Genetic Scheduling Algorithm for High-Level Synthesis

High-level synthesis (HLS) aims at constructing the optimal hardware or software structure from a given high-level speci cation. This process involves a number of optimization steps, from which scheduling is the most crucial one, concerning both the running time of the process and the quality of the found solution. In this paper, we present a genetic algorithm for the scheduling problem in HLS. We describe in detail how this general heuristic could be applied to the particular engineering problem at hand. It is demonstrated with several measurements on industry benchmarks that the new algorithm performs better than previous methods.