Hierarchical Static Scheduling of Dataflow Graphs onto Multiple Processors

5. Conclusions

In this paper, we have introduced a hierarchical scheduling framework for SDF graphs being mapped onto multiple processors. This framework can drastically reduce the number of nodes present in the final DAG graph, which is used for parallel scheduling. We have shown a practical example where this scheduling technique has greatly improved the final schedule.

We plan to augment the SDF graph clustering techniques by specializing some of the DAG clustering heuristics found in multiprocessor schedulers for direct use on the SDF graph. The objective is to hide only that parallelism that would not be exploited anyway, and in doing so, simplifying the DAG.

Hierarchical Static Scheduling of Dataflow Graphs onto Multiple Processors

[Next] [Previous] [Top]