1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
|
/* Single-precision floating point 2^x.
Copyright (C) 1997-2016 Free Software Foundation, Inc.
This file is part of the GNU C Library.
Contributed by Geoffrey Keating <geoffk@ozemail.com.au>
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, see
<http://www.gnu.org/licenses/>. */
/* The basic design here is from
Shmuel Gal and Boris Bachelis, "An Accurate Elementary Mathematical
Library for the IEEE Floating Point Standard", ACM Trans. Math. Soft.,
17 (1), March 1991, pp. 26-45.
It has been slightly modified to compute 2^x instead of e^x, and for
single-precision.
*/
#ifndef _GNU_SOURCE
# define _GNU_SOURCE
#endif
#include <stdlib.h>
#include <float.h>
#include <ieee754.h>
#include <math.h>
#include <fenv.h>
#include <inttypes.h>
#include <math_private.h>
#include "t_exp2f.h"
static const float TWOM100 = 7.88860905e-31;
static const float TWO127 = 1.7014118346e+38;
float
__ieee754_exp2f (float x)
{
static const float himark = (float) FLT_MAX_EXP;
static const float lomark = (float) (FLT_MIN_EXP - FLT_MANT_DIG - 1);
/* Check for usual case. */
if (isless (x, himark) && isgreaterequal (x, lomark))
{
static const float THREEp14 = 49152.0;
int tval, unsafe;
float rx, x22, result;
union ieee754_float ex2_u, scale_u;
if (fabsf (x) < FLT_EPSILON / 4.0f)
return 1.0f + x;
{
SET_RESTORE_ROUND_NOEXF (FE_TONEAREST);
/* 1. Argument reduction.
Choose integers ex, -128 <= t < 128, and some real
-1/512 <= x1 <= 1/512 so that
x = ex + t/512 + x1.
First, calculate rx = ex + t/256. */
rx = x + THREEp14;
rx -= THREEp14;
x -= rx; /* Compute x=x1. */
/* Compute tval = (ex*256 + t)+128.
Now, t = (tval mod 256)-128 and ex=tval/256 [that's mod, NOT %;
and /-round-to-nearest not the usual c integer /]. */
tval = (int) (rx * 256.0f + 128.0f);
/* 2. Adjust for accurate table entry.
Find e so that
x = ex + t/256 + e + x2
where -7e-4 < e < 7e-4, and
(float)(2^(t/256+e))
is accurate to one part in 2^-64. */
/* 'tval & 255' is the same as 'tval%256' except that it's always
positive.
Compute x = x2. */
x -= __exp2f_deltatable[tval & 255];
/* 3. Compute ex2 = 2^(t/255+e+ex). */
ex2_u.f = __exp2f_atable[tval & 255];
tval >>= 8;
/* x2 is an integer multiple of 2^-30; avoid intermediate
underflow from the calculation of x22 * x. */
unsafe = abs(tval) >= -FLT_MIN_EXP - 32;
ex2_u.ieee.exponent += tval >> unsafe;
scale_u.f = 1.0;
scale_u.ieee.exponent += tval - (tval >> unsafe);
/* 4. Approximate 2^x2 - 1, using a second-degree polynomial,
with maximum error in [-2^-9 - 2^-14, 2^-9 + 2^-14]
less than 1.3e-10. */
x22 = (.24022656679f * x + .69314736128f) * ex2_u.f;
}
/* 5. Return (2^x2-1) * 2^(t/512+e+ex) + 2^(t/512+e+ex). */
result = x22 * x + ex2_u.f;
if (!unsafe)
return result;
else
{
result *= scale_u.f;
math_check_force_underflow_nonneg (result);
return result;
}
}
/* Exceptional cases: */
else if (isless (x, himark))
{
if (isinf (x))
/* e^-inf == 0, with no error. */
return 0;
else
/* Underflow */
return TWOM100 * TWOM100;
}
else
/* Return x, if x is a NaN or Inf; or overflow, otherwise. */
return TWO127*x;
}
strong_alias (__ieee754_exp2f, __exp2f_finite)
|
