Implicit conversion from float -7.215189873417728 to int loses precision