Mesh-of-Trees and Alternative Interconnection Networks for Single Chip Parallel Processing (Extended Abstract)

Abstract

Many applications have stimulated the recent surge of interest single-chip parallel processing. In such machines, it is crucial to implement a high-throughput low-latency interconnection network to connect the on-chip components, especially the processing units and the memory units. In this paper, we propose a new mesh of trees (MoT) implementation of the interconnection network and evaluate it relative to metrics such as wire area, register count, total switch delay, maximum throughput, latency-throughput relation and delay effects of long wires. We show that on-chip interconnection networks can facilitate higher bandwidth between processors and shared first-level cache than previously considered possible. This has significant impact for chip multiprocessing. MoT is also compared, both analytically and experimentally, to some other traditional network topologies, such as hypercube, butterfly, fat trees and butterfly fat trees. When we evaluate a 64-terminal MoT network at 65nm technology, concrete results show that MoT provides higher throughput and lower latency especially when the input traffic (or the on-chip parallelism) is high, at the cost of larger area. A recurring problem in networking and communication is that of achieving good sustained throughput in contrast to just high theoretical peak performance that does not materialize for typical work loads. Our quantitative results demonstrate a clear advantage of the proposed MoT network in the context of single-chip parallel processing.