Implicit conversion from float -163.125 to int loses precision