NoC处理单元随机舍入的启发式应用映射

摘要

With the performance constraints, a unified approach is presented to reduce energy and solve the application-mapping problem on a heterogeneous network-on-chip (NoC) platform, which incorporates heterogeneous intellectual properties (Ips), is interconnected in a hierarchical star topology and operating at multiple voltage levels. Mixed Integer Linear Program is then used to formulate the problem. Finally, a randomized rounding based heuristic algorithm (HAMU) is proposed for energy efficient application mapping on NoC processing elements. Experimental results demonstrated that the HAMU approach is more efficient for application-mapping comparing with other heuristic algorithms.%在性能约束的前提下,为了降低能量消耗,针对一个集成了异构IP块以分层星型拓扑互连的片上网络(NoC),采用多电压电平操作,运用一种统一方法来解决应用映射问题,并用混合整数线性程序公式化问题,提出了基于随机舍入的NoC处理单元启发式高效应用映射--HAMU法.实验结果表明,HAMU法的效能明显优于其他启发式方法.