Implicit conversion from float -1.290322580645153 to int loses precision