Implicit conversion from float -36.21722846441948 to int loses precision