| Summary: | 碩士 === 大同工學院 === 資訊工程研究所 === 86 === There are many parallelizing compiling techniques for expoit the potentialparallelism in loop. Data dependence test is one of the important techniques.Many data dependence tests have been proposed, however, each of them has itsown favor on loop input data and not proper for all input cases. We propose a new method, Weight test (W test), which will be suitable for all cases. In thesis, an approach that integrates existing test methods and make good use oftheir advantages. This approach chooses the appropriate test by computing theweight of input features of loops, and applies the proper test to detect data dependences on loops. From our experimental results, the W test gives exactsolution in most of input cases and less overhead than other integrated methods.
|
|---|
