一种自适应的混合型无线NoC拓扑结构

发布时间：2018-04-30 01:00

本文选题：无线片上网络 + 自适应拓扑　；参考：《电子学报》2016年06期

【摘要】：针对传统大规模片上网络(Network-on-Chip,No C)远距离核间多跳通信所带来的高能耗与延时问题,提出了一种基于虚Torus的自适应的混合型无线No C拓扑结构(VT-AWi No C).该结构通过引入链路拥塞测度作为感知参数,基于此采用热点无线链路自动探测与带宽动态分配机制,并设计实现发送器动态分配的控制电路模块,以达到根据不同的通信流量模型,于片内自适应地调整拓扑结构及链路带宽的目的.通过建立混合型无线No C的延时与功耗评估模型,对该结构的无线No C进行性能评估.实验结果表明,该自适应拓扑与其它混合型无线No C相比,在随机流量模型下,网络平均延时降低了16.52~23.27%;在20%的热点流量模型下,包平均能耗节省了39.19%;以真实应用FFT作为基准测试,平均延时降低了17.20%~21.68%,并节省了23.49%的包平均能耗.该结构以较小的面积开销获得了更优的性能.
[Abstract]:Aiming at the problem of high energy consumption and delay caused by the traditional large-scale on-chip network (Network-on-ChipNo No) communication between long distance cores, an adaptive hybrid wireless No C topology based on virtual Torus is proposed. This structure adopts the link congestion measure as the sensing parameter, adopts the hot wireless link automatic detection and the bandwidth dynamic allocation mechanism based on this, and designs the control circuit module of the transmitter dynamic allocation. In order to adjust the topology and link bandwidth adaptively according to different traffic models. The performance of the hybrid wireless No C is evaluated by establishing a delay and power evaluation model of the hybrid wireless No C. The experimental results show that compared with other hybrid wireless No C, the network average delay is reduced by 16.52% and 23.27 in the random traffic model, and 20% in the hot traffic model. The average energy consumption of the package was saved by 39.19, and the average delay was reduced by 17.20% and 23.49% by using FFT as the benchmark. The structure achieves better performance with smaller area overhead.
【作者单位】：南京航空航天大学电子信息工程学院;台州学院计算机应用研究所;
【基金】：国家自然科学基金(No.61376025) 江苏省产学研前瞻性联合研究项目(No.BY2013003-11) 浙江省自然科学基金(No.LY13F020012)
【分类号】：TN47

【相似文献】