Implicit conversion from float -32.02247191011236 to int loses precision