Implicit conversion from float -3.1451612903225623 to int loses precision