num_traits::pow::float_impls