Implicit conversion from float -2.1348314606741496 to int loses precision