数据相关性分析中替代约束的有效性  

Effectiveness of surrogating constraints in data dependence analysis

作　　者：乔林[1] 黄维通[1] 孟威[1] 汤志忠[2] 

机构地区：[1]清华大学计算机与信息管理中心,北京100084 [2]清华大学计算机系高性能计算所,北京100084

出　　处：《计算机工程与设计》2006年第5期724-727,共4页Computer Engineering and Design

基　　金：国家自然科学基金项目(69773028;60173010)

摘　　要：数据相关性分析经常使用诸如方向向量或体差不等式等作为替代约束以加强求解丢番图方程的能力。给出了替代约束有效性的几个测度,详细分析了什么样的替代约束是可行的、有效的、良性的或恰当的,并证明在大多数情况下体差不等式约束要比方向向量约束更有效,体差不等式测试方法的收敛性要强于其它使用方向向量作为替代约束的传统数据相关性测试算法。Most data dependence analysis methods often use surrogating constraints, e.g., direction vectors and dependence difference inequalities, to enforce their capabilities for solving the Diophantine equation. Several measures of effectiveness of surrogating constraints were presented, what surrogating constraints are feasible, effective, well and exact was discussed in detail. In most cases, as the paper proved, dependence difference inequalities arc more effective than direction vectors, and this leads to that the convergence of dependence difference inequality test is better than other traditional data dependence analysis methods by using direction vectors as surrogating con- straints.

关 键 词：软件流水 数据相关性 替代约束 迭代向量 体差不等式测试 有效约束 

分 类 号：TP33[自动化与计算机技术—计算机系统结构]