论文部分内容阅读
为了满足片上网络虚拟化过程中产生的不规则拓扑子网内部的通信需要,提出了一种二级优先维序路由算法。该算法通过2 b的配置信息,决定在不同拓扑结构中每个路由器的两个优先路由方向,使其能快速有效地在以下两类拓扑中路由:3个Mesh以任意方式连接成的拓扑结构,或者是任何满足最小路径原则的拓扑结构。该文分析了该算法的路由连通性和死锁避免问题,并在FPGA(field-pro-grammable gate arrays)器件EP2S180F1020C3上的进行了综合仿真。仿真结果表明:跟传统的维序路由算法相比,该算法路由器增加面积只有1.2%,即以几乎可以忽略的代价有效地实现了两类拓扑集的路由。
In order to meet the need of communication within the irregular topological sub-network generated during virtualization of on-chip network, a two-level priority dimension routing algorithm is proposed. Through the configuration information of 2 b, the algorithm determines the two preferential routing directions of each router in different topologies so that it can be quickly and efficiently routed in two types of topologies: three topologies in which Mesh is connected in an arbitrary manner , Or any topology that meets the minimum path principle. This paper analyzes the problem of routing connectivity and deadlock avoidance of the algorithm, and conducts comprehensive simulation on FPGA (field-pro-grammable gate arrays) EP2S180F1020C3. The simulation results show that compared with the traditional routing algorithm, the proposed router increases the area by only 1.2%, that is, the routing of the two types of topologies can be effectively implemented with almost negligible cost.