Co-Evolutionary Scheduling and Mapping for High-Level Synthesis

The main approach of this paper is utilizing bio-inspired evolutionary strategies for high level synthesis. In this paper, a genetic algorithm (GA) is implemented that schedules a data-flow graph considering design constraints. Also, mapping is performed with another genetic algorithm concurrently, considering resource constraints. The register mapping is performed using a conventional deterministic approach. Then, a co-evolutionary strategy merges two genetic algorithms with the deterministic register allocation algorithm targeting testability improvement. The final result is a high level synthesis tool which synthesizes a data-flow graph using a combination of a co-evolutionary strategy and a deterministic algorithm.