Implicit conversion from float -36.63934426229508 to int loses precision