Massively Parallel Search for Transition-Tables of Polyautomata

One of the fundamental tasks in automata theory is to look for transition-tables that implement a given speciication. In principle, most of this task can be performed by a computer. But a combinatorial explosion in the number of possible transition-tables quickly renders brute force search impractical. This paper demonstrates two approaches to extend the frontier of tractable problem sizes. Firstly, an eecient heuristic technique is used which dramatically prunes the search space without giving up completeness. Secondly, a massively parallel implementation is described which achieves near linear speedup on as many as 16384 Processors. These techniques yield some new results regarding two open problems involving cellular automata and trellis automata.