Implicit conversion from float -20.2540834845735 to int loses precision