论文部分内容阅读
在多簇处理器情况下,指令应用所带来的簇间数据交互问题已经成为制约处理器性能的关键问题。针对此问题提出了在一般的调度后进行一次后溯重调度优化过程,减少了簇间的数据交互量,提高了编译器关于处理器的利用率,同时减少了编译生成的指令序列运行时所消耗的功耗。实验结果表明,利用该方法进行调度,比列表调度算法簇间数据交互量减少平均44.36%,调度后的指令执行时间的平均减少量为24.93%,比UAS(unified assign and schedule)调度算法簇间数据交互量减少平均31.25%,调度后的指令执行时间的平均减少量为14.62%。
In the case of multi-cluster processors, the problem of inter-cluster data exchange brought by the instruction application has become a key problem that restricts the performance of the processor. In order to solve this problem, we proposed a post-traverse scheduling optimization process after the general scheduling to reduce the amount of data interaction between clusters and improve the compiler’s utilization rate of the processor, while reducing the compiler-generated instruction sequence runtime Consumed power consumption. The experimental results show that the proposed algorithm can reduce the amount of data interaction between clusters by an average of 44.36% and the average reduction of scheduled execution time by 24.93% The average amount of data exchanged is reduced by 31.25% and the average reduction of the scheduled execution time is 14.62%.