Transformations techniques for extracting parallelism in non-uniform nested loops

Executing a program in parallel machines needs not only to find sufficient parallelism in a program, but it is also important that we minimize the synchronization and communication overheads in the parallelized program. This yields to improve the performance by reducing the time needed for executing the program. Parallelizing and partitioning of nested loops requires efficient iteration dependence analysis. Although many loop transformations techniques exist for nested loop partitioning, most of these transformation techniques perform poorly when parallelizing nested loops with non-uniform (irregular) dependences. In this paper the affine and unimodular transformations are applied to solve the problem of parallelism in nested loops with non-uniform dependence vectors. To solve these problem few researchers converted the non-uniform nested loops to uniform nested loops and then find the parallelism. We propose applying directly the two approaches affine and unimodular transformations to extract and improve the parallelism in nested loops with non-uniform dependences. The study shows that unimodular transformation is better than affine transformation when the dependences in nested loops exist only in one statement. While affine transformation is more effective when the nested loops have a sequence of statements and the dependence exists between these different statements.