Conjugate gradient method with graphics processing unit acceleration: CUDA vs OpenCL. (September 2017)

• Record Type: Journal Article
• Authors: Cheik Ahamed, Abal-Kassim; Magoulès, Frédéric
• Is Part Of: Advances in engineering software. Volume 111(2017)
• Page Start: 32
• Page End: 42
• Publication Date: 2017-09
• Subjects: Conjugate gradient method -- Iterative method -- Linear algebra -- Sparse matrix-vector product -- Parallel computing -- GPU -- CUDA -- OpenCL; Computer-aided engineering -- Periodicals; Engineering -- Computer programs -- Periodicals; Engineering -- Software -- Periodicals; Periodicals; 620.0028553
• DOI: 10.1016/j.advengsoft.2016.10.002
• Languages: English
• ISSNs: 0965-9978

Abstract

Highlights: Advances in Engineering Software. Authors response to the Editor on Conjugate gradient method with graphics processing unit acceleration: CUDA vs OpenCL. by Frdéric Magoulès and Abal-Kassim Cheik Ahamed We are grateful for the comments of the reviewers and the time they have taken to read our paper. We have modified the manuscript accordingly. A brief response to each comment and a description of the associated changes to the manuscript are appended below (in red text) to the original review. Abstract: Performance computations depend on the machine architecture, the operating system, the problem studied and obviously on the programming implementation. Solving partial differential equations by numerical methods such as the finite element method requires the solution of large sparse linear systems. Graphics processing unit (GPU) is now commonly used to accelerate numerical simulations and most supercomputers provide large number of GPUs to their users. This paper proposes a comparison of both CUDA and OpenCL GPU languages to take the highest performance of multi-GPUs clusters. We analyse, evaluate and compare their respective performances for computing linear algebra operations and for solving large sparse linear systems with the conjugate gradient iterative method on multi-GPUs clusters.