Implicit conversion from float -56.400000000000006 to int loses precision