X-Network: An Area-Efficient and High-Performance On-Chip Wormhole-Switching Network
X-Network: An Area-Efficient and High-Performance On-Chip Wormhole-Switching Network
复制标题
X-Network：面积高效且高性能的片上虫洞交换网络
DOI：
10.1109/hpcc.2010.110
复制
发表时间：
2010
期刊：
影响因子：
--
通讯作者：
L. Bandi
中科院分区：
文献类型：
--
作者：
Xiaofang Wang;L. Bandi
分组交换片上网络（NoCs）已经成为一个有前途的范例，为未来的芯片多处理器和复杂的片上系统（SoC）设计可扩展的通信基础设施。然而，对高性能网络的追求导致了非常消耗面积和复杂的路由器，而性能回报很少。本文介绍了X网络，一个低面积和高性能的虫洞交换NoC，是建立在一个新的PE（处理单元）路由器组织。在X-Network中，每个路由器由四个PE共享，并且除了相邻PE之间的NEWS（北、东、西、南）连接之外，每个通用PE还可以访问四个直连路由器。通过在PE之间共享路由器，网络减少了分组的平均跳数，从而减少了延迟并提高了网络的吞吐量。我们的设计不仅减少了路由器的总数量为给定数量的PE，但也提供了更多的路由灵活性相比，现有的基于网格的解决方案。任何路由算法都可以在X-Network中使用，只要加入我们的路由器选择功能。使用合成工作负载和SPLASH-2应用程序的广泛模拟结果表明，对于具有64个PE的系统，X-Network可以将平均网络延迟降低51%。使用完全自适应的路由算法，网络饱和点扩展到约100%。
Packet-switching networks on chip (NoCs) have emerged as a promising paradigm for designing scalable communication infrastructures for future chip multiprocessors and complex Systems on Chip (SoCs). However, the quest for highperformance networks has led to very area-consuming and complicated routers with little return in performance. This paper presents X-Network, a low-area and high-performance wormhole-switching NoC that is built on a novel PE (Processing Element)-router organization. In X-Network, each router is shared by four PEs and each general PE has access to four directly-connected routers in addition to NEWS (North, East, West, South) connections between neighboring PEs. By sharing routers among PEs, the network reduces the average hop count for packets thereby reducing the latency and improving the throughput of the network. Our design not only reduces the total number of routers for a given number of PEs, but also offers much more routing flexibility compared to existing mesh-based solutions. Any routing algorithm can be used in X-Network after incorporating our router selection functions. Extensive simulation results using both synthetic workloads and the SPLASH-2 applications show that X-Network can reduce average network latency by 51% for a system with 64 PEs. The network saturation point is extended by up to approximately 100% using a fully-adaptive routing algorithm.