Instruction Placement for an EDGE Multicore Processor Using Reinforcement Learning

Communication overheads are one of the fundamental challenges in a multiprocessor system. As the number of processors on a chip increases, communication overheads and the distribution of computation and data become increasingly significant. The granularity of communication between processors on a single chip in future systems will determine how significant these overheads will be. Fine-grained communication may expose more parallelism, but at the cost of higher overheads. Coarser-grained communication limits the impact of these overheads, but it may be difficult to expose the same degree of parallelism without the benefits of fine-grained communication. In this work, we attempt to show that the compiler can be effectively used to minimize these fine-grained communication delays in an EDGE processor. We use a reinforcment learning technique called Neuro-Evolution of Augmenting Topologies (NEAT) to aid the compiler in achieving this goal.