Implicit conversion from float -78.1640625 to int loses precision