Implicit conversion from float -9.755859375 to int loses precision