高级检索
王海琪, 董社勤. 面向应用的片上网络的网络拓扑生成算法[J]. 计算机辅助设计与图形学学报, 2011, 23(9): 1576-1584.
引用本文: 王海琪, 董社勤. 面向应用的片上网络的网络拓扑生成算法[J]. 计算机辅助设计与图形学学报, 2011, 23(9): 1576-1584.
Wang Haiqi, Dong Sheqin. Topology Generation Algorithm for Application-Specific Network-on-Chip[J]. Journal of Computer-Aided Design & Computer Graphics, 2011, 23(9): 1576-1584.
Citation: Wang Haiqi, Dong Sheqin. Topology Generation Algorithm for Application-Specific Network-on-Chip[J]. Journal of Computer-Aided Design & Computer Graphics, 2011, 23(9): 1576-1584.

面向应用的片上网络的网络拓扑生成算法

Topology Generation Algorithm for Application-Specific Network-on-Chip

  • 摘要: 针对面向应用的片上网络,提出了一种三阶段的低功耗网络拓扑生成算法.首先基于内核通信量和物理坐标信息做划分驱动的布图规划,以确定内核的摆放位置以及内核和转换器之间的映射关系;其次考虑转换器和网络接口的面积消耗,并把它们的同时插入问题抽象成整数线性规划模型,通过求解此优化模型确定其最佳插入位置,生成互连网络;最后通过路由分配策略确定互连网络上的通信量分布,进一步优化功耗.实验结果表明,该算法平均能节省35.2%的功耗开销以及5.7%的中转转换器数目.

     

    Abstract: A three-phase low-power topology generation algorithm for application-specific network-on-chip is proposed in this paper.Firstly, partition-driven floorplan is done to determine the cores' positions and the mapping relations between cores and switches.Then, with the area and power consuming of network components (such as switches and network interfaces) taken into consideration, the simultaneous insertion of network components is formulated as an integer linear programming (ILP) , which is solved to determine the optimal positions for switches and network interfaces.Finally, path allocation is carried out to distribute the communications among switches.The experimental results show that our algorithm can save power by 35.2% and reduce the number of hops by 5.7%.

     

/

返回文章
返回