Extended cycle shrinking: A restructuring method for parallel compilation

Abstract

An important part of a parallelizing compiler is the restructuring phase, which extracts parallelism from a sequential program. We consider an important restructuring transformation, called cycle shrinking [1], which partitions the iteration space of a loop so that the iterations within each partition can be executed in parallel. We propose a new cycle shrinking transformation, called extended cycle shrinking, which is an improvement over the existing methods. We present the conditions under which our method can be applied, and give an algorithm which performs this transformation. Further, we present results to show that our method always leads to a minimal number of partitions, whereas the earlier methods do not. Thus our algorithms is, in this sense, optimal.