...
首页> 外文期刊>Journal of Parallel and Distributed Computing >Design and evaluation of ZMesh topology for on-chip interconnection networks
【24h】

Design and evaluation of ZMesh topology for on-chip interconnection networks

机译:片上互连网络的ZMesh拓扑的设计和评估

获取原文
获取原文并翻译 | 示例

摘要

AbstractThis article presents the design and evaluation of a scalable and energy efficient Network-on-Chip topology with diagonal links, called ZMesh. A heuristic technique for mapping applications onto the ZMesh topology has been proposed. Further, a mapping technique based on particle swarm optimization combined with simulated annealing has also been considered for a fair comparison of the performance of ZMesh with that of other topologies. Experimental evaluation shows that ZMesh when compared with baseline Mesh, shows an average improvement of up to 1.25×and 1.20×in energy consumption and execution time, respectively while running real application traces. Compared to the state-of-the-art topologies such as DMesh, ZMesh shows an average improvement of up to 2.39×in terms of energy consumption, while suffering a degradation of up to 9.37% in execution time, for running real application traces. For the synthetic as well as real applications with more uniform traffic pattern, such as constant geometry algorithms, ZMesh achieves better energy efficiency than the other considered topologies. Evaluation of ZMesh for multicast kind of traffic using a proposed multicast routing algorithm shows improvement in link energy consumption by up to 55%, for the considered scenarios, as compared to unicast-based multicast routing.HighlightsA novel NoC topology with diagonal links, called ZMesh, has been proposed.Analysis of ZMesh topology in terms of various parameters has been performed.A novel heuristic to map applications onto ZMesh has been proposed.Performance evaluation of ZMesh for various traffic scenarios has been presented.Brief analysis on ZMesh support for multicast traffic has been presented.
机译: 摘要 本文介绍了可扩展且节能的网络-具有对角链接的片上拓扑,称为ZMesh。已经提出了一种用于将应用程序映射到ZMesh拓扑的启发式技术。此外,还考虑了基于粒子群优化与模拟退火相结合的映射技术,以公平地比较ZMesh与其他拓扑的性能。实验评估表明,与基线Mesh相比,ZMesh的平均改进幅度高达1.25 × 和1.20 × 分别在运行实际应用程序跟踪时的能耗和执行时间。与DMesh等最新拓扑相比,ZMesh的平均改进幅度高达2.39 × 就能耗而言,尽管运行时间下降了9.37%,但仍可运行实际的应用程序跟踪。对于具有更统一流量模式的合成应用程序和实际应用程序(例如恒定几何算法),ZMesh的能源效率均高于其他考虑的拓扑。与基于单播的多播路由相比,使用所建议的多播路由算法对ZMesh进行多播路由算法的评估显示,在考虑的情况下,链路能耗最多可降低55%。 Highlights < / ce:section-title> < ce:list id = “ d1e2400 ”> 已提出了一种新颖的带有对角链接的NoC拓扑,称为ZMesh。 < ce:label>• 已根据各种参数对ZMesh拓扑进行了分析。 < / ce:列表项> 一种新颖的启发式方法将应用程序映射到ZMesh 已针对各种流量场景对ZMesh进行了性能评估。 已经对ZMesh对多播流量的支持进行了简要分析。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号