Abstract
We present a new technique for partitioning processes in distributed embedded systems. Our heuristic algorithm minimizes both context switch and communication overhead under real-time deadline and process size constraints; it also tries to allocate functions to processors which are well-suited to that function. The algorithm analyzes the sensitivity of the latency of the task graph to changes in vertices hierarchical clustering, splitting and border adjusting. This algorithm can be used for initial partitioning during co-synthesis of distributed embedded systems. Synthesis of examples partitioned by our algorithm with implementations synthesized directly from the original example shows that our partitioning algorithm significantly improves the results obtainable by practical co-synthesis algorithms.

This publication has 15 references indexed in Scilit: