Implicit conversion from float -9.920127795527165 to int loses precision