- underflow/overflow handling for type narrowing.
  Eg, double -> float, int64 -> int32