基于调用链分析的访存优化技术
Memory Operation Optimization Based on Call-graph Analysis
-
摘要: 本文提出一个基于调用链分析的访存优化编译技术.该技术通过建立一个扩展的调用关系图,分析程序从而估算出最小的寄存器保存集合,达到提高寄存器的利用率,并减少冗余访存操作的目的.本文将该算法在IM-PACT1编译框架中实现,并对DSPstone测试集进行实验.实验数据表明,该算法能够平均减少40.36%的访存操作,并获得平均2.99%的性能提升.Abstract: In this paper, an optimal compiler algorithm is proposed for better utilization of registers and elimination of redundant memory operation.By building an extended call graph, the algorithm is able to saving least context during function call.The experimental results show a decrease in memory accesses by an average of 40.36% and an improvement in performance by an average of 2.99%.applications.
下载:
