Symmetric Tridiagonal Eigenvalue Solver Across CPU Graphics Processing Unit (GPU) Nodes

In this work, an improved and scalable implementation of Cuppen’s algorithm for diagonalizing symmetric tridiagonal matrices is presented. This approach uses a hybrid-heterogeneous parallelization technique, taking advantage of GPU and CPU in a distributed hardware architecture. Cuppen’s algorithm i...

Full description

Saved in:
Bibliographic Details
Main Authors: Erika Hernández-Rubio, Alberto Estrella-Cruz, Amilcar Meneses-Viveros, Jorge Alberto Rivera-Rivera, Liliana Ibeth Barbosa-Santillán, Sergio Víctor Chapa-Vergara
Format: Article
Language:English
Published: MDPI AG 2024-11-01
Series:Applied Sciences
Subjects:
Online Access:https://www.mdpi.com/2076-3417/14/22/10716
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846154423603036160
author Erika Hernández-Rubio
Alberto Estrella-Cruz
Amilcar Meneses-Viveros
Jorge Alberto Rivera-Rivera
Liliana Ibeth Barbosa-Santillán
Sergio Víctor Chapa-Vergara
author_facet Erika Hernández-Rubio
Alberto Estrella-Cruz
Amilcar Meneses-Viveros
Jorge Alberto Rivera-Rivera
Liliana Ibeth Barbosa-Santillán
Sergio Víctor Chapa-Vergara
author_sort Erika Hernández-Rubio
collection DOAJ
description In this work, an improved and scalable implementation of Cuppen’s algorithm for diagonalizing symmetric tridiagonal matrices is presented. This approach uses a hybrid-heterogeneous parallelization technique, taking advantage of GPU and CPU in a distributed hardware architecture. Cuppen’s algorithm is a theoretical concept and a powerful tool in various scientific and engineering applications. It is a key player in matrix diagonalization, finding its use in Functional Density Theory (FDT) and Spectral Clustering. This highly efficient and numerically stable algorithm computes eigenvalues and eigenvectors of symmetric tridiagonal matrices, making it a crucial component in many computational methods. One of the challenges in parallelizing algorithms for GPUs is their limited memory capacity. However, we overcome this limitation by utilizing multiple nodes with both CPUs and GPUs. This enables us to solve subproblems that fit within the memory of each device in parallel and subsequently combine these subproblems to obtain the complete solution. The hybrid-heterogeneous approach proposed in this work outperforms the state-of-the-art libraries and also maintains a high degree of accuracy in terms of orthogonality and quality of eigenvectors. Furthermore, the sequential version of the algorithm with our approach in this work demonstrates superior performance and potential for practical use. In the experiments carried out, it was possible to verify that the performance of the implementation that was carried out scales by 2× using two graphic cards in the same node. Notably, Symmetric Tridiagonal Eigenvalue Solvers are fundamental to solving more general eigenvalue problems. Additionally, the divide-and-conquer approach employed in this implementation can be extended to singular value solvers. Given the wide range of eigenvalue problems encountered in scientific and engineering domains, this work is essential in advancing computational methods for efficient and accurate matrix diagonalization.
format Article
id doaj-art-c8b95c3c1f9848ad82824777418fa2e7
institution Kabale University
issn 2076-3417
language English
publishDate 2024-11-01
publisher MDPI AG
record_format Article
series Applied Sciences
spelling doaj-art-c8b95c3c1f9848ad82824777418fa2e72024-11-26T17:49:53ZengMDPI AGApplied Sciences2076-34172024-11-0114221071610.3390/app142210716Symmetric Tridiagonal Eigenvalue Solver Across CPU Graphics Processing Unit (GPU) NodesErika Hernández-Rubio0Alberto Estrella-Cruz1Amilcar Meneses-Viveros2Jorge Alberto Rivera-Rivera3Liliana Ibeth Barbosa-Santillán4Sergio Víctor Chapa-Vergara5Sección de Estudios de Posgrado e Invetigación, Escuela Superior de Cómputo, Instituto Politécnico Nacional, Mexico City 07320, MexicoDepartamento de Computación, Cinvestav-IPN, Mexico City 07360, MexicoDepartamento de Computación, Cinvestav-IPN, Mexico City 07360, MexicoEscuela Superior de Cómputo, Instituto Politécnico Nacional, Mexico City 07320, MexicoDepartamento de Ciencias Computacionales, Instituto Tecnológico y de Estudios Superiores de Monterrey, Monterrey 45138, MexicoDepartamento de Computación, Cinvestav-IPN, Mexico City 07360, MexicoIn this work, an improved and scalable implementation of Cuppen’s algorithm for diagonalizing symmetric tridiagonal matrices is presented. This approach uses a hybrid-heterogeneous parallelization technique, taking advantage of GPU and CPU in a distributed hardware architecture. Cuppen’s algorithm is a theoretical concept and a powerful tool in various scientific and engineering applications. It is a key player in matrix diagonalization, finding its use in Functional Density Theory (FDT) and Spectral Clustering. This highly efficient and numerically stable algorithm computes eigenvalues and eigenvectors of symmetric tridiagonal matrices, making it a crucial component in many computational methods. One of the challenges in parallelizing algorithms for GPUs is their limited memory capacity. However, we overcome this limitation by utilizing multiple nodes with both CPUs and GPUs. This enables us to solve subproblems that fit within the memory of each device in parallel and subsequently combine these subproblems to obtain the complete solution. The hybrid-heterogeneous approach proposed in this work outperforms the state-of-the-art libraries and also maintains a high degree of accuracy in terms of orthogonality and quality of eigenvectors. Furthermore, the sequential version of the algorithm with our approach in this work demonstrates superior performance and potential for practical use. In the experiments carried out, it was possible to verify that the performance of the implementation that was carried out scales by 2× using two graphic cards in the same node. Notably, Symmetric Tridiagonal Eigenvalue Solvers are fundamental to solving more general eigenvalue problems. Additionally, the divide-and-conquer approach employed in this implementation can be extended to singular value solvers. Given the wide range of eigenvalue problems encountered in scientific and engineering domains, this work is essential in advancing computational methods for efficient and accurate matrix diagonalization.https://www.mdpi.com/2076-3417/14/22/10716Cuppen’s algorithmEigenvalue SolverGraphics Processing UnitHybrid-Heterogeneous computing
spellingShingle Erika Hernández-Rubio
Alberto Estrella-Cruz
Amilcar Meneses-Viveros
Jorge Alberto Rivera-Rivera
Liliana Ibeth Barbosa-Santillán
Sergio Víctor Chapa-Vergara
Symmetric Tridiagonal Eigenvalue Solver Across CPU Graphics Processing Unit (GPU) Nodes
Applied Sciences
Cuppen’s algorithm
Eigenvalue Solver
Graphics Processing Unit
Hybrid-Heterogeneous computing
title Symmetric Tridiagonal Eigenvalue Solver Across CPU Graphics Processing Unit (GPU) Nodes
title_full Symmetric Tridiagonal Eigenvalue Solver Across CPU Graphics Processing Unit (GPU) Nodes
title_fullStr Symmetric Tridiagonal Eigenvalue Solver Across CPU Graphics Processing Unit (GPU) Nodes
title_full_unstemmed Symmetric Tridiagonal Eigenvalue Solver Across CPU Graphics Processing Unit (GPU) Nodes
title_short Symmetric Tridiagonal Eigenvalue Solver Across CPU Graphics Processing Unit (GPU) Nodes
title_sort symmetric tridiagonal eigenvalue solver across cpu graphics processing unit gpu nodes
topic Cuppen’s algorithm
Eigenvalue Solver
Graphics Processing Unit
Hybrid-Heterogeneous computing
url https://www.mdpi.com/2076-3417/14/22/10716
work_keys_str_mv AT erikahernandezrubio symmetrictridiagonaleigenvaluesolveracrosscpugraphicsprocessingunitgpunodes
AT albertoestrellacruz symmetrictridiagonaleigenvaluesolveracrosscpugraphicsprocessingunitgpunodes
AT amilcarmenesesviveros symmetrictridiagonaleigenvaluesolveracrosscpugraphicsprocessingunitgpunodes
AT jorgealbertoriverarivera symmetrictridiagonaleigenvaluesolveracrosscpugraphicsprocessingunitgpunodes
AT lilianaibethbarbosasantillan symmetrictridiagonaleigenvaluesolveracrosscpugraphicsprocessingunitgpunodes
AT sergiovictorchapavergara symmetrictridiagonaleigenvaluesolveracrosscpugraphicsprocessingunitgpunodes