Implicit conversion from float -20.828025477707 to int loses precision