Abstract
From execution time analysis of programs, nested loops are often the most time-consuming part and the main source of parallelism. In this paper, an index shift method is proposed to improve the parallelism of nested loops with only slight modification to original programs. The idea behind this index shift method is to defer or advance the execution steps of some statements such that the parallelism of the loops can be increased under the Hyperplane method. Based on this method, an algorithm, under the assumption that a time function is fixed, is devised to improve the parallelism of nested loops.