mirror of
https://github.com/boostorg/multiprecision.git
synced 2025-05-12 14:01:48 +00:00
601 lines
20 KiB
C++
601 lines
20 KiB
C++
///////////////////////////////////////////////////////////////
|
|
// Copyright 2012 John Maddock. Distributed under the Boost
|
|
// Software License, Version 1.0. (See accompanying file
|
|
// LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_
|
|
|
|
#ifndef BOOST_MATH_FLOAT_BACKEND_HPP
|
|
#define BOOST_MATH_FLOAT_BACKEND_HPP
|
|
|
|
#include <iostream>
|
|
#include <iomanip>
|
|
#include <sstream>
|
|
#include <boost/cstdint.hpp>
|
|
#include <boost/lexical_cast.hpp>
|
|
#include <boost/math/concepts/real_concept.hpp>
|
|
#include <boost/multiprecision/number.hpp>
|
|
#include <boost/math/common_factor_rt.hpp>
|
|
|
|
namespace boost{
|
|
namespace multiprecision{
|
|
namespace backends{
|
|
|
|
#ifdef BOOST_MSVC
|
|
# pragma warning(push)
|
|
# pragma warning(disable:4389 4244 4018 4244 4127)
|
|
#endif
|
|
|
|
template <class Arithmetic>
|
|
struct arithmetic_backend
|
|
{
|
|
typedef mpl::list<short, int, long, long long> signed_types;
|
|
typedef mpl::list<unsigned short, unsigned, unsigned long, unsigned long long> unsigned_types;
|
|
typedef mpl::list<float, double, long double> float_types;
|
|
typedef int exponent_type;
|
|
|
|
arithmetic_backend(){}
|
|
arithmetic_backend(const arithmetic_backend& o)
|
|
{
|
|
m_value = o.m_value;
|
|
}
|
|
template <class A>
|
|
arithmetic_backend(const A& o, const typename enable_if<is_arithmetic<A> >::type* = 0) : m_value(o) {}
|
|
template <class A>
|
|
arithmetic_backend(const arithmetic_backend<A>& o) : m_value(o.data()) {}
|
|
arithmetic_backend& operator = (const arithmetic_backend& o)
|
|
{
|
|
m_value = o.m_value;
|
|
return *this;
|
|
}
|
|
template <class A>
|
|
typename enable_if<is_arithmetic<A>, arithmetic_backend&>::type operator = (A i)
|
|
{
|
|
m_value = i;
|
|
return *this;
|
|
}
|
|
template <class A>
|
|
arithmetic_backend& operator = (const arithmetic_backend<A>& i)
|
|
{
|
|
m_value = i.data();
|
|
return *this;
|
|
}
|
|
arithmetic_backend& operator = (const char* s)
|
|
{
|
|
try
|
|
{
|
|
m_value = boost::lexical_cast<Arithmetic>(s);
|
|
}
|
|
catch(const bad_lexical_cast&)
|
|
{
|
|
throw std::runtime_error(std::string("Unable to interpret the string provided: \"") + s + std::string("\" as a compatible number type."));
|
|
}
|
|
return *this;
|
|
}
|
|
void swap(arithmetic_backend& o)
|
|
{
|
|
std::swap(m_value, o.m_value);
|
|
}
|
|
std::string str(std::streamsize digits, std::ios_base::fmtflags f)const
|
|
{
|
|
std::stringstream ss;
|
|
ss.flags(f);
|
|
ss << std::setprecision(digits ? digits : std::numeric_limits<Arithmetic>::digits10 + 4) << m_value;
|
|
return ss.str();
|
|
}
|
|
void do_negate(const mpl::true_&)
|
|
{
|
|
m_value = 1 + ~m_value;
|
|
}
|
|
void do_negate(const mpl::false_&)
|
|
{
|
|
m_value = -m_value;
|
|
}
|
|
void negate()
|
|
{
|
|
do_negate(is_unsigned<Arithmetic>());
|
|
}
|
|
int compare(const arithmetic_backend& o)const
|
|
{
|
|
return m_value > o.m_value ? 1 : (m_value < o.m_value ? -1 : 0);
|
|
}
|
|
template <class A>
|
|
typename enable_if<is_arithmetic<A>, int>::type compare(A i)const
|
|
{
|
|
return m_value > static_cast<Arithmetic>(i) ? 1 : (m_value < static_cast<Arithmetic>(i) ? -1 : 0);
|
|
}
|
|
Arithmetic& data() { return m_value; }
|
|
const Arithmetic& data()const { return m_value; }
|
|
private:
|
|
Arithmetic m_value;
|
|
};
|
|
|
|
template <class R, class Arithmetic>
|
|
inline void eval_convert_to(R* result, const arithmetic_backend<Arithmetic>& backend)
|
|
{
|
|
*result = backend.data();
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline bool eval_eq(const arithmetic_backend<Arithmetic>& a, const arithmetic_backend<Arithmetic>& b)
|
|
{
|
|
return a.data() == b.data();
|
|
}
|
|
template <class Arithmetic, class A2>
|
|
inline typename enable_if<is_arithmetic<A2>, bool>::type eval_eq(const arithmetic_backend<Arithmetic>& a, const A2& b)
|
|
{
|
|
return a.data() == static_cast<Arithmetic>(b);
|
|
}
|
|
template <class Arithmetic>
|
|
inline bool eval_lt(const arithmetic_backend<Arithmetic>& a, const arithmetic_backend<Arithmetic>& b)
|
|
{
|
|
return a.data() < b.data();
|
|
}
|
|
template <class Arithmetic, class A2>
|
|
inline typename enable_if<is_arithmetic<A2>, bool>::type eval_lt(const arithmetic_backend<Arithmetic>& a, const A2& b)
|
|
{
|
|
return a.data() < static_cast<Arithmetic>(b);
|
|
}
|
|
template <class Arithmetic>
|
|
inline bool eval_gt(const arithmetic_backend<Arithmetic>& a, const arithmetic_backend<Arithmetic>& b)
|
|
{
|
|
return a.data() > b.data();
|
|
}
|
|
template <class Arithmetic, class A2>
|
|
inline typename enable_if<is_arithmetic<A2>, bool>::type eval_gt(const arithmetic_backend<Arithmetic>& a, const A2& b)
|
|
{
|
|
return a.data() > static_cast<Arithmetic>(b);
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_add(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
result.data() += o.data();
|
|
}
|
|
template <class Arithmetic>
|
|
inline void eval_subtract(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
result.data() -= o.data();
|
|
}
|
|
template <class Arithmetic>
|
|
inline void eval_multiply(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
result.data() *= o.data();
|
|
}
|
|
template <class Arithmetic>
|
|
inline typename enable_if_c<std::numeric_limits<Arithmetic>::has_infinity>::type eval_divide(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
result.data() /= o.data();
|
|
}
|
|
template <class Arithmetic>
|
|
inline typename disable_if_c<std::numeric_limits<Arithmetic>::has_infinity>::type eval_divide(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
if(!o.data())
|
|
BOOST_THROW_EXCEPTION(std::overflow_error("Divide by zero"));
|
|
result.data() /= o.data();
|
|
}
|
|
|
|
template <class Arithmetic, class A2>
|
|
inline typename enable_if<is_arithmetic<A2> >::type eval_add(arithmetic_backend<Arithmetic>& result, const A2& o)
|
|
{
|
|
result.data() += o;
|
|
}
|
|
template <class Arithmetic, class A2>
|
|
inline typename enable_if<is_arithmetic<A2> >::type eval_subtract(arithmetic_backend<Arithmetic>& result, const A2& o)
|
|
{
|
|
result.data() -= o;
|
|
}
|
|
template <class Arithmetic, class A2>
|
|
inline typename enable_if<is_arithmetic<A2> >::type eval_multiply(arithmetic_backend<Arithmetic>& result, const A2& o)
|
|
{
|
|
result.data() *= o;
|
|
}
|
|
template <class Arithmetic, class A2>
|
|
inline typename enable_if_c<(is_arithmetic<A2>::value && !std::numeric_limits<Arithmetic>::has_infinity)>::type
|
|
eval_divide(arithmetic_backend<Arithmetic>& result, const A2& o)
|
|
{
|
|
if(!o)
|
|
BOOST_THROW_EXCEPTION(std::overflow_error("Divide by zero"));
|
|
result.data() /= o;
|
|
}
|
|
template <class Arithmetic, class A2>
|
|
inline typename enable_if_c<(is_arithmetic<A2>::value && std::numeric_limits<Arithmetic>::has_infinity)>::type
|
|
eval_divide(arithmetic_backend<Arithmetic>& result, const A2& o)
|
|
{
|
|
result.data() /= o;
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_add(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const arithmetic_backend<Arithmetic>& b)
|
|
{
|
|
result.data() = a.data() + b.data();
|
|
}
|
|
template <class Arithmetic>
|
|
inline void eval_subtract(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const arithmetic_backend<Arithmetic>& b)
|
|
{
|
|
result.data() = a.data() - b.data();
|
|
}
|
|
template <class Arithmetic>
|
|
inline void eval_multiply(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const arithmetic_backend<Arithmetic>& b)
|
|
{
|
|
result.data() = a.data() * b.data();
|
|
}
|
|
template <class Arithmetic>
|
|
inline typename enable_if_c<std::numeric_limits<Arithmetic>::has_infinity>::type eval_divide(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const arithmetic_backend<Arithmetic>& b)
|
|
{
|
|
result.data() = a.data() / b.data();
|
|
}
|
|
template <class Arithmetic>
|
|
inline typename disable_if_c<std::numeric_limits<Arithmetic>::has_infinity>::type eval_divide(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const arithmetic_backend<Arithmetic>& b)
|
|
{
|
|
if(!b.data())
|
|
BOOST_THROW_EXCEPTION(std::overflow_error("Divide by zero"));
|
|
result.data() = a.data() / b.data();
|
|
}
|
|
|
|
template <class Arithmetic, class A2>
|
|
inline typename enable_if<is_arithmetic<A2> >::type eval_add(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const A2& b)
|
|
{
|
|
result.data() = a.data() + b;
|
|
}
|
|
template <class Arithmetic, class A2>
|
|
inline typename enable_if<is_arithmetic<A2> >::type eval_subtract(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const A2& b)
|
|
{
|
|
result.data() = a.data() - b;
|
|
}
|
|
template <class Arithmetic, class A2>
|
|
inline typename enable_if<is_arithmetic<A2> >::type eval_multiply(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const A2& b)
|
|
{
|
|
result.data() = a.data() * b;
|
|
}
|
|
template <class Arithmetic, class A2>
|
|
inline typename enable_if_c<(is_arithmetic<A2>::value && !std::numeric_limits<Arithmetic>::has_infinity)>::type
|
|
eval_divide(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const A2& b)
|
|
{
|
|
if(!b)
|
|
BOOST_THROW_EXCEPTION(std::overflow_error("Divide by zero"));
|
|
result.data() = a.data() / b;
|
|
}
|
|
template <class Arithmetic, class A2>
|
|
inline typename enable_if_c<(is_arithmetic<A2>::value && std::numeric_limits<Arithmetic>::has_infinity)>::type
|
|
eval_divide(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const A2& b)
|
|
{
|
|
result.data() = a.data() / b;
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline bool eval_is_zero(const arithmetic_backend<Arithmetic>& val)
|
|
{
|
|
return val.data() == 0;
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline typename enable_if_c<
|
|
(!std::numeric_limits<Arithmetic>::is_specialized
|
|
|| std::numeric_limits<Arithmetic>::is_signed), int>::type
|
|
eval_get_sign(const arithmetic_backend<Arithmetic>& val)
|
|
{
|
|
return val.data() == 0 ? 0 : val.data() < 0 ? -1 : 1;
|
|
}
|
|
template <class Arithmetic>
|
|
inline typename disable_if_c<
|
|
(std::numeric_limits<Arithmetic>::is_specialized
|
|
|| std::numeric_limits<Arithmetic>::is_signed), int>::type
|
|
eval_get_sign(const arithmetic_backend<Arithmetic>& val)
|
|
{
|
|
return val.data() == 0 ? 0 : 1;
|
|
}
|
|
|
|
template <class T>
|
|
inline typename enable_if<is_unsigned<T>, T>::type abs(T v) { return v; }
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_abs(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
using std::abs;
|
|
using boost::multiprecision::backends::abs;
|
|
result.data() = abs(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_fabs(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
result.data() = std::abs(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_floor(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = floor(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_ceil(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = ceil(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_sqrt(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = sqrt(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline int eval_fpclassify(const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
return (boost::math::fpclassify)(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_trunc(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = trunc(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_round(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = round(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_frexp(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, int* v)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = frexp(a.data(), v);
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_ldexp(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, int v)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = ldexp(a.data(), v);
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_exp(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = exp(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_log(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = log(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_log10(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = log10(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_sin(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = sin(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_cos(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = cos(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_tan(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = tan(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_acos(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = acos(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_asin(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = asin(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_atan(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = atan(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_sinh(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = sinh(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_cosh(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = cosh(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_tanh(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& o)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = tanh(o.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_fmod(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const arithmetic_backend<Arithmetic>& b)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = fmod(a.data(), b.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_pow(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const arithmetic_backend<Arithmetic>& b)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = pow(a.data(), b.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_atan2(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const arithmetic_backend<Arithmetic>& b)
|
|
{
|
|
BOOST_MATH_STD_USING
|
|
result.data() = atan2(a.data(), b.data());
|
|
}
|
|
|
|
template <class Arithmetic, class I>
|
|
inline void eval_left_shift(arithmetic_backend<Arithmetic>& result, I val)
|
|
{
|
|
result.data() <<= val;
|
|
}
|
|
|
|
template <class Arithmetic, class I>
|
|
inline void eval_right_shift(arithmetic_backend<Arithmetic>& result, I val)
|
|
{
|
|
result.data() >>= val;
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_modulus(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a)
|
|
{
|
|
result.data() %= a.data();
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_bitwise_and(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a)
|
|
{
|
|
result.data() &= a.data();
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_bitwise_or(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a)
|
|
{
|
|
result.data() |= a.data();
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_bitwise_xor(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a)
|
|
{
|
|
result.data() ^= a.data();
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_complement(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a)
|
|
{
|
|
result.data() = ~a.data();
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_gcd(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const arithmetic_backend<Arithmetic>& b)
|
|
{
|
|
result.data() = boost::math::gcd(a.data(), b.data());
|
|
}
|
|
|
|
template <class Arithmetic>
|
|
inline void eval_lcm(arithmetic_backend<Arithmetic>& result, const arithmetic_backend<Arithmetic>& a, const arithmetic_backend<Arithmetic>& b)
|
|
{
|
|
result.data() = boost::math::lcm(a.data(), b.data());
|
|
}
|
|
|
|
#ifdef BOOST_MSVC
|
|
# pragma warning(pop)
|
|
#endif
|
|
|
|
} // namespace backends
|
|
|
|
using boost::multiprecision::backends::arithmetic_backend;
|
|
|
|
template <class Arithmetic>
|
|
struct number_category<arithmetic_backend<Arithmetic> > : public mpl::int_<is_integral<Arithmetic>::value ? number_kind_integer : number_kind_floating_point>{};
|
|
|
|
namespace detail{
|
|
|
|
template <class Backend>
|
|
struct double_precision_type;
|
|
|
|
template<class Arithmetic, boost::multiprecision::expression_template_option ET>
|
|
struct double_precision_type<number<arithmetic_backend<Arithmetic>, ET> >
|
|
{
|
|
typedef number<arithmetic_backend<typename double_precision_type<Arithmetic>::type>, ET> type;
|
|
};
|
|
template<>
|
|
struct double_precision_type<arithmetic_backend<boost::int32_t> >
|
|
{
|
|
typedef arithmetic_backend<boost::int64_t> type;
|
|
};
|
|
|
|
}
|
|
|
|
}} // namespaces
|
|
|
|
namespace boost{ namespace math{ namespace tools{
|
|
|
|
template <>
|
|
inline double real_cast<double, concepts::real_concept>(concepts::real_concept r)
|
|
{
|
|
return static_cast<double>(r.value());
|
|
}
|
|
|
|
}}}
|
|
|
|
|
|
namespace std{
|
|
|
|
template <class Arithmetic, boost::multiprecision::expression_template_option ExpressionTemplates>
|
|
class numeric_limits<boost::multiprecision::number<boost::multiprecision::arithmetic_backend<Arithmetic>, ExpressionTemplates > > : public std::numeric_limits<Arithmetic>
|
|
{
|
|
typedef std::numeric_limits<Arithmetic> base_type;
|
|
typedef boost::multiprecision::number<boost::multiprecision::arithmetic_backend<Arithmetic>, ExpressionTemplates> number_type;
|
|
public:
|
|
BOOST_STATIC_CONSTEXPR number_type (min)() BOOST_NOEXCEPT { return (base_type::min)(); }
|
|
BOOST_STATIC_CONSTEXPR number_type (max)() BOOST_NOEXCEPT { return (base_type::max)(); }
|
|
BOOST_STATIC_CONSTEXPR number_type lowest() BOOST_NOEXCEPT { return -(max)(); }
|
|
BOOST_STATIC_CONSTEXPR number_type epsilon() BOOST_NOEXCEPT { return base_type::epsilon(); }
|
|
BOOST_STATIC_CONSTEXPR number_type round_error() BOOST_NOEXCEPT { return epsilon() / 2; }
|
|
BOOST_STATIC_CONSTEXPR number_type infinity() BOOST_NOEXCEPT { return base_type::infinity(); }
|
|
BOOST_STATIC_CONSTEXPR number_type quiet_NaN() BOOST_NOEXCEPT { return base_type::quiet_NaN(); }
|
|
BOOST_STATIC_CONSTEXPR number_type signaling_NaN() BOOST_NOEXCEPT { return base_type::signaling_NaN(); }
|
|
BOOST_STATIC_CONSTEXPR number_type denorm_min() BOOST_NOEXCEPT { return base_type::denorm_min(); }
|
|
};
|
|
|
|
template<>
|
|
class numeric_limits<boost::math::concepts::real_concept> : public std::numeric_limits<long double>
|
|
{
|
|
typedef std::numeric_limits<long double> base_type;
|
|
typedef boost::math::concepts::real_concept number_type;
|
|
public:
|
|
static const number_type (min)() BOOST_NOEXCEPT { return (base_type::min)(); }
|
|
static const number_type (max)() BOOST_NOEXCEPT { return (base_type::max)(); }
|
|
static const number_type lowest() BOOST_NOEXCEPT { return -(max)(); }
|
|
static const number_type epsilon() BOOST_NOEXCEPT { return base_type::epsilon(); }
|
|
static const number_type round_error() BOOST_NOEXCEPT { return epsilon() / 2; }
|
|
static const number_type infinity() BOOST_NOEXCEPT { return base_type::infinity(); }
|
|
static const number_type quiet_NaN() BOOST_NOEXCEPT { return base_type::quiet_NaN(); }
|
|
static const number_type signaling_NaN() BOOST_NOEXCEPT { return base_type::signaling_NaN(); }
|
|
static const number_type denorm_min() BOOST_NOEXCEPT { return base_type::denorm_min(); }
|
|
};
|
|
|
|
}
|
|
|
|
#include <boost/multiprecision/detail/integer_ops.hpp>
|
|
|
|
#endif
|