Buffer optimal static scheduling with a throughput constraint for synchronous dataflow applications on multiprocessors

Abstract

This paper determines a static scheduling and the minimal size of arc buffers for a given synchronous dataflow (SDF) graph, satisfying a throughput constraint. Unlike the previous work, we assume that the target architecture and the mapping information are given. In addition we consider the unfolding of the SDF graph to improve the throughput. To solve this problem, we adopt answer set programming (ASP) with constraint programming (CP) technique which always finds optimal solutions. The proposed ASP+CP formulation is compact enough to list the scheduling rules in 27 lines and could be applied to a small-but-practical size of coarse-grain SDF graphs successfully.