Sprout/sprout/math/float_sig_exp.hpp
2013-04-26 00:25:35 +09:00

55 lines
1.9 KiB
C++

#ifndef SPROUT_MATH_FLOAT_SIG_EXP_HPP
#define SPROUT_MATH_FLOAT_SIG_EXP_HPP
#include <climits>
#include <limits>
#include <type_traits>
#include <sprout/config.hpp>
#include <sprout/detail/pow.hpp>
#include <sprout/math/detail/config.hpp>
#include <sprout/math/isnan.hpp>
#include <sprout/math/ilogb.hpp>
#include <sprout/type_traits/enabler_if.hpp>
#include <sprout/utility/pair/pair.hpp>
namespace sprout {
namespace math {
namespace detail {
template<typename T>
inline SPROUT_CONSTEXPR sprout::pair<T, int>
float_sig_exp_impl(T x, int exp) {
typedef sprout::pair<T, int> type;
return type(x / sprout::detail::pow_n(T(std::numeric_limits<T>::radix), exp), exp);
}
template<
typename FloatType,
typename sprout::enabler_if<std::is_floating_point<FloatType>::value>::type = sprout::enabler
>
inline SPROUT_CONSTEXPR sprout::pair<FloatType, int>
float_sig_exp(FloatType x) {
typedef sprout::pair<FloatType, int> type;
return x == std::numeric_limits<FloatType>::infinity() ? type(std::numeric_limits<FloatType>::infinity(), 0)
: x == -std::numeric_limits<FloatType>::infinity() ? type(-std::numeric_limits<FloatType>::infinity(), 0)
: sprout::math::isnan(x) ? type(std::numeric_limits<FloatType>::quiet_NaN(), FP_ILOGBNAN)
: x == 0 ? type(x, 0)
: sprout::math::detail::float_sig_exp_impl(x, sprout::math::ilogb(x) + 1)
;
}
template<
typename IntType,
typename sprout::enabler_if<std::is_integral<IntType>::value>::type = sprout::enabler
>
inline SPROUT_CONSTEXPR sprout::pair<double, int>
float_sig_exp(IntType x) {
return sprout::math::detail::float_sig_exp(static_cast<double>(x));
}
} // namespace detail
using sprout::math::detail::float_sig_exp;
} // namespace math
using sprout::math::float_sig_exp;
} // namespace sprout
#endif // #ifndef SPROUT_MATH_FLOAT_SIG_EXP_HPP