Yes.
(and if we talk about optimizing compilation things are getting more complicated as there int31, int32 and float64 can all coexist and you can have 31-bit integer stored in a float64 value - if operation was specialized for floating point values)