Improving system level design space exploration by incorporating SAT-solvers into multi-objective evolutionary algorithms

Automatic design space exploration at the system level is the task of finding optimal or close to optimal mappings for a set of applications onto an optimized architecture. Especially, finding a feasible binding of processes onto resources that permit the communications imposed by data dependencies is known to be a NP-complete task which demands the use of heuristic optimization approaches. Nearly all optimization approaches known from literature will fail in design spaces containing only a few feasible solutions. In this paper, we propose a novel approach based on the combination of multi-objective evolutionary algorithms and SAT-solvers to overcome these drawbacks. We provide experimental results showing the efficiency of our novel methodology for synthetic and real life test cases.