The traditional approach to fault tolerant computing involves replicating computation units and applying a majority vote operation on individual result bits. This approach, however, has several limitations; the most severe is the resource requirement. This paper presents a new method for fault tolerant computing where for a given error rate, the hamming distance between correct inputs and faulty inputs as well as the hamming distance between a correct result and a faulty result is preserved throughout processing thereby enabling correction of up to transient faults per computation cycle. The new method is compared and contrasted with current protection methods and its cost / performance is analyzed.
|State||Published - 2011|