Implicit conversion from float -108.52941176470588 to int loses precision