Due to the importance of data dependence test for exploiting parallelisms in loop, an approach that integrates existing test methods and makes good use of their advantages is proposed in this study. The proposed method, called weighting characteristic test, will be suitable for all cases due to considering their characteristics of input data. By computing the weight of characteristics for loop input data, the proposed approach chooses the appropriate test and applies it to detect data dependences on loops. The weighting characteristic test gives exact solution in most of input cases and less overhead than other integrated methods from our experimental results. Thus, the proposed method is more applicable than existed data dependence tests for exploiting loop parallelism indeed.
展开▼
