An Optimal In-Place Parallel Quicksort

Abstract

We give a complete analysis of the parallel quick-sort algorithm which we call partswap. Unlike many other parallel quicksort algorithms, the partswap algorithm is in-place, that is, no auxiliary memory is needed. We Will also show that partswap achieves maximal speed-up on the average. Average-case analyses are also given to justify the choice of the load distribution strategy over a straightforward and seemingly equally efficient load distribution strategy.