Implicit conversion from float -1.9718309859155028 to int loses precision