Implicit conversion from float -21.24818577648766 to int loses precision