Implicit conversion from float -20.919540229885058 to int loses precision