Implicit conversion from float -91.16343490304709 to int loses precision