On Fast Array Data Dependence Tests

Array data-dependence analysis is an important part of any optimizing compiler for scientiic programs. The Omega test is an exact test for integer solutions to aane constraints and can be used for array data dependence. There are other tests that are less exact but are intended to be faster. Many of these less exact tests are rather complicated and designed to be as accurate as possible while still being fast. In this paper, we describe the Epsilon test, intended to be as simple and as fast as possible, while not being embarrassingly inaccurate. We explore the relative speed and accuracy of the Epsilon and Omega test, and discuss how they might be joined. We also point out serious errors in recent published work on array data dependence tests.