Implicit conversion from float -32.396449704142015 to int loses precision