Implicit conversion from float -80.9375 to int loses precision