Implicit conversion from float -4.820717131474112 to int loses precision