Implementação de overthread em uma aplicação de método numérico HOPMOC através de um algoritmo híbrido MPI/OpenMP
Abstract
This article presents a study of an overthread strategy applied to the Hopmoc numerical method, which consists of allocating a number of processes and threads higher than the number of available physical cores. The results obtained are presented and analyzed through different metrics. By the overthread strategy, we showed that it was possible to increase the performance of the application in a parallel environment beyond the maximum physical cores capacity of the machine, and that this gain is possible due to the elimination of native OpenMP synchronization barriers.References
Bassi, F., Colombo, A., Crivellini, A., and Franciolini, M. (2016). Hybrid openmp/mpi In 7th parallelization of a high–order discontinuous galerkin cfd/caa solver. European Congress on Computational Methods in Applied Sciences and Engineering, ECCOMAS Congress, pages 7992–8012.
Cabral, F. L., Gonzaga de Oliveira, S. L., Osthoff, C., Costa, G. P., Brandão, D. N., and Kischinhevsky, M. (2019). An evaluation of MPI and OpenMP paradigms in nite-difference explicit methods for PDEs on shared-memory multi-and manycore systems. Concurrency and Computation: Practice and Experience, page e5642.
Cabral, F. L., Osthoff, C., Costa, G. P., Brandão, D., Kischinhevsky, M., and Gonzaga de Oliveira, S. L. (2017). Tuning up the TVD-HOPMOC method on Intel MIC Xeon Phi architectures with Intel Parallel Studio tools. In 2017 International Symposium on Computer Architecture and High Performance Computing Workshops (SBAC-PADW), pages 19–24. IEEE.
Cabral, F. L., Osthoff, C., Costa, G. P., Gonzaga de Oliveira, S. L., Brandão, D., and Kischinhevsky, M. (2018a). An OpenMP implementation of the TVD–Hopmoc method based on a synchronization mechanism using locks between adjacent threads In International Conference on Computational on Xeon Phi(TM) accelerators. Science, pages 701–707. Springer.
Cabral, F. L., Osthoff, C., Souto, R. P., Costa, G. P., Gonzaga de Oliveira, S. L., Brandão, D., and Kischinhevsky, M. (2018b). Fine-tuning an OpenMP-based TVD–Hopmoc method using Intel R(cid:13) Parallel Studio XE Tools on Intel R(cid:13) Xeon R(cid:13) architectures. In Latin American High Performance Computing Conference, pages 194–209. Springer.
Costa, G., Cabral, F., and Osthoff, C. (2019). Otimização do método hopmoc 1d com auxílio das ferramentas intel parallel studio. In Anais Estendidos do XX Simpósio em Sistemas Computacionais de Alto Desempenho, pages 41–48, Porto Alegre, RS, Brasil. SBC.
Diener, M., White, S., Kale, L. V., Campbell, M., Bodony, D. J., and Freund, J. B. (2017). Improving the memory access locality of hybrid mpi applications. In Proceedings of the 24th European MPI Users' Group Meeting, page 11. ACM.
Jeffers, J., Reinders, J., and Sodani, A. (2016). Intel Xeon Phi Processor High Performance Programming – Knights Landing Edition. Morgan Kaufmann, Burlington, MA, 2 edition.
Cabral, F. L., Gonzaga de Oliveira, S. L., Osthoff, C., Costa, G. P., Brandão, D. N., and Kischinhevsky, M. (2019). An evaluation of MPI and OpenMP paradigms in nite-difference explicit methods for PDEs on shared-memory multi-and manycore systems. Concurrency and Computation: Practice and Experience, page e5642.
Cabral, F. L., Osthoff, C., Costa, G. P., Brandão, D., Kischinhevsky, M., and Gonzaga de Oliveira, S. L. (2017). Tuning up the TVD-HOPMOC method on Intel MIC Xeon Phi architectures with Intel Parallel Studio tools. In 2017 International Symposium on Computer Architecture and High Performance Computing Workshops (SBAC-PADW), pages 19–24. IEEE.
Cabral, F. L., Osthoff, C., Costa, G. P., Gonzaga de Oliveira, S. L., Brandão, D., and Kischinhevsky, M. (2018a). An OpenMP implementation of the TVD–Hopmoc method based on a synchronization mechanism using locks between adjacent threads In International Conference on Computational on Xeon Phi(TM) accelerators. Science, pages 701–707. Springer.
Cabral, F. L., Osthoff, C., Souto, R. P., Costa, G. P., Gonzaga de Oliveira, S. L., Brandão, D., and Kischinhevsky, M. (2018b). Fine-tuning an OpenMP-based TVD–Hopmoc method using Intel R(cid:13) Parallel Studio XE Tools on Intel R(cid:13) Xeon R(cid:13) architectures. In Latin American High Performance Computing Conference, pages 194–209. Springer.
Costa, G., Cabral, F., and Osthoff, C. (2019). Otimização do método hopmoc 1d com auxílio das ferramentas intel parallel studio. In Anais Estendidos do XX Simpósio em Sistemas Computacionais de Alto Desempenho, pages 41–48, Porto Alegre, RS, Brasil. SBC.
Diener, M., White, S., Kale, L. V., Campbell, M., Bodony, D. J., and Freund, J. B. (2017). Improving the memory access locality of hybrid mpi applications. In Proceedings of the 24th European MPI Users' Group Meeting, page 11. ACM.
Jeffers, J., Reinders, J., and Sodani, A. (2016). Intel Xeon Phi Processor High Performance Programming – Knights Landing Edition. Morgan Kaufmann, Burlington, MA, 2 edition.
Published
2020-10-21
How to Cite
COSTA, Gabriel; TEIXEIRA, Thiago; CABRAL, Frederico; OSTHOFF, Carla.
Implementação de overthread em uma aplicação de método numérico HOPMOC através de um algoritmo híbrido MPI/OpenMP. In: UNDERGRADUATE RESEARCH WORKSHOP - SYMPOSIUM ON HIGH PERFORMANCE COMPUTING SYSTEMS (SSCAD), 21. , 2020, Evento Online.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2020
.
p. 70-77.
DOI: https://doi.org/10.5753/wscad_estendido.2020.14091.
