一种基于气泡流控的改进多播路由算法
发布时间:2018-10-30 12:23
【摘要】:多播通信在片上多核系统中占据很重要的地位,并会对多核系统上并行应用程序的性能产生很大影响。现有的多播路由算法大多存在资源利用不均衡的不足,从而导致片上缓存资源的利用率较低。提出一种新的基于气泡流控的多播路由算法,该方法充分利用了片上网络的缓存资源,并通过向网络注入气泡避免了死锁现象的产生。实验表明,该方法可以有效提高片上网络中多播通信的性能,在合成负载下,该方法相对于均衡自适应多播路由算法可以实现18.1%的网络平均时延的降低以及16.7%的网络饱和吞吐量的增加。
[Abstract]:Multicast communication plays an important role in on-chip multicore systems and will have a great impact on the performance of parallel applications on multi-core systems. Most of the existing multicast routing algorithms have the deficiency of unbalanced utilization of resources, which leads to low utilization of on-chip cache resources. A new multicast routing algorithm based on bubble flow control is proposed, which makes full use of the buffer resources of the on-chip network and avoids the deadlock by injecting bubbles into the network. Experiments show that this method can effectively improve the performance of multicast communication in the on-chip network. Compared with the equalization adaptive multicast routing algorithm, this method can reduce the average delay by 18.1% and increase the saturated throughput by 16.7%.
【作者单位】: 国防科学技术大学计算机学院;
【基金】:国家863计划资助项目(2013AA014301,2013AA01A208) 全国优秀博士学位论文作者专项资金资助项目(201450) 国家自然科学基金资助项目(61272482,61303066)
【分类号】:TN47;TP332
[Abstract]:Multicast communication plays an important role in on-chip multicore systems and will have a great impact on the performance of parallel applications on multi-core systems. Most of the existing multicast routing algorithms have the deficiency of unbalanced utilization of resources, which leads to low utilization of on-chip cache resources. A new multicast routing algorithm based on bubble flow control is proposed, which makes full use of the buffer resources of the on-chip network and avoids the deadlock by injecting bubbles into the network. Experiments show that this method can effectively improve the performance of multicast communication in the on-chip network. Compared with the equalization adaptive multicast routing algorithm, this method can reduce the average delay by 18.1% and increase the saturated throughput by 16.7%.
【作者单位】: 国防科学技术大学计算机学院;
【基金】:国家863计划资助项目(2013AA014301,2013AA01A208) 全国优秀博士学位论文作者专项资金资助项目(201450) 国家自然科学基金资助项目(61272482,61303066)
【分类号】:TN47;TP332
【相似文献】
相关期刊论文 前10条
1 程文斐;谭小彬;徐鹏;;基于贪婪策略的高效可靠多播路由协议[J];计算机工程;2012年21期
2 范俊峰,封国剑;一种基于新型优化算法的多播路由选择方法[J];计算机工程与应用;2005年31期
3 陈琳,杨志云,徐正全;基于遗传操作的带度约束的多播路由算法[J];计算机工程;2005年02期
4 乔保军;石峰;计卫星;;基三分层网络中的受限多播路由算法[J];计算机应用;2007年04期
5 吴家皋;;覆盖多播路由的算法及协议研究综述[J];计算机科学;2007年06期
6 蔡奎生;;基于量子遗传退火策略的多播路由优化模型及其算法[J];计算机工程与科学;2009年10期
7 陈琳,余健,杨志云,徐正全;一种延迟约束的动态多播路由算法[J];江汉石油学院学报;2003年S1期
8 陈琳,杨志云,徐正全;基于遗传操作的延迟和度约束的多播路由启发式算法[J];武汉大学学报(理学版);2004年03期
9 刘芳,杨海潮;一种基于克隆策略的多播路由算法[J];电子与信息学报;2004年11期
10 胡光岷;CHANG Rocky;;可扩展的多目标最优化多播路由(英文)[J];软件学报;2008年06期
相关会议论文 前6条
1 高玲玲;李伟生;;一种新的动态QoS多播路由启发式算法[A];2006年全国通信软件学术会议论文集[C];2006年
2 王琳珠;单_,
本文编号:2300003
本文链接:https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2300003.html