Consider enabling Boost.Multiprecision. This may mess up the use of concepts, such as std::floating_point