Implicit conversion from float -11.125 to int loses precision