Implicit conversion from float -1.0631741140215638 to int loses precision