依赖距离主导的向量化方法研究  

Vectorization method research dominated by dependence distance

作　　者：丁丽丽 韩林 王冬 张素平 王鹏翔 于海宁 

机构地区：[1]信息工程大学数学工程与先进计算国家重点实验室,郑州450001

出　　处：《计算机应用研究》2017年第5期1311-1315,共5页Application Research of Computers

基　　金："核高基"国家科技重大专项资助项目(2009ZX01036-001-001-2)

摘　　要：向量寄存器的非满载使用方式为大量迭代次数不足的循环提供了向量化的机会,但也导致向量化的并行宽度不固定,传统的向量因子主导的依赖测试方法不再适用。提出了一种依赖距离主导的依赖测试方法,通过分析依赖图中所有依赖环的破环关键边所携带的依赖距离,选择其中最小的依赖距离来决定并行宽度,破除依赖环,实现基于向量寄存器非满载使用方式的向量化。实验结果表明,该方法能够有效增加循环向量化的机会和提高向量寄存器的使用率,测试用例的向量化加速比平均提高14.6%。Though the usage of the non- full loaded method of the vector register provided an opportunity number of loops which were lacking in iteration numbers , however , it also caused problem as the instability of inter-iterationparallelism of the vector, which made the tra ditional dependent test methods dominated by vector factor no longer applicabIn order to solve this problem,this paper pesented a dependent test method dominated by dependence distance. During itsprocedure, this method chose the smallest dependence distance by analyzing the dependence distance carriewhich could break dependence circle in the dependency graph,thus determined the inte r-iteration pdependence circle, in order to achieve the vectorization based on the usage of the non-full loaded method of theExperimental results demonstrate that this method can effectively promote the opportunities of vectoring loopsvantage of non-full length usage of vector register, it improves the performance of programs 14. 6 % in average.

关 键 词：依赖测试 依赖距离 向量因子 并行宽度 向量化 向量寄存器 

分 类 号：TP302.7[自动化与计算机技术—计算机系统结构]