Implicit conversion from float -70.44568245125348 to int loses precision