Public international benchmarks for parallel computers: PARKBENCH committee: Report-1
Scientific Programming
MPI-The Complete Reference, Volume 1: The MPI Core
MPI-The Complete Reference, Volume 1: The MPI Core
Flattening on the Fly: Efficient Handling of MPI Derived Datatypes
Proceedings of the 6th European PVM/MPI Users' Group Meeting on Recent Advances in Parallel Virtual Machine and Message Passing Interface
SKaMPI: A Detailed, Accurate MPI Benchmark
Proceedings of the 5th European PVM/MPI Users' Group Meeting on Recent Advances in Parallel Virtual Machine and Message Passing Interface
Reproducible Measurements of MPI Performance Characteristics
Proceedings of the 6th European PVM/MPI Users' Group Meeting on Recent Advances in Parallel Virtual Machine and Message Passing Interface
Exploiting Transparent Remote Memory Access for Non-Contiguous- and One-Sided-Communication
IPDPS '02 Proceedings of the 16th International Parallel and Distributed Processing Symposium
Performance expectations and guidelines for MPI derived datatypes
EuroMPI'11 Proceedings of the 18th European MPI Users' Group conference on Recent advances in the message passing interface
Automatic memory optimizations for improving MPI derived datatype performance
EuroPVM/MPI'06 Proceedings of the 13th European PVM/MPI User's Group conference on Recent advances in parallel virtual machine and message passing interface
Micro-applications for communication data access patterns and MPI datatypes
EuroMPI'12 Proceedings of the 19th European conference on Recent Advances in the Message Passing Interface
| Hi-index | 0.00 |
We present an extension of the SKaMPI benchmark for MPI implementations to cover the derived datatype mechanism of MPI. All MPI constructors for derived datatypes are covered by the benchmark, and varied along different dimensions. This is controlled bya set of pre-defined patterns which can be instantiated bypa rameters given by the user in a configurations file. We classify the patterns into fixed types, dynamic types, nested types, and special types. We show results from the SKaMPI ping-pong measurement with the fixed and special types on three platforms: CrayT3E/900, IBM RS 6000SP, NEC SX-5. The machines show quite some difference in handling datatypes, with typically a significant penalty for nested types for the Cray (up to a factor of 16) and the IBM (up to a factor of 8), whereas the NEC treats these types very uniformly (overhead of between 2 and 4). Such results illustrate the need for a systematic datatype benchmark to help the MPI programmer select the most efficient data representation for a particular machine.