Tiwari, M and Vadhiyar, S (2022) Strategies for Efficient Execution of Pipelined Conjugate Gradient Method on GPU Systems. In: 37th International Conference on High Performance Computing , ISC High Performance 2022, 29 May - 2 June 2022, Hamburg, pp. 77-89.
PDF
lncs_2022.pdf - Published Version Restricted to Registered users only Download (544kB) | Request a copy |
Abstract
The Preconditioned Conjugate Gradient (PCG) method is widely used for solving linear systems of equations with sparse matrices. A recent version of PCG, Pipelined PCG (PIPECG), eliminates the dependencies in the computations of the PCG algorithm so that the non-dependent computations can be overlapped with communication. In this paper, we develop three methods for efficient execution of the Pipelined PCG algorithm on GPU accelerated heterogeneous architectures. The first two methods achieve task-parallelism using asynchronous executions of different tasks on multi-core CPU and a GPU. The third method achieves data parallelism by decomposing the workload between multi-core CPU and GPU based on a performance model. We performed experiments on both the K40 and V100 GPU systems and our methods give up to 8x speedup and on average 3x speedup over PCG CPU implementation of Paralution and PETSc libraries. They also give up to 5x speedup and on average 1.45x speedup over PCG GPU implementation of Paralution and PETSc libraries. The third method also provides an efficient solution for solving problems that cannot be fit into the GPU memory and gives up to 6.8x speedup for such problems. © 2022, Springer Nature Switzerland AG.
Item Type: | Conference Paper |
---|---|
Publication: | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
Publisher: | Springer Science and Business Media Deutschland GmbH |
Additional Information: | The copyright for this article belongs to Springer Science and Business Media Deutschland GmbH. |
Keywords: | Conjugate gradient method; Libraries; Linear systems; Pipelines, Asynchronous executions; Conjugate-gradient method; Heterogeneous architectures; Linear systems of equations; Multi-cores; PETSc libraries; Pipelined method; Preconditioned conjugate gradient; Preconditioned conjugate gradient algorithms; Preconditioned conjugate gradient method, Graphics processing unit |
Department/Centre: | Division of Interdisciplinary Sciences > Computational and Data Sciences |
Date Deposited: | 14 Mar 2023 07:14 |
Last Modified: | 14 Mar 2023 07:14 |
URI: | https://eprints.iisc.ac.in/id/eprint/81004 |
Actions (login required)
View Item |