A loop transformation using two parallel region partitioning method

  • Authors:
  • Sam Jin Jeong;Jung Soo Han

  • Affiliations:
  • Division of Information and Communication Engineering, Cheonan University, Cheonan City, Korea;Division of Information and Communication Engineering, Cheonan University, Cheonan City, Korea

  • Venue:
  • APPT'05 Proceedings of the 6th international conference on Advanced Parallel Processing Technologies
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

Loop parallelization is an important optimization issue in the execution of scientific programs. This paper proposes loop transformation techniques for finding parallel regions within nested loops with non-uniform dependences in order to improve parallelism. By parallelizing anti dependence region using variable renaming, there remains only flow dependence in the loop. We then divide the iteration space into FDT (Flow Dependence Tail set) and FDH (Flow Dependence Head set). By two given equations, we show how to determine whether the intersection of FDT and FDH is empty or not. So, we can find two parallel regions for doubly nested loops with non-uniform dependences. In the case that FDT does not overlap FDH, we will divide the iteration space into two parallel regions by a line.