Implicit conversion from float -355.54455445544556 to int loses precision