Implicit conversion from float -21.2271540469974 to int loses precision