Implicit conversion from float -32.06106870229007 to int loses precision