Seemingly it's still faster to let mirc round $pi than to call a custom identifier.