Dependence Testing Without Induction Variable Substitution

International audienceWe present a new approach to dependence testing in the presence of induction variables. Instead of looking forclosed form expressions, our method computes monotonic evolution which captures the direction in which the valueof a variable changes. This information is used for dependence testing of array references. Under this scheme, closedform computation and induction variable substitution can be delayed until after the dependence test and be performedon-demand. The technique can be extended to dynamic data structures, using either pointer-based implementationsor standard object-oriented containers. To improve efficiency, we also propose an optimized (non-iterative) data-flowalgorithm to compute evolution. Experimental results show that dependence tests based on evolution informationmatch the accuracy of that based on closed-form computation (implemented in Polaris), and when no closed formexpressions can be calculated, our method is more accurate than that of Polaris