摘要
PADDING是一种提高程序中循环Cache命中率的有效方法,其有效性的关键在于选取合适的PADDING大小以尽可能地消除循环中数组引用之间的Cache冲突。本文从CME理论出发,根据循环中数组引用重用向量的不同,建立对应的CME模型并对其进行分析,给出两种不同的计算PADDING大小的算法。我们运用这两种算法对BenchmarkSPEC95中部分测试程序进行Cache优化试验,取得了较好的效果。
PADDING is a kind of efficient method to improve the cache hit rate of loops in the program.The key of its efficiency is to select proper PADDING value to reduce the cache conflict between array references to the utmost.This paper, based on the theory of CME, establishes the CME model,analyses it according to the different reuse vector of array reference in the program,and gives two algorithms of caculating PADDING value.The authors use these algorithms in the experiment to improve the cache hit rate of some testing program in Benchmark SPEC95 and gain some good results.
出处
《信息工程大学学报》
2002年第4期35-39,共5页
Journal of Information Engineering University