摘要
现在FPGA已被广泛的应用,但三模冗余(TMR)结构不能充分利用FPGA资源,基于主/从版本技术,提出一种实时任务容错调度算法。算法通过后向调度从版本,使得从版本在其截止期限内尽可能地推迟执行,从而同一任务的主从版本在执行时间上没有重叠或较少重叠,当任务主版本成功执行时释放任务从版本所占用的资源。仿真实验表明,与TMR相比,此方法能更有效地利用FPGA资源,提高硬件任务的接受率。
FPGA has been used widely now, but the architecture of TMR( triple module redundancy) can not make full use of FPGA. This paper proposed a fault-tolerant scheduling algorithm of hardware real-time tasks based on primary/slave copy. By scheduled backwards, the salve copy executed as late as possible before its deadline. So that the primary copy had no overlap or as small overlap as possible with the slave copy on the executing time. When the primary copy executed successfully, released the slave copy. The result of experiment demonstrates that the method can use the FPGA more efficiently than TMR and the accept ratio of hardware tasks increases.
出处
《计算机应用研究》
CSCD
北大核心
2009年第5期1729-1732,共4页
Application Research of Computers
关键词
可重构计算
实时调度
容错调度
放置策略
reconfigurable computing
real-time scheduling
fault-tolerant scheduling
placement scheme