Implicit conversion from float -3.063583815028906 to int loses precision