摘要
在片上网络(NoC)的网络分配与任务映射相配合的路径分配中,单维序路由策略会限制可行解空间。为此,提出一种基于双维序路由策略的网络分配方法。在路径分配步骤中采用双维序路由法,设计以带宽、延时和无死锁为约束条件、以降低动态及静态能耗为优化目标的遗传算法。实验结果表明,该方法可以扩大任务映射的可行解空间,求解最小所需带宽比单维序法平均减少6.3%,且在各种带宽场合时均能求得更低能耗解。
As single dimension-ordered routing used in path allocation limits the solution space of mapping problem for Network-on-Chip（NoC）, a network assignment approach based on dual dimension-ordered routing is proposed. Dual dimension-ordered routing method is adopted in the path allocation step, and a Genetic Algorithrn（GA） is developed to minimize dynamic and static energy consumption subject to bandwidth, delay and lock-free constraints. Experimental results show that the proposed algorithm is able to enlarge solution space, reduce link bandwidth requirements by 6.3% in average, and fred less energy consumption solutions at any bandwidth circumstance.
出处
《计算机工程》
CAS
CSCD
2012年第13期13-16,21,共5页
Computer Engineering
基金
核高基重大专项"自主知识产权高性能嵌入式CPU的研发及产业化"(2009ZX01030-001-002)
关键词
片上网络
网络分配
低能耗
双维序路由
遗传算法
多约束优化
Network-on-Chip（NoC）
network assignment
low energy
dual dimension-ordered routing
Genetic Algorithm（GA）
multi-constrain optimization
