Implicit conversion from float -3.095599393019725 to int loses precision