Implicit conversion from float -5.2554744525547505 to int loses precision