Evolutionary Acyclic Graph Partitioning
This addresses the challenge of efficiently mapping streaming applications to resource-constrained embedded multiprocessors, representing an incremental improvement in domain-specific optimization.
The paper tackles the problem of parallelizing streaming applications on embedded multiprocessors by developing a multi-level algorithm for acyclic graph partitioning, and engineers an evolutionary algorithm to reduce communication cost, improve load balancing, and decrease scheduling makespan.
Directed graphs are widely used to model data flow and execution dependencies in streaming applications. This enables the utilization of graph partitioning algorithms for the problem of parallelizing computation for multiprocessor architectures. However due to resource restrictions, an acyclicity constraint on the partition is necessary when mapping streaming applications to an embedded multiprocessor. Here, we contribute a multi-level algorithm for the acyclic graph partitioning problem. Based on this, we engineer an evolutionary algorithm to further reduce communication cost, as well as to improve load balancing and the scheduling makespan on embedded multiprocessor architectures.