Implicit conversion from float -1.3881748071979416 to int loses precision