Implicit conversion from float -2.1487603305785115 to int loses precision