Shared Memory-based Algorithms for Vector Collective Operations of the MPI Standard
https://doi.org/10.55648/1998-6920-2022-16-3-38-47
Abstract
Algorithms for MPI_Scatterv, MPI_Gatherv, MPI_Allgatherv collective operations using the shared memory of multiprocessor SMP/NUMA systems are proposed. The algorithms use an approach based on copying message fragments via queues in a shared memory segment (Copy-In-Copy-Out). The algorithms were implemented based on Open MPI and the execution time is reduced on average by 20-40% for the MPI_Scatterv, MPI_Gatherv, MPI_Allgatherv operations compared to the coll/tuned component of the Open MPI library.
About the Authors
A. A. Romanyuta
Siberian State University of Telecommunications and Information Science
Russian Federation
Russian Federation
Alexey A. Romanyuta - Computer systems department, SibSUTIS.
Novosibirsk.
M. G. Kurnosov
Siberian State University of Telecommunications and Information Science; Rzhanov Institute of Semiconductor Physics of the Siberian Branch of the RAS
Russian Federation
Russian Federation
Mikhail G. Kurnosov - Doctor of Technical Science, Professor, SibSUTIS; Senior Research Scientist, Computer Systems Laboratory, Rzhanov Institute of Semiconductor Physics of the Siberian Branch of the RAS.
Novosibirsk.
References
1. MPI: A Message-Passing Interface Standard Version 4.0, available at: http://www.mpi-forum.org/docs/mpi-4.0/mpi40-report.pdf (accessed: 15.05.2022).
2. Open MPI: Open Source High Performance Computing, available at: http://www.open-mpi.org (accessed: 15.05.2022)
3. CMA: Cross Memory Attach, available at: https://lwn.net/Articles/405284/ (accessed: 20.05.2022)
4. KNEM High-Performance Intra-Node MPI Communication, available at: https://knem.gitlabpages.inria.fr (accessed: 20.05.2022)
5. XPMEM: Linux Cross-Memory Attach, available at: https://github.com/hjelmn/xpmem (accessed: 20.05.2022)
6. Kurnosov, M. G., Tokmasheva E. I. Algoritm shirokoveshchatel'noj peredachi standarta MPI na baze razdelyaemoj pamyati mnogoprocessornyh NUMA-uzlov [MPI Broadcast Algorithm Based on Shared Memory of Multiprocessor NUMA Nodes]. Vestnik SibGUTI, 2020, no. 1, pp. 42-59.
7. Graham R. L., Shipman G. MPI Support for Multi-core Architectures: Optimized Shared Memory Collectives. Proc. of the 15th European PVM/MPI Users’ Group Meeting, 2008, pp. 130–140.
8. Jain S., Kaleem R., Balmana M., Langer A., Durnov D., Sannikov A. and Garzaran M. Framework for Scalable Intra-Node Collective Operations using Shared Memory. Proc. of the International Conference for High Performance Computing, Networking, Storage, and Analysis (SC-2018), 2018, pp. 374–385.
Review
For citations:
Romanyuta A.A., Kurnosov M.G. Shared Memory-based Algorithms for Vector Collective Operations of the MPI Standard. The Herald of the Siberian State University of Telecommunications and Information Science. 2022;(3):38-47. (In Russ.) https://doi.org/10.55648/1998-6920-2022-16-3-38-47
Views:
3479
Email this article 