A combined interval and floating-point comparator

Interval arithmetic provides an efficient method for monitoring and controlling errors in numerical calculations. However, existing software packages for interval arithmetic are often too slow for numerically intensive calculations. This paper presents the design of a comparator that performs either interval or two's complement/floating-point comparisons. This comparator requires only slightly more area than a conventional floating point comparator, and provides a significant performance improvement over software implementations of interval comparisons. Both serial and parallel designs are examined and the results are implemented in AMI 0.60 /spl mu/m technology.