异构信号处理平台中层次性流水线调度算法  

Multi-level Pipeline Scheduling Algorithm in Heterogeneous Signal Processing Platform

作　　者：杨平平 岳春生[1] 胡泽明[1] YANG Pingping;YUE Chunsheng;HU Zeming(College of Information System Engineering,Information Engineering University,Zhengzhou 450001,China)

机构地区：[1]信息工程大学信息系统工程学院,郑州450001

出　　处：《计算机工程》2018年第11期83-89,共7页Computer Engineering

基　　金：国家科技支撑计划项目(2013BAH56F00)

摘　　要：现有实时任务调度算法在系统异构性方面没有考虑处理节点计算能力的差异,导致任务划分不均衡。为此,根据异构信号处理平台实时任务的特点,提出一种面向异构信号处理平台的层次性流水线调度算法。基于数据流图的多层次任务划分算法并借鉴多层图划分思想,实现任务调度负载均衡和低通信同步开销。采用同步流水线调度方法达到低延迟实时数据处理的目的。实验结果表明,与事件触发数据流调度算法相比,该算法能有效提高异构信号处理平台的实时处理能力。The existing real-time task scheduling algorithm do not consider the difference of node computing power in system heterogeneity,which leads to the imbalance of task partition.Therefore,according to the characteristics of real-time task in the heterogeneous signal processing platform,a multi-level pipeline scheduling algorithm for heterogeneous signal processing platform is proposed.The multi-level task partition algorithm based on data flow graph and the idea of multi-layer graph partition are used to realize the task scheduling load balancing and low communication synchronization overhead.The synchronous pipeline scheduling method is used to achieve the purpose of low delay real-time data processing.Experimental results show that,compared with the event-triggered data flow scheduling algorithm,this algorithm can effectively improve the real-time processing capability of heterogeneous signal processing platform.

关 键 词：异构信号处理平台 有向无环图 负载均衡 任务调度 同步流水线 

分 类 号：TN911.7[电子电信—通信与信息系统]