Efficient constrained encoding for VLSI sequential logic synthesis

Abstract

A fast heuristic algorithm called ENCORE is proposed for the dichotomy-based constrained encoding problem. Its implementation has been tested on MCNC synchronous sequential logic benchmarks. For the case of complete encoding, ENCORE generates the same or shorter encoding lengths than the programs KISS, NOVA and DIET, for most of the benchmarks, and uses much less CPU time. For bounded-length encoding, ENCORE produces better overall quality than both random encoding and NOVA. ENCORE has also been applied to state assignment problems for asynchronous machines; it consistently obtains optimal or near-optimal results for a variety of examples found in the literature.<>