Implicit conversion from float -12.536715620827778 to int loses precision