3 * Common logarithm, 128-bit long double precision
9 * long double x, y, log10l();
17 * Returns the base 10 logarithm of x.
19 * The argument is separated into its exponent and fractional
20 * parts. If the exponent is between -1 and +1, the logarithm
21 * of the fraction is approximated by
23 * log(1+x) = x - 0.5 x^2 + x^3 P(x)/Q(x).
25 * Otherwise, setting z = 2(x-1)/x+1),
27 * log(x) = z + z^3 P(z)/Q(z).
34 * arithmetic domain # trials peak rms
35 * IEEE 0.5, 2.0 30000 2.3e-34 4.9e-35
36 * IEEE exp(+-10000) 30000 1.0e-34 4.1e-35
38 * In the tests over the interval exp(+-10000), the logarithms
39 * of the random arguments were uniformly distributed over
45 Cephes Math Library Release 2.2: January, 1991
46 Copyright 1984, 1991 by Stephen L. Moshier
47 Adapted for glibc November, 2001
49 This library is free software; you can redistribute it and/or
50 modify it under the terms of the GNU Lesser General Public
51 License as published by the Free Software Foundation; either
52 version 2.1 of the License, or (at your option) any later version.
54 This library is distributed in the hope that it will be useful,
55 but WITHOUT ANY WARRANTY; without even the implied warranty of
56 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
57 Lesser General Public License for more details.
59 You should have received a copy of the GNU Lesser General Public
60 License along with this library; if not, see <http://www.gnu.org/licenses/>.
63 #include "quadmath-imp.h"
65 /* Coefficients for ln(1+x) = x - x**2/2 + x**3 P(x)/Q(x)
66 * 1/sqrt(2) <= x < sqrt(2)
67 * Theoretical peak relative error = 5.3e-37,
68 * relative peak error spread = 2.3e-14
70 static const __float128 P
[13] =
72 1.313572404063446165910279910527789794488E4Q
,
73 7.771154681358524243729929227226708890930E4Q
,
74 2.014652742082537582487669938141683759923E5Q
,
75 3.007007295140399532324943111654767187848E5Q
,
76 2.854829159639697837788887080758954924001E5Q
,
77 1.797628303815655343403735250238293741397E5Q
,
78 7.594356839258970405033155585486712125861E4Q
,
79 2.128857716871515081352991964243375186031E4Q
,
80 3.824952356185897735160588078446136783779E3Q
,
81 4.114517881637811823002128927449878962058E2Q
,
82 2.321125933898420063925789532045674660756E1Q
,
83 4.998469661968096229986658302195402690910E-1Q
,
84 1.538612243596254322971797716843006400388E-6Q
86 static const __float128 Q
[12] =
88 3.940717212190338497730839731583397586124E4Q
,
89 2.626900195321832660448791748036714883242E5Q
,
90 7.777690340007566932935753241556479363645E5Q
,
91 1.347518538384329112529391120390701166528E6Q
,
92 1.514882452993549494932585972882995548426E6Q
,
93 1.158019977462989115839826904108208787040E6Q
,
94 6.132189329546557743179177159925690841200E5Q
,
95 2.248234257620569139969141618556349415120E5Q
,
96 5.605842085972455027590989944010492125825E4Q
,
97 9.147150349299596453976674231612674085381E3Q
,
98 9.104928120962988414618126155557301584078E2Q
,
99 4.839208193348159620282142911143429644326E1Q
100 /* 1.000000000000000000000000000000000000000E0L, */
103 /* Coefficients for log(x) = z + z^3 P(z^2)/Q(z^2),
104 * where z = 2(x-1)/(x+1)
105 * 1/sqrt(2) <= x < sqrt(2)
106 * Theoretical peak relative error = 1.1e-35,
107 * relative peak error spread 1.1e-9
109 static const __float128 R
[6] =
111 1.418134209872192732479751274970992665513E5Q
,
112 -8.977257995689735303686582344659576526998E4Q
,
113 2.048819892795278657810231591630928516206E4Q
,
114 -2.024301798136027039250415126250455056397E3Q
,
115 8.057002716646055371965756206836056074715E1Q
,
116 -8.828896441624934385266096344596648080902E-1Q
118 static const __float128 S
[6] =
120 1.701761051846631278975701529965589676574E6Q
,
121 -1.332535117259762928288745111081235577029E6Q
,
122 4.001557694070773974936904547424676279307E5Q
,
123 -5.748542087379434595104154610899551484314E4Q
,
124 3.998526750980007367835804959888064681098E3Q
,
125 -1.186359407982897997337150403816839480438E2Q
126 /* 1.000000000000000000000000000000000000000E0L, */
129 static const __float128
132 L102B
= -1.14700043360188047862611052755069732318101185E-2Q
,
135 L10EB
= -6.570551809674817234887108108339491770560299E-2Q
,
137 SQRTH
= 7.071067811865475244008443621048490392848359E-1Q
;
141 /* Evaluate P[n] x^n + P[n-1] x^(n-1) + ... + P[0] */
144 neval (__float128 x
, const __float128
*p
, int n
)
159 /* Evaluate x^n+1 + P[n] x^(n) + P[n-1] x^(n-1) + ... + P[0] */
162 deval (__float128 x
, const __float128
*p
, int n
)
179 log10q (__float128 x
)
186 /* Test for domain */
187 GET_FLT128_WORDS64 (hx
, lx
, x
);
188 if (((hx
& 0x7fffffffffffffffLL
) | lx
) == 0)
189 return (-1 / fabsq (x
)); /* log10l(+-0)=-inf */
191 return (x
- x
) / (x
- x
);
192 if (hx
>= 0x7fff000000000000LL
)
198 /* separate mantissa from exponent */
200 /* Note, frexp is used so that denormal numbers
201 * will be handled properly.
206 /* logarithm using log(x) = z + z**3 P(z)/Q(z),
207 * where z = 2(x-1)/x+1)
209 if ((e
> 2) || (e
< -2))
212 { /* 2( 2x-1 )/( 2x+1 ) */
218 { /* 2 (x-1)/(x+1) */
225 y
= x
* (z
* neval (z
, R
, 5) / deval (z
, S
, 5));
230 /* logarithm using log(1+x) = x - .5x**2 + x**3 P(x)/Q(x) */
235 x
= 2.0 * x
- 1; /* 2x - 1 */
242 y
= x
* (z
* neval (x
, P
, 12) / deval (x
, Q
, 11));
247 /* Multiply log of fraction by log10(e)
248 * and base 2 exponent by log10(2).