The existing parallelizing compilers can parallelize most of the loops with uniform dependences, but they do not satisfactorily handle loops with non-uniform dependences. Most of the time, the compiler leaves such loops running sequentially. Unfortunately, loops with non-uniform dependences are not so uncommon in the real world. This paper presents the two parallel region partitioning method of nested loops with non-uniform dependences for maximizing parallelism. By parallelizing anti dependence region using variable renaming, we will divide the iteration space into two parallel regions by a line in case that FDT (Flow Dependence Tail set) does not overlap FDH (Flow Dependence Head set). Comparison with some related works shows more parallelism than other existing methods. © Springer-Verlag Berlin Heidelberg 2005.
CITATION STYLE
Jeong, S. J. (2005). Maximizing parallelism for non-uniform dependence Loops using two parallel region partitioning method. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 3605 LNCS, pp. 88–93). Springer Verlag. https://doi.org/10.1007/11535409_12
Mendeley helps you to discover research relevant for your work.