A parallel logic programming approach to combinatorial optimization in design

An approach to the solution of combinatorial optimization problems using parallel logic programming is presented. The solution space is divided with a separate process (worker process) assigned to search each subspace. In order to have information regarding the most recent optimal combination available to all worker processes, a tree of manager processes is established to relay discoveries at a given worker process to other worker processes. The performance of this method is tested through the application to a structural design problem. The execution time was measured while varying the number of processors and super-linear speed-up was attained.