 = (((.0096181293647031180 | |
112 | * x + .055504110254308625) | |
113 | * x + .240226506959100583) | |
114 | * x + .69314718055994495) * ex2_u.d; | |
d38f1dba | 115 | math_opt_barrier (x22); |
63640cb7 UD |
116 | |
117 | /* 5. Return (2^x2-1) * 2^(t/512+e+ex) + 2^(t/512+e+ex). */ | |
d38f1dba | 118 | libc_fesetenv (&oldenv); |
63640cb7 UD |
119 | |
120 | result = x22 * x + ex2_u.d; | |
121 | ||
122 | if (!unsafe) | |
123 | return result; | |
124 | else | |
125 | return result * scale_u.d; | |
126 | } | |
63640cb7 UD |
127 | else |
128 | /* Return x, if x is a NaN or Inf; or overflow, otherwise. */ | |
129 | return TWO1023*x; | |
130 | } | |
0ac5ae23 | 131 | strong_alias (__ieee754_exp2, __exp2_finite) |