New basic linear algebra methods for simulation on GPUs
Pages 258 - 265
Abstract
We have used Graphics Processing Units (GPUs) to accelerate the solution of the types of equations typically encountered in dynamic system simulators. Compared to commercial matrix solvers that run on a CPU, we realized speedups ranging from 5 (for system size ≈ 700) to 460 (for system size ≈ 5800). While calculation time for the commercial matrix solver increased with matrix size ≈ O(N)2.3, our new GPU-based Preconditioned Generalized Minimal Residual (PGM-RES) technique yielded scaling as ≈ O(N)1.2. A significant component of this performance was achieved by development of new Basic Linear Algebra routines for the NVIDIA Tesla GPU that directly address characteristics typical of matrices that describe the time domain response of naturally-coupled dynamic systems.
References
[1]
Levy, M and Agarwal, A. The kill rule for multicore. In the 44th annual conference on Design automation, San Diego, California, pages 750--753, 2007.
[2]
J Andrzej and Robert.P.B. The parallel algorithm of conjugate gradient method. In IWCC 2001, LNCS 2326, pages 156--165, 2002.
[3]
Farmer I. Grinspun E. Schroder P. Bolz, J. Sparse matrix solvers on the gpu:conjugate gradients and multigrid. In ACM SIGGRAPH 2003 Papers, San Diego, California, pages 917--924, 2003.
[4]
C.T.Kelly. Iterative Methods for Linear and Nonlinear Equations. SIAM Books, 1995.
[5]
J.W Dammel. Applied Numerical Linear Algebra. SIAM Books, 1997.
[6]
Cunha D. Hopkins T. Dias, R. A parallel implementation of the restarted gmres iterative algorithm for nonsymmetric systems of linear equations. In Adv.Comp.Math 2, pages 261--277, 1994.
[7]
Galoppo, N. Govindaraju N. K, Henson M. Manocha D. Lu-gpu: Efficient algorithms for solving dense linear systems on graphics hardware. In Proceedings of the 2005 ACM/IEEE conference on Supercomputing., p. 3. IEEE Computer Society, Los Alamitos, 2005.
[8]
M. Garland. Sparse matrix computations on manycore gpu's. In Anaheim, California, ACM, pages 2--6, 2008.
[9]
Yiwei, Zh. Jian, S. Jijun, T. Roger, D. Method for partitioning large system models when using latency insertion method to speed network solution. In Proceedings of the 2009 Grand Challenges in Modeling Simulation (GCMS 09), book ISBN 1-56555-333-0, pages 259--263
[10]
Bergere G. Petiton S. He, H. Gmres method on lightweight grid system. In Proceedings of the 4th International Symposium on Parallel and Distributed Computing, pages 74--82, 2005.
[11]
Govindaraju N. K. Luo Q. Smith B. He, B. Efficient gather and scatter operations on graphics processors. In Proceedings of the 2007 ACM/IEEE conference on Supercomputing, Reno, Nevada ACM Press, pages 1--12, 2007.
[12]
Yang K. Fang R. Lu M. Govindaraju N. Luo Q. Sander P. He, B. Relational joins on graphics processors. In Vancouver, Canada ACM, pages 511--524, 2008.
[13]
M.R. Hestenes and E. Steifeel. Methods of conjugate gradient for solving linear systems. In J. of Res. nat. Bureau Standards, 49, pages 409--436, 1952.
[14]
E Jocelyne. A parallel gmres version for general sparse matrices. Electronic Transactions on Numerical Analysis, 3:160--176, 1995.
[15]
T. A. Manteuffel and S. Parter. Preconditioning and boundary conditions. In SIAM J. Numer. Anal., 27, pages 656--694, 1990.
[16]
Glanville R. S. Akeley K. Kilgard M. J. Mark, W. R. Cg: a system for programming graphics hardware in a c-like language. In ACM SIGGRAPH 2003 Papers, San Diego, California, ACM Press, pages 896--907, 2003.
[17]
Manish P. Hari S. Mingliang W., Hector K. Solving sparse linear systems on nvidia tesla gpus. Lecture Notes in Computer Science, 5544/2009:864--873, 2009.
[18]
S. C. Reddy N. M. Nachtigal and L. N. Trefethen. How fast are nonsymmetric matrix iterations. In SIAM J. Matrix Anal. Appl., 13, pages 778--795, 1992.
[19]
G. H. Golub P. Concus and G. Meurant. Block preconditioning for the conjugate gradient method. In SIAM J. Sci. Statist. Comput., 6, pages 220--252, 1984.
[20]
Hardy D. J. Stone J. E. Schulten K. Hwu W. M. W. Rodrigues, C. I. Gpu acceleration of cutoff pair potentials for molecular modeling applications. In Proceedings of the 2008 conference on Computing frontiers, Ischia, Italy ACM, pages 273--282, 2008.
[21]
R.J. Rost. OpenGL(R) Shading Language. Addison-Wesley Professional, Reading, 2004.
[22]
T. A. Manteuffel S. F. Ashby and J. S. Otto. A comparison of adaptive chebyshev and least squares polynomial preconditioning for hermetian positive definite linear systems. In SIAM J. Sci. Statist. Comput., 13, pages 1--29, 1992.
[23]
Y. Saad and M. Schultz. Gmres a generalized minimal residual algorithm for solving nonsymmetric linear systems. In SIAM J. Sci. Statist. Comput., 7, pages 856--869, 1986.
[24]
NVIDIA CUDA Team. NVIDIA CUDA C Programming Guide, 2010. Version 3.2.
[25]
Volkov, V James, D. Lu, qr and cholesky factorizations using vector capabilities of gpus. Technical Report No. UCB/EECS-2008-49, 2008.
[26]
Ward, C David, K. Numerical mathematics and computing, 6th Edition. Cengage Learning, 2007.
[27]
Bergere G. Petiton S. Zhang, Y. A parallel hybrid method of gmes on grid system. In Parallel and Distributed Processing Symposium, IPDPS 2007, IEEE international, pages 1--7, 2007.
[28]
John D. Owens, David Luebke, Naga Govindaraju, Mark Harris, Jens Kruger, Aaron E. Lefohn, and Timothy J. Purcel. A Survey of General-Purpose Computation on Graphics Hardware. In Computer Graphics Forum, 26(1), March 2007, pages 80--113, 2007.
- New basic linear algebra methods for simulation on GPUs
Recommendations
Sparse systems solving on GPUs with GMRES
Scientific applications very often rely on solving one or more linear systems. When matrices are sparse, iterative methods are preferred to direct ones. Nevertheless, the value of nonzero elements and their distribution (i.e., the sketch of the matrix) ...
Benchmarking GPUs to tune dense linear algebra
SC '08: Proceedings of the 2008 ACM/IEEE conference on SupercomputingWe present performance results for dense linear algebra using recent NVIDIA GPUs. Our matrix-matrix multiply routine (GEMM) runs up to 60% faster than the vendor's implementation and approaches the peak of hardware capabilities. Our LU, QR and Cholesky ...
Preconditioned Krylov solvers on GPUs
In this paper, we study the effect of enhancing GPU-accelerated Krylov solvers with preconditioners. We consider the BiCGSTAB, CGS, QMR, and IDR(s) Krylov solvers. For a large set of test matrices, we assess the impact of Jacobi and incomplete ...
Comments
Information & Contributors
Information
Published In
Sponsors
- SCS: Society for Modeling and Simulation International
In-Cooperation
Publisher
Society for Modeling & Simulation International
Vista, CA
Publication History
Published: 27 June 2011
Check for updates
Author Tags
Qualifiers
- Research-article
Conference
SummerSim '11
Sponsor:
- SCS
Contributors
Other Metrics
Bibliometrics & Citations
Bibliometrics
Article Metrics
- 0Total Citations
- 74Total Downloads
- Downloads (Last 12 months)6
- Downloads (Last 6 weeks)0
Reflects downloads up to 03 Oct 2024
Other Metrics
Citations
View Options
Get Access
Login options
Check if you have access through your login credentials or your institution to get full access on this article.
Sign in