摘要
随着对NoC平台研究的逐步深入，如何将规模庞大的应用合理地映射到NoC半台上成为亟待解决的问题之一．本文基于二维网格结构NoC平台，建立了旨在优化系统通信能耗和执行时间的统一目标函数．提出了通过优化链路负载分布间接优化延时的方法，避免了NoC等待延时精确建模的难题．并且采用蚁群算法实现了面向能耗和延时的NoC映射．调整参数λ，可以选择单一目标或者联合目标优化．本文还对映射结果进行了执行时间模拟．实验结果显示：与随机映射相比，单一目标优化在通信能耗和执行时间上分别能节省（30％～47％）和（20％～39％），而联合目标优化则能在能量支配的映射方案中进一步挖掘时间维度的潜力．
With the research on the NoC deeper and deeper, it becomes a stringent task to map a complex application onto an appointed NoC platform efficiently. Based on 2-D mesh NoC platform, this paper sets up a united object function in order to optimize communication energy and execution time. It is also proposed that the delay can be optimized indirectly by means of optimizing distribution of link load, which can avoid the difficulty to model waiting time in NoC accurately. Ant colony algorithm is used to realize the energy-and delay-aware NoC mapping. Single object （energy/delay） or collaborate object can be selected to optimize the mapping by modifying parameter 2. Furthermore, the optimized solutions are simulated on NoC platform finally. Experimental re- suits show single object optimization mapping can reduce communication energy and execution time by （30% -47% ） and （20% 39% ） respectively compared with random mapping. And collaborate object optimization mapping can provide a seductive tradeoff between energy and execution time.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2008年第5期937-942,共6页
Acta Electronica Sinica
基金
国家自然科学基金(No.90307011,No.60576034)
江苏省高技术研究项目(No.BG2005030)
关键词
片上网络
映射
能耗
延时
蚁群算法
network on a chip（NoC）
mapping
energy
delay
ant colony algorithm
