Implicit conversion from float -32.14854111405836 to int loses precision