Implicit conversion from float -15.125 to int loses precision