mirror of
https://github.com/boostorg/math.git
synced 2026-01-29 19:52:08 +00:00
Change not to use cstdfloat as that header may not actually define any types!
This commit is contained in:
@@ -328,7 +328,7 @@ namespace tools
|
||||
{
|
||||
|
||||
template <>
|
||||
inline concepts::real_concept make_big_value<concepts::real_concept>(boost::floatmax_t val, const char* , mpl::false_ const&, mpl::false_ const&)
|
||||
inline concepts::real_concept make_big_value<concepts::real_concept>(boost::math::tools::largest_float val, const char* , mpl::false_ const&, mpl::false_ const&)
|
||||
{
|
||||
return val; // Can't use lexical_cast here, sometimes it fails....
|
||||
}
|
||||
|
||||
@@ -342,7 +342,7 @@ namespace tools
|
||||
{
|
||||
|
||||
template <>
|
||||
inline concepts::std_real_concept make_big_value<concepts::std_real_concept>(boost::floatmax_t val, const char*, mpl::false_ const&, mpl::false_ const&)
|
||||
inline concepts::std_real_concept make_big_value<concepts::std_real_concept>(boost::math::tools::largest_float val, const char*, mpl::false_ const&, mpl::false_ const&)
|
||||
{
|
||||
return val; // Can't use lexical_cast here, sometimes it fails....
|
||||
}
|
||||
|
||||
@@ -12,31 +12,48 @@
|
||||
#include <boost/lexical_cast.hpp>
|
||||
#endif
|
||||
#include <boost/type_traits/is_convertible.hpp>
|
||||
#include <boost/math/cstdfloat/cstdfloat_types.hpp>
|
||||
|
||||
namespace boost{ namespace math{
|
||||
|
||||
namespace tools{
|
||||
|
||||
template <class T>
|
||||
inline BOOST_CONSTEXPR_OR_CONST T make_big_value(boost::floatmax_t v, const char*, mpl::true_ const&, mpl::false_ const&)
|
||||
struct numeric_traits : public std::numeric_limits< T > {};
|
||||
|
||||
#ifdef BOOST_MATH_USE_FLOAT128
|
||||
typedef __float128 largest_float;
|
||||
#define BOOST_MATH_LARGEST_FLOAT_C(x) x##Q
|
||||
template <>
|
||||
struct numeric_traits<__float128>
|
||||
{
|
||||
static const int digits = 113;
|
||||
static const int digits10 = 34;
|
||||
static const int max_exponent = 16384;
|
||||
};
|
||||
#else
|
||||
typedef long double largest_float;
|
||||
#define BOOST_MATH_LARGEST_FLOAT_C(x) x##L
|
||||
#endif
|
||||
|
||||
template <class T>
|
||||
inline BOOST_CONSTEXPR_OR_CONST T make_big_value(largest_float v, const char*, mpl::true_ const&, mpl::false_ const&)
|
||||
{
|
||||
return static_cast<T>(v);
|
||||
}
|
||||
template <class T>
|
||||
inline BOOST_CONSTEXPR_OR_CONST T make_big_value(boost::floatmax_t v, const char*, mpl::true_ const&, mpl::true_ const&)
|
||||
inline BOOST_CONSTEXPR_OR_CONST T make_big_value(largest_float v, const char*, mpl::true_ const&, mpl::true_ const&)
|
||||
{
|
||||
return static_cast<T>(v);
|
||||
}
|
||||
#ifndef BOOST_MATH_NO_LEXICAL_CAST
|
||||
template <class T>
|
||||
inline T make_big_value(boost::floatmax_t, const char* s, mpl::false_ const&, mpl::false_ const&)
|
||||
inline T make_big_value(largest_float, const char* s, mpl::false_ const&, mpl::false_ const&)
|
||||
{
|
||||
return boost::lexical_cast<T>(s);
|
||||
}
|
||||
#endif
|
||||
template <class T>
|
||||
inline BOOST_CONSTEXPR const char* make_big_value(boost::floatmax_t, const char* s, mpl::false_ const&, mpl::true_ const&)
|
||||
inline BOOST_CONSTEXPR const char* make_big_value(largest_float, const char* s, mpl::false_ const&, mpl::true_ const&)
|
||||
{
|
||||
return s;
|
||||
}
|
||||
@@ -46,20 +63,20 @@ inline BOOST_CONSTEXPR const char* make_big_value(boost::floatmax_t, const char*
|
||||
//
|
||||
#define BOOST_MATH_BIG_CONSTANT(T, D, x)\
|
||||
boost::math::tools::make_big_value<T>(\
|
||||
BOOST_FLOATMAX_C(x), \
|
||||
BOOST_MATH_LARGEST_FLOAT_C(x), \
|
||||
BOOST_STRINGIZE(x), \
|
||||
mpl::bool_< (is_convertible<boost::floatmax_t, T>::value) && \
|
||||
((D <= std::numeric_limits<boost::floatmax_t>::digits) \
|
||||
mpl::bool_< (is_convertible<boost::math::tools::largest_float, T>::value) && \
|
||||
((D <= boost::math::tools::numeric_traits<boost::math::tools::largest_float>::digits) \
|
||||
|| is_floating_point<T>::value \
|
||||
|| (std::numeric_limits<T>::is_specialized && \
|
||||
(std::numeric_limits<T>::digits10 <= std::numeric_limits<boost::floatmax_t>::digits10))) >(), \
|
||||
|| (boost::math::tools::numeric_traits<T>::is_specialized && \
|
||||
(boost::math::tools::numeric_traits<T>::digits10 <= boost::math::tools::numeric_traits<boost::math::tools::largest_float>::digits10))) >(), \
|
||||
boost::is_convertible<const char*, T>())
|
||||
//
|
||||
// For constants too huge for any conceivable long double (and which generate compiler errors if we try and declare them as such):
|
||||
//
|
||||
#define BOOST_MATH_HUGE_CONSTANT(T, D, x)\
|
||||
boost::math::tools::make_big_value<T>(0.0L, BOOST_STRINGIZE(x), \
|
||||
mpl::bool_<is_floating_point<T>::value || (std::numeric_limits<T>::is_specialized && std::numeric_limits<T>::max_exponent <= std::numeric_limits<long double>::max_exponent && std::numeric_limits<T>::digits <= std::numeric_limits<long double>::digits)>(), \
|
||||
mpl::bool_<is_floating_point<T>::value || (boost::math::tools::numeric_traits<T>::is_specialized && boost::math::tools::numeric_traits<T>::max_exponent <= boost::math::tools::numeric_traits<boost::math::tools::largest_float>::max_exponent && boost::math::tools::numeric_traits<T>::digits <= boost::math::tools::numeric_traits<boost::math::tools::largest_float>::digits)>(), \
|
||||
boost::is_convertible<const char*, T>())
|
||||
|
||||
}}} // namespaces
|
||||
|
||||
Reference in New Issue
Block a user