Efficient list algorithms for irregular block redistribution in parallelizing compilers

In parallelizing compilers on distributed memory systems, distributions of irregular sized array blocks are provided for load balancing and irregular problems. The irregular redistribution is different from the regular block-cyclic redistribution. This paper is devoted to developing algorithms for irregular redistribution that attempt to obtain near optimal scheduling while satisfying the minimal communication costs condition and the minimal step condition. Efficient algorithms are developed and their experimental results are compared. One improved list algorithm provides more chance for conflict messages in its relocation phase. It allocates conflict messages through methods used in a divide-and-conquer algorithm and a relocation algorithm proposed previously. The method of selecting the smallest relocation cost guarantees that the improved list algorithm is more efficient than others in average.