Implicit conversion from float -9.636871508379897 to int loses precision