Implicit conversion from float -5.350318471337573 to int loses precision