1 | // This file is part of Eigen, a lightweight C++ template library |
2 | // for linear algebra. |
3 | // |
4 | // Copyright (C) 2014 Pedro Gonnet (pedro.gonnet@gmail.com) |
5 | // Copyright (C) 2016 Gael Guennebaud <gael.guennebaud@inria.fr> |
6 | // |
7 | // This Source Code Form is subject to the terms of the Mozilla |
8 | // Public License v. 2.0. If a copy of the MPL was not distributed |
9 | // with this file, You can obtain one at http://mozilla.org/MPL/2.0/. |
10 | |
11 | #ifndef EIGEN_MATHFUNCTIONSIMPL_H |
12 | #define EIGEN_MATHFUNCTIONSIMPL_H |
13 | |
14 | namespace Eigen { |
15 | |
16 | namespace internal { |
17 | |
18 | /** \internal \returns the hyperbolic tan of \a a (coeff-wise) |
19 | Doesn't do anything fancy, just a 13/6-degree rational interpolant which |
20 | is accurate up to a couple of ulp in the range [-9, 9], outside of which |
21 | the tanh(x) = +/-1. |
22 | |
23 | This implementation works on both scalars and packets. |
24 | */ |
25 | template<typename T> |
26 | T generic_fast_tanh_float(const T& a_x) |
27 | { |
28 | // Clamp the inputs to the range [-9, 9] since anything outside |
29 | // this range is +/-1.0f in single-precision. |
30 | const T plus_9 = pset1<T>(9.f); |
31 | const T minus_9 = pset1<T>(-9.f); |
32 | // NOTE GCC prior to 6.3 might improperly optimize this max/min |
33 | // step such that if a_x is nan, x will be either 9 or -9, |
34 | // and tanh will return 1 or -1 instead of nan. |
35 | // This is supposed to be fixed in gcc6.3, |
36 | // see: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=72867 |
37 | const T x = pmax(minus_9,pmin(plus_9,a_x)); |
38 | // The monomial coefficients of the numerator polynomial (odd). |
39 | const T alpha_1 = pset1<T>(4.89352455891786e-03f); |
40 | const T alpha_3 = pset1<T>(6.37261928875436e-04f); |
41 | const T alpha_5 = pset1<T>(1.48572235717979e-05f); |
42 | const T alpha_7 = pset1<T>(5.12229709037114e-08f); |
43 | const T alpha_9 = pset1<T>(-8.60467152213735e-11f); |
44 | const T alpha_11 = pset1<T>(2.00018790482477e-13f); |
45 | const T alpha_13 = pset1<T>(-2.76076847742355e-16f); |
46 | |
47 | // The monomial coefficients of the denominator polynomial (even). |
48 | const T beta_0 = pset1<T>(4.89352518554385e-03f); |
49 | const T beta_2 = pset1<T>(2.26843463243900e-03f); |
50 | const T beta_4 = pset1<T>(1.18534705686654e-04f); |
51 | const T beta_6 = pset1<T>(1.19825839466702e-06f); |
52 | |
53 | // Since the polynomials are odd/even, we need x^2. |
54 | const T x2 = pmul(x, x); |
55 | |
56 | // Evaluate the numerator polynomial p. |
57 | T p = pmadd(x2, alpha_13, alpha_11); |
58 | p = pmadd(x2, p, alpha_9); |
59 | p = pmadd(x2, p, alpha_7); |
60 | p = pmadd(x2, p, alpha_5); |
61 | p = pmadd(x2, p, alpha_3); |
62 | p = pmadd(x2, p, alpha_1); |
63 | p = pmul(x, p); |
64 | |
65 | // Evaluate the denominator polynomial p. |
66 | T q = pmadd(x2, beta_6, beta_4); |
67 | q = pmadd(x2, q, beta_2); |
68 | q = pmadd(x2, q, beta_0); |
69 | |
70 | // Divide the numerator by the denominator. |
71 | return pdiv(p, q); |
72 | } |
73 | |
74 | template<typename RealScalar> |
75 | EIGEN_STRONG_INLINE |
76 | RealScalar positive_real_hypot(const RealScalar& x, const RealScalar& y) |
77 | { |
78 | EIGEN_USING_STD_MATH(sqrt); |
79 | RealScalar p, qp; |
80 | p = numext::maxi(x,y); |
81 | if(p==RealScalar(0)) return RealScalar(0); |
82 | qp = numext::mini(y,x) / p; |
83 | return p * sqrt(RealScalar(1) + qp*qp); |
84 | } |
85 | |
86 | template<typename Scalar> |
87 | struct hypot_impl |
88 | { |
89 | typedef typename NumTraits<Scalar>::Real RealScalar; |
90 | static inline RealScalar run(const Scalar& x, const Scalar& y) |
91 | { |
92 | EIGEN_USING_STD_MATH(abs); |
93 | return positive_real_hypot<RealScalar>(abs(x), abs(y)); |
94 | } |
95 | }; |
96 | |
97 | } // end namespace internal |
98 | |
99 | } // end namespace Eigen |
100 | |
101 | #endif // EIGEN_MATHFUNCTIONSIMPL_H |
102 | |