Most efficient floating-point types at least as wide as float and double respectively

Evaluates to positive infinity or the value guaranteed to overflow a float

Evaluates to a quiet NaN of type float