Implicit conversion from float -8.22646657571623 to int loses precision