Implicit conversion from float -2.9745042492917833 to int loses precision