Implicit conversion from float -5.31835205992509 to int loses precision