• Improving the performance of c...

Treffer: Improving the performance of classical linear algebra iterative methods via hybrid parallelism

Title:
Improving the performance of classical linear algebra iterative methods via hybrid parallelism
Authors:
Martínez Ferrer, Pedro José, Arslan, Tufan, Beltran Querol, Vicenç
Contributors:
Universitat Politècnica de Catalunya. Departament d'Arquitectura de Computadors, Barcelona Supercomputing Center, Universitat Politècnica de Catalunya. PM - Programming Models
Publisher Information:
Elsevier
Publication Year:
2023
Collection:
Universitat Politècnica de Catalunya, BarcelonaTech: UPCommons - Global access to UPC knowledge
Subject Terms:
Àrees temàtiques de la UPC::Informàtica::Arquitectura de computadors::Arquitectures paral·leles, Àrees temàtiques de la UPC::Informàtica::Informàtica teòrica::Algorísmica i teoria de la complexitat, Algebras, Linear, Parallel processing (Electronic computers), Linear algebra, Hybrid parallelism, Distributed-memory, Shared-memory, MPI, Àlgebra lineal, Processament en paral·lel (Ordinadors)
Document Type:
Fachzeitschrift article in journal/newspaper
File Description:
application/pdf
Language:
English
Relation:
https://www.sciencedirect.com/science/article/abs/pii/S0743731523000746?via%3Dihub.; Martínez, P.; Arslan, T.; Beltran, V. Improving the performance of classical linear algebra iterative methods via hybrid parallelism. "Journal of parallel and distributed computing", Setembre 2023, vol. 179, article 104711.; https://arxiv.org/abs/2305.05988; http://hdl.handle.net/2117/387845
DOI:
10.1016/j.jpdc.2023.04.012
Availability:
http://hdl.handle.net/2117/387845
https://arxiv.org/abs/2305.05988
https://doi.org/10.1016/j.jpdc.2023.04.012
Rights:
Attribution-NonCommercial-NoDerivatives 4.0 International ; http://creativecommons.org/licenses/by-nc-nd/4.0/ ; Open Access
Accession Number:
edsbas.6364E205
Database:
BASE

Weitere Informationen

We propose fork-join and task-based hybrid implementations of four classical linear algebra iterative methods (Jacobi, Gauss–Seidel, conjugate gradient and biconjugate gradient stabilized) on CPUs as well as variations of them. This class of algorithms, that are ubiquitous in computational frameworks, are duly documented and the corresponding source code is made publicly available for reproducibility. Both weak and strong scalability benchmarks are conducted to statistically analyse their relative efficiencies. The weak scalability results assert the superiority of a task-based hybrid parallelisation over MPI-only and fork-join hybrid implementations. Indeed, the task-based model is able to achieve speedups of up to 25% larger than its MPI-only counterpart depending on the numerical method and the computational resources used. For strong scalability scenarios, hybrid methods based on tasks remain more efficient with moderate computational resources where data locality does not play an important role. Fork-join hybridisation often yields mixed results and hence does not seem to bring a competitive advantage over a much simpler MPI approach ; This work has received funding from the European High Performance Computing Joint Undertaking (EuroHPC JU) initiative [grant number 956416] via the exaFOAM research project. The JU receives support from the European Union's Horizon 2020 research and innovation programme and France, Germany, Spain, Italy, Croatia, Greece, and Portugal. In Spain, it has received complementary funding from MCIN/AEI/10.13039/501100011033 [grant number PCI2021-121961]. This work has also benefited financially from the Ramón y Cajal programme [grant number RYC2019-027592-I] funded by MCIN/AEI and ESF/10.13039/501100004895 as well as the Severo Ochoa Centre of Excellence accreditation [grant number CEX2021-001148-S] funded by MCIN/AEI. The Programming Models research group at BSC-UPC received financial support from Departament de Recerca i Universitats de la Generalitat de Catalunya [grant number ...