Implicit conversion from float -2.324246771879487 to int loses precision