不跨层行操作并行RCA互连时延性能评估
Interconnect Delay Performance Evaluation for Non-Crossing Level and Row Operands Parallel RCA

针对三类行并行重构单元阵列互连时延性能评估问题, 提出了一种通过节点映射和运行机制来评测互连时延的方法, 基于前驱回溯不加旁节点不跨层时域映射算法, 对点到点、路由传输、行列总线等互连RCA进行了时延分析和计算．实验结果表明, 与路由传输和行列总线互连相比, 点到点互连在最大不跨层互连、不跨层累加互连、考虑互连执行总时延等方面均是最小的, 从而表明了点到点重构单元阵列的互连时延优于路由器传输和行列总线互连．
In order to assess three types of row parallel recon?gurable cell array(RCA)interconnect delay performance，this paper presented a method for evaluating interconnect delay by means of node mapping and operation running mechanism. Based on preorder traversing backtracking no adding-bypass-node(PTBNA)non-crossing level temporal mapping algorithm，this paper analyzed and computed RCA interconnect delay of point to point(PP)，router transmission(RT)，and row column bus(RCB). Compared with RT and RCB，PP can get the least in maximum non-crossing level interconnect delay，non-crossing level accumulation interconnect delay and considering interconnect execution total delay. Thus PP-RCA interconnect delay is better than that of RT and RCB