This paper presents an efficient method of handling a large number of paths during path-delay fault testing. Instead of handling the corresponding set of signals, an identifier is derived for every path. We handle up to three billion paths because the memory requirement is only about three bits per path. Compared to former approaches, experimental results show fast access, small memory requirements, and negligible CPU-times for the management of huge path sets
Date of Conference: 28 Feb-3 Mar 1994