A characterization of instruction-level error derating and its implications for error detection