The error introduced when a real number is rounded to a machine-representable number. Most floating-point calculations incur roundoff error. For any one floating-point operation, IEEE Std 754 specifies that the result shall not incur more than one rounding error.