Portable randomized list ranking on multiprocessors using MPI

Purchase on Springer.com

$29.95 / €24.95 / £19.95*

* Final gross prices may vary according to local VAT.

Get Access

Abstract

We describe a simple multiprocessor list ranking algorithm with low communication volume and simple communication pattern. With p processors the algorithm performs < 4p (pipelined) communication rounds involving only point-to-point communication. For lists with N elements the algorithm runs in O(N ln p/p+p) time. Experiments with an implementation using MPI on a network of workstations and an IBM SP-2 comparing the algorithm to the well-known pointer jumping algorithm are reported. On the NOW the new algorithm is significantly better than pointer jumping. On the IBM SP-2 only the new algorithm was able to produce (modest) speed-up.