1 /* Quad-precision floating point e^x.
2 Copyright (C) 1999-2022 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4
5 The GNU C Library is free software; you can redistribute it and/or
6 modify it under the terms of the GNU Lesser General Public
7 License as published by the Free Software Foundation; either
8 version 2.1 of the License, or (at your option) any later version.
9
10 The GNU C Library is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 Lesser General Public License for more details.
14
15 You should have received a copy of the GNU Lesser General Public
16 License along with the GNU C Library; if not, see
17 <https://www.gnu.org/licenses/>. */
18
19 /* The basic design here is from
20 Abraham Ziv, "Fast Evaluation of Elementary Mathematical Functions with
21 Correctly Rounded Last Bit", ACM Trans. Math. Soft., 17 (3), September 1991,
22 pp. 410-423.
23
24 We work with number pairs where the first number is the high part and
25 the second one is the low part. Arithmetic with the high part numbers must
26 be exact, without any roundoff errors.
27
28 The input value, X, is written as
29 X = n * ln(2)_0 + arg1[t1]_0 + arg2[t2]_0 + x
30 - n * ln(2)_1 + arg1[t1]_1 + arg2[t2]_1 + xl
31
32 where:
33 - n is an integer, 16384 >= n >= -16495;
34 - ln(2)_0 is the first 93 bits of ln(2), and |ln(2)_0-ln(2)-ln(2)_1| < 2^-205
35 - t1 is an integer, 89 >= t1 >= -89
36 - t2 is an integer, 65 >= t2 >= -65
37 - |arg1[t1]-t1/256.0| < 2^-53
38 - |arg2[t2]-t2/32768.0| < 2^-53
39 - x + xl is whatever is left, |x + xl| < 2^-16 + 2^-53
40
41 Then e^x is approximated as
42
43 e^x = 2^n_1 ( 2^n_0 e^(arg1[t1]_0 + arg1[t1]_1) e^(arg2[t2]_0 + arg2[t2]_1)
44 + 2^n_0 e^(arg1[t1]_0 + arg1[t1]_1) e^(arg2[t2]_0 + arg2[t2]_1)
45 * p (x + xl + n * ln(2)_1))
46 where:
47 - p(x) is a polynomial approximating e(x)-1
48 - e^(arg1[t1]_0 + arg1[t1]_1) is obtained from a table
49 - e^(arg2[t2]_0 + arg2[t2]_1) likewise
50 - n_1 + n_0 = n, so that |n_0| < -LDBL_MIN_EXP-1.
51
52 If it happens that n_1 == 0 (this is the usual case), that multiplication
53 is omitted.
54 */
55
56 #ifndef _GNU_SOURCE
57 #define _GNU_SOURCE
58 #endif
59 #include <float.h>
60 #include <ieee754.h>
61 #include <math.h>
62 #include <fenv.h>
63 #include <inttypes.h>
64 #include <math-barriers.h>
65 #include <math_private.h>
66 #include <math-underflow.h>
67 #include <stdlib.h>
68 #include "t_expl.h"
69 #include <libm-alias-finite.h>
70
71 static const _Float128 C[] = {
72 /* Smallest integer x for which e^x overflows. */
73 #define himark C[0]
74 L(11356.523406294143949491931077970765),
75
76 /* Largest integer x for which e^x underflows. */
77 #define lomark C[1]
78 L(-11433.4627433362978788372438434526231),
79
80 /* 3x2^96 */
81 #define THREEp96 C[2]
82 L(59421121885698253195157962752.0),
83
84 /* 3x2^103 */
85 #define THREEp103 C[3]
86 L(30423614405477505635920876929024.0),
87
88 /* 3x2^111 */
89 #define THREEp111 C[4]
90 L(7788445287802241442795744493830144.0),
91
92 /* 1/ln(2) */
93 #define M_1_LN2 C[5]
94 L(1.44269504088896340735992468100189204),
95
96 /* first 93 bits of ln(2) */
97 #define M_LN2_0 C[6]
98 L(0.693147180559945309417232121457981864),
99
100 /* ln2_0 - ln(2) */
101 #define M_LN2_1 C[7]
102 L(-1.94704509238074995158795957333327386E-31),
103
104 /* very small number */
105 #define TINY C[8]
106 L(1.0e-4900),
107
108 /* 2^16383 */
109 #define TWO16383 C[9]
110 L(5.94865747678615882542879663314003565E+4931),
111
112 /* 256 */
113 #define TWO8 C[10]
114 256,
115
116 /* 32768 */
117 #define TWO15 C[11]
118 32768,
119
120 /* Chebyshev polynom coefficients for (exp(x)-1)/x */
121 #define P1 C[12]
122 #define P2 C[13]
123 #define P3 C[14]
124 #define P4 C[15]
125 #define P5 C[16]
126 #define P6 C[17]
127 L(0.5),
128 L(1.66666666666666666666666666666666683E-01),
129 L(4.16666666666666666666654902320001674E-02),
130 L(8.33333333333333333333314659767198461E-03),
131 L(1.38888888889899438565058018857254025E-03),
132 L(1.98412698413981650382436541785404286E-04),
133 };
134
135 _Float128
__ieee754_expl(_Float128 x)136 __ieee754_expl (_Float128 x)
137 {
138 /* Check for usual case. */
139 if (isless (x, himark) && isgreater (x, lomark))
140 {
141 int tval1, tval2, unsafe, n_i;
142 _Float128 x22, n, t, result, xl;
143 union ieee854_long_double ex2_u, scale_u;
144 fenv_t oldenv;
145
146 feholdexcept (&oldenv);
147 #ifdef FE_TONEAREST
148 fesetround (FE_TONEAREST);
149 #endif
150
151 /* Calculate n. */
152 n = x * M_1_LN2 + THREEp111;
153 n -= THREEp111;
154 x = x - n * M_LN2_0;
155 xl = n * M_LN2_1;
156
157 /* Calculate t/256. */
158 t = x + THREEp103;
159 t -= THREEp103;
160
161 /* Compute tval1 = t. */
162 tval1 = (int) (t * TWO8);
163
164 x -= __expl_table[T_EXPL_ARG1+2*tval1];
165 xl -= __expl_table[T_EXPL_ARG1+2*tval1+1];
166
167 /* Calculate t/32768. */
168 t = x + THREEp96;
169 t -= THREEp96;
170
171 /* Compute tval2 = t. */
172 tval2 = (int) (t * TWO15);
173
174 x -= __expl_table[T_EXPL_ARG2+2*tval2];
175 xl -= __expl_table[T_EXPL_ARG2+2*tval2+1];
176
177 x = x + xl;
178
179 /* Compute ex2 = 2^n_0 e^(argtable[tval1]) e^(argtable[tval2]). */
180 ex2_u.d = __expl_table[T_EXPL_RES1 + tval1]
181 * __expl_table[T_EXPL_RES2 + tval2];
182 n_i = (int)n;
183 /* 'unsafe' is 1 iff n_1 != 0. */
184 unsafe = abs(n_i) >= 15000;
185 ex2_u.ieee.exponent += n_i >> unsafe;
186
187 /* Compute scale = 2^n_1. */
188 scale_u.d = 1;
189 scale_u.ieee.exponent += n_i - (n_i >> unsafe);
190
191 /* Approximate e^x2 - 1, using a seventh-degree polynomial,
192 with maximum error in [-2^-16-2^-53,2^-16+2^-53]
193 less than 4.8e-39. */
194 x22 = x + x*x*(P1+x*(P2+x*(P3+x*(P4+x*(P5+x*P6)))));
195 math_force_eval (x22);
196
197 /* Return result. */
198 fesetenv (&oldenv);
199
200 result = x22 * ex2_u.d + ex2_u.d;
201
202 /* Now we can test whether the result is ultimate or if we are unsure.
203 In the later case we should probably call a mpn based routine to give
204 the ultimate result.
205 Empirically, this routine is already ultimate in about 99.9986% of
206 cases, the test below for the round to nearest case will be false
207 in ~ 99.9963% of cases.
208 Without proc2 routine maximum error which has been seen is
209 0.5000262 ulp.
210
211 union ieee854_long_double ex3_u;
212
213 #ifdef FE_TONEAREST
214 fesetround (FE_TONEAREST);
215 #endif
216 ex3_u.d = (result - ex2_u.d) - x22 * ex2_u.d;
217 ex2_u.d = result;
218 ex3_u.ieee.exponent += LDBL_MANT_DIG + 15 + IEEE854_LONG_DOUBLE_BIAS
219 - ex2_u.ieee.exponent;
220 n_i = abs (ex3_u.d);
221 n_i = (n_i + 1) / 2;
222 fesetenv (&oldenv);
223 #ifdef FE_TONEAREST
224 if (fegetround () == FE_TONEAREST)
225 n_i -= 0x4000;
226 #endif
227 if (!n_i) {
228 return __ieee754_expl_proc2 (origx);
229 }
230 */
231 if (!unsafe)
232 return result;
233 else
234 {
235 result *= scale_u.d;
236 math_check_force_underflow_nonneg (result);
237 return result;
238 }
239 }
240 /* Exceptional cases: */
241 else if (isless (x, himark))
242 {
243 if (isinf (x))
244 /* e^-inf == 0, with no error. */
245 return 0;
246 else
247 /* Underflow */
248 return TINY * TINY;
249 }
250 else
251 /* Return x, if x is a NaN or Inf; or overflow, otherwise. */
252 return TWO16383*x;
253 }
254 libm_alias_finite (__ieee754_expl, __expl)
255