Implicit conversion from float -5.497630331753555 to int loses precision