MPI Support for Multi-core Architectures: Optimized Shared Memory Collectives
Proceedings of the 15th European PVM/MPI Users' Group Meeting on Recent Advances in Parallel Virtual Machine and Message Passing Interface
Hi-index | 0.00 |
In this paper we investigate two strategies to improve the performance of collective communications on a cluster of SMPs. In the first part of the paper we explore the advantages and disadvantages of the commonly used strategy "optimize shared memory send and receive only". In the second part of the paper we show a model to design collective communications on SMP clusters which can be used to improve the disadvantages of the former strategy. Several collective operations are implemented through the later process, and the experimental results show better performance than vendor supplied implementations.