Computational Technology Resources - CCP

Paper 222

Comparison of GPU-Based Parallel Assembly and Assembly-Free Sparse Matrix Vector Multiplication for Finite Element Analysis of Three-Dimensional Structures

A. Akbariyeh, B.H. Dennis, B.P. Wang and K.L. Lawrence

Mechanical and Aerospace Engineering, University of Texas at Arlington, United States of America

doi:10.4203/ccp.108.222

purchase the full-text of this paper

A. Akbariyeh, B.H. Dennis, B.P. Wang, K.L. Lawrence, "Comparison of GPU-Based Parallel Assembly and Assembly-Free Sparse Matrix Vector Multiplication for Finite Element Analysis of Three-Dimensional Structures", in J. Kruis, Y. Tsompanakis, B.H.V. Topping, (Editors), "Proceedings of the Fifteenth International Conference on Civil, Structural and Environmental Engineering Computing", Civil-Comp Press, Stirlingshire, UK, Paper 222, 2015. doi:10.4203/ccp.108.222

Keywords: finite element, parallel assembly, sparse matrix vector multiplication, SpMV, assembly free, matrix free, scientific computing, GPU, CUDA, parallel programming, block compressed sparse row.

Summary

In this paper we present a GPU assembly free kernel for sparse matrix vector multiplication in the context of the finite element method. An element based approach, has been implemented, for unstructured four-node tetrahedral meshes. In this approach the explicit assembly and storage of the global stiffness matrix is avoided, thus saving memory and reducing time spent accessing main memory. A colouring technique is used to subdivide the mesh into disjoint elements of same colours. Sparse matrix vector multiplication and parallel assembly on GPUs with the assembly free kernel is compared. All GPU codes are written in the CUDA language but the algorithm is not restricted to specific hardware.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £75 +P&P)

	Computational & Technology Resources an online resource for computational, engineering & technology publications
	not logged in - login
Front Page Browse CCP CSETS CTR IJRT Other Authors Search Purchase Guide FAQ Contact us	Civil-Comp Proceedings ISSN 1759-3433 CCP: 108 PROCEEDINGS OF THE FIFTEENTH INTERNATIONAL CONFERENCE ON CIVIL, STRUCTURAL AND ENVIRONMENTAL ENGINEERING COMPUTING Edited by: J. Kruis, Y. Tsompanakis and B.H.V. Topping Paper 222 Comparison of GPU-Based Parallel Assembly and Assembly-Free Sparse Matrix Vector Multiplication for Finite Element Analysis of Three-Dimensional Structures A. Akbariyeh, B.H. Dennis, B.P. Wang and K.L. Lawrence Mechanical and Aerospace Engineering, University of Texas at Arlington, United States of America doi:10.4203/ccp.108.222 purchase the full-text of this paper Full Bibliographic Reference for this paper A. Akbariyeh, B.H. Dennis, B.P. Wang, K.L. Lawrence, "Comparison of GPU-Based Parallel Assembly and Assembly-Free Sparse Matrix Vector Multiplication for Finite Element Analysis of Three-Dimensional Structures", in J. Kruis, Y. Tsompanakis, B.H.V. Topping, (Editors), "Proceedings of the Fifteenth International Conference on Civil, Structural and Environmental Engineering Computing", Civil-Comp Press, Stirlingshire, UK, Paper 222, 2015. doi:10.4203/ccp.108.222 Keywords: finite element, parallel assembly, sparse matrix vector multiplication, SpMV, assembly free, matrix free, scientific computing, GPU, CUDA, parallel programming, block compressed sparse row. Summary In this paper we present a GPU assembly free kernel for sparse matrix vector multiplication in the context of the finite element method. An element based approach, has been implemented, for unstructured four-node tetrahedral meshes. In this approach the explicit assembly and storage of the global stiffness matrix is avoided, thus saving memory and reducing time spent accessing main memory. A colouring technique is used to subdivide the mesh into disjoint elements of same colours. Sparse matrix vector multiplication and parallel assembly on GPUs with the assembly free kernel is compared. All GPU codes are written in the CUDA language but the algorithm is not restricted to specific hardware. purchase the full-text of this paper (price £20) go to the previous paper go to the next paper return to the table of contents return to the book description purchase this book (price £75 +P&P)
Back to top	©Civil-Comp Limited 2023 - terms & conditions