On supported architectures, change the rounding mode of the floating point arithmetics (to nearest, up, down, to zero) or change the precision of the arithmetical operations (single, double, double extended). Experimentally test the properties of the floating point arithmetics.