Implicit conversion from float -9.908759124087595 to int loses precision