Achieving Performance Portability with SKaMPI for High-Performance MPI Programs
ICCS '01 Proceedings of the International Conference on Computational Science-Part II
On Benchmarking Collective MPI Operations
Proceedings of the 9th European PVM/MPI Users' Group Meeting on Recent Advances in Parallel Virtual Machine and Message Passing Interface
Using SKaMPI for developing high-performance MPI programs with performance portability
Future Generation Computer Systems - Tools for program development and analysis
A Method for MPI Broadcast in Computational Grids
IPDPS '05 Proceedings of the 19th IEEE International Parallel and Distributed Processing Symposium (IPDPS'05) - Workshop 13 - Volume 14
SKaMPI: a comprehensive benchmark for public benchmarking of MPI
Scientific Programming
MGF: A grid-enabled MPI library
Future Generation Computer Systems
Enabling collective communications between components
Proceedings of the 2007 symposium on Component and framework technology in high-performance and scientific computing
Proceedings of the 15th European PVM/MPI Users' Group Meeting on Recent Advances in Parallel Virtual Machine and Message Passing Interface
Exploring unexpected behavior in MPI
HPCC'06 Proceedings of the Second international conference on High Performance Computing and Communications
MGF: a grid-enabled MPI library with a delegation mechanism to improve collective operations
PVM/MPI'05 Proceedings of the 12th European PVM/MPI users' group conference on Recent Advances in Parallel Virtual Machine and Message Passing Interface
An efficient collective communication method using a shortest path algorithm in a computational grid
GCC'05 Proceedings of the 4th international conference on Grid and Cooperative Computing
Improving multilevel approach for optimizing collective communications in computational grids
EGC'05 Proceedings of the 2005 European conference on Advances in Grid Computing
| Hi-index | 0.00 |
An MPI library's implementation of broadcast communication can significantly affect the performance of applications built with that library. In order to choose between similar implementations or to evaluate available libraries, accurate measurements of broadcast performance are required. As we demonstrate, existing methods for measuring broadcast performance are either inaccurate or inadequate. Fortunately, we have designed an accurate method for measuring broadcast performance, even in a challenging grid environment.Measuring broadcast performance is not easy. Simply sending one broadcast after another allows them to proceed through the network concurrently, thus resulting in inaccurate per broadcast timings. Existing methods either fail to eliminate this pipelining effect or eliminate it by introducing overheads that are as difficult to measure as the performance of the broadcast itself. This problem becomes even more challenging in grid environments. Latencies along different links can vary significantly. Thus, an algorithm's performance is difficult to predict from it's communication pattern. Even when accurate prediction is possible, the pattern is often unknown. Our method introduces a measurable overhead to eliminate the pipelining effect, regardless of variations in link latencies.