Implicit conversion from float -12.64200792602378 to int loses precision