pub const fn at_least_float(d: DType) -> DType
d if it is floating point or complex, otherwise the promotion of d with float32. Transcendental functions and division compute in this type.
d