AN EFFICIENT HEURISTIC METHOD FOR STATE ASSIGNMENT OF LARGE SEQUENTIAL MACHINES

Abstract

VLSI circuit design is a “trial and error” process that consists of solving a number of design problems. An optimal state assignment is one of the most important problems in the logic synthesis for sequential machines and it consists of choosing a binary representation for symbolic internal states of a sequential machine, so that the resulting logic is optimal for a given objective. This problem belongs to the class of most complex computational problems in VLSI design – it is NP hard. In a strict sense, it has never been solved, except for exhaustive search, which is impossible for large machines even with a computer. A structural heuristic approach uses specific knowledge about the structure of a given problem in order to reduce the search space to a manageable size and to maintain high quality solutions at the same time. Using the state assignment problem as an example, we determined the importance of the structural heuristic approach in CAD for VLSI and we showed how to search for suitable heuristics. We discussed a new heuristic method for state assignments concentrating on heuristic aspects such as: the solution space, the generation procedure and its operators, the evaluation functions etc. We provided some experimental results to show that the new method is very efficient. The structural heuristic search can be highly efficient. Its efficiency is limited more by the capacity of the human brain to think heuristically than by the complexity of the problem itself.