Implicit conversion from float -65.55944055944056 to int loses precision