Implicit conversion from float -3.583617747440286 to int loses precision