1 ///////////////////////////////////////////////////////////////
2 //  Copyright 2013 John Maddock. Distributed under the Boost
3 //  Software License, Version 1.0. (See accompanying file
4 //  LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_
5 
6 #ifndef BOOST_MULTIPRECISION_CPP_BIN_FLOAT_TRANSCENDENTAL_HPP
7 #define BOOST_MULTIPRECISION_CPP_BIN_FLOAT_TRANSCENDENTAL_HPP
8 
9 namespace boost{ namespace multiprecision{ namespace backends{
10 
11 template <unsigned Digits, digit_base_type DigitBase, class Allocator, class Exponent, Exponent MinE, Exponent MaxE>
eval_exp_taylor(cpp_bin_float<Digits,DigitBase,Allocator,Exponent,MinE,MaxE> & res,const cpp_bin_float<Digits,DigitBase,Allocator,Exponent,MinE,MaxE> & arg)12 void eval_exp_taylor(cpp_bin_float<Digits, DigitBase, Allocator, Exponent, MinE, MaxE> &res, const cpp_bin_float<Digits, DigitBase, Allocator, Exponent, MinE, MaxE> &arg)
13 {
14    static const int bits = cpp_bin_float<Digits, DigitBase, Allocator, Exponent, MinE, MaxE>::bit_count;
15    //
16    // Taylor series for small argument, note returns exp(x) - 1:
17    //
18    res = limb_type(0);
19    cpp_bin_float<Digits, DigitBase, Allocator, Exponent, MinE, MaxE> num(arg), denom, t;
20    denom = limb_type(1);
21    eval_add(res, num);
22 
23    for(unsigned k = 2; ; ++k)
24    {
25       eval_multiply(denom, k);
26       eval_multiply(num, arg);
27       eval_divide(t, num, denom);
28       eval_add(res, t);
29       if(eval_is_zero(t) || (res.exponent() - bits > t.exponent()))
30          break;
31    }
32 }
33 
34 template <unsigned Digits, digit_base_type DigitBase, class Allocator, class Exponent, Exponent MinE, Exponent MaxE>
eval_exp(cpp_bin_float<Digits,DigitBase,Allocator,Exponent,MinE,MaxE> & res,const cpp_bin_float<Digits,DigitBase,Allocator,Exponent,MinE,MaxE> & arg)35 void eval_exp(cpp_bin_float<Digits, DigitBase, Allocator, Exponent, MinE, MaxE> &res, const cpp_bin_float<Digits, DigitBase, Allocator, Exponent, MinE, MaxE> &arg)
36 {
37    //
38    // This is based on MPFR's method, let:
39    //
40    // n = floor(x / ln(2))
41    //
42    // Then:
43    //
44    // r = x - n ln(2) : 0 <= r < ln(2)
45    //
46    // We can reduce r further by dividing by 2^k, with k ~ sqrt(n),
47    // so if:
48    //
49    // e0 = exp(r / 2^k) - 1
50    //
51    // With e0 evaluated by taylor series for small arguments, then:
52    //
53    // exp(x) = 2^n (1 + e0)^2^k
54    //
55    // Note that to preserve precision we actually square (1 + e0) k times, calculating
56    // the result less one each time, i.e.
57    //
58    // (1 + e0)^2 - 1 = e0^2 + 2e0
59    //
60    // Then add the final 1 at the end, given that e0 is small, this effectively wipes
61    // out the error in the last step.
62    //
63    using default_ops::eval_multiply;
64    using default_ops::eval_subtract;
65    using default_ops::eval_add;
66    using default_ops::eval_convert_to;
67    using default_ops::eval_increment;
68 
69    int type = eval_fpclassify(arg);
70    bool isneg = eval_get_sign(arg) < 0;
71    if(type == (int)FP_NAN)
72    {
73       res = arg;
74       errno = EDOM;
75       return;
76    }
77    else if(type == (int)FP_INFINITE)
78    {
79       res = arg;
80       if(isneg)
81          res = limb_type(0u);
82       else
83          res = arg;
84       return;
85    }
86    else if(type == (int)FP_ZERO)
87    {
88       res = limb_type(1);
89       return;
90    }
91    cpp_bin_float<Digits, DigitBase, Allocator, Exponent, MinE, MaxE> t, n;
92    if(isneg)
93    {
94       t = arg;
95       t.negate();
96       eval_exp(res, t);
97       t.swap(res);
98       res = limb_type(1);
99       eval_divide(res, t);
100       return;
101    }
102 
103    eval_divide(n, arg, default_ops::get_constant_ln2<cpp_bin_float<Digits, DigitBase, Allocator, Exponent, MinE, MaxE> >());
104    eval_floor(n, n);
105    eval_multiply(t, n, default_ops::get_constant_ln2<cpp_bin_float<Digits, DigitBase, Allocator, Exponent, MinE, MaxE> >());
106    eval_subtract(t, arg);
107    t.negate();
108    if (t.compare(default_ops::get_constant_ln2<cpp_bin_float<Digits, DigitBase, Allocator, Exponent, MinE, MaxE> >()) > 0)
109    {
110       // There are some rare cases where the multiply rounds down leaving a remainder > ln2
111       // See https://github.com/boostorg/multiprecision/issues/120
112       eval_increment(n);
113       t = limb_type(0);
114    }
115    if(eval_get_sign(t) < 0)
116    {
117       // There are some very rare cases where arg/ln2 is an integer, and the subsequent multiply
118       // rounds up, in that situation t ends up negative at this point which breaks our invariants below:
119       t = limb_type(0);
120    }
121 
122    Exponent k, nn;
123    eval_convert_to(&nn, n);
124 
125    if (nn == (std::numeric_limits<Exponent>::max)())
126    {
127       // The result will necessarily oveflow:
128       res = std::numeric_limits<number<cpp_bin_float<Digits, DigitBase, Allocator, Exponent, MinE, MaxE> > >::infinity().backend();
129       return;
130    }
131 
132    BOOST_ASSERT(t.compare(default_ops::get_constant_ln2<cpp_bin_float<Digits, DigitBase, Allocator, Exponent, MinE, MaxE> >()) < 0);
133 
134    k = nn ? Exponent(1) << (msb(nn) / 2) : 0;
135    k = (std::min)(k, (Exponent)(cpp_bin_float<Digits, DigitBase, Allocator, Exponent, MinE, MaxE>::bit_count / 4));
136    eval_ldexp(t, t, -k);
137 
138    eval_exp_taylor(res, t);
139    //
140    // Square 1 + res k times:
141    //
142    for(Exponent s = 0; s < k; ++s)
143    {
144       t.swap(res);
145       eval_multiply(res, t, t);
146       eval_ldexp(t, t, 1);
147       eval_add(res, t);
148    }
149    eval_add(res, limb_type(1));
150    eval_ldexp(res, res, nn);
151 }
152 
153 }}} // namespaces
154 
155 #endif
156 
157