Implicit conversion from float -2.0289855072463894 to int loses precision