Implicit conversion from float -8.04878048780489 to int loses precision