Single-port and multi-port collective communication operations on single and dual Cell BE processor systems

Abstract

Recently, a set of factors has been leading high-performance processor architectures toward designs that feature multiple processing cores on a single chip (a.k.a. CMP). The cell broadband engine (BE) shows potential to provide high-performance to parallel applications (e.g., MPI applications). An efficient implementation of collective communication operations is one of the key issues to reach high-performance and scalability in parallel applications. In this work, we implement several collective communications and investigate their performance in terms of latency and the associated components. For this, broadcast, all-gather and total-exchange functions are implemented on the Cell BE processor.