GPU implementation of Jacobi Method and Gauss-Seidel Method for Data Arrays that Exceed GPU-dedicated Memory Size



The paper proposes a method to extend the dimension of grids that GPU-aided implicit finite difference method is capable to work with. The approach is based on the pyramid method. A predictive mathematical model for computation duration is proposed. This model allows to find optimal algorithm parameters. The paper provides computation experiment results that has shown the model to be accurate enough to predict optimal algorithm parameters.


Pyramid method Finite-difference method Parallel computing GPU computing Jacobi method Red-black Gauss-Seidel method 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Golovashkin, D.L., Vorotnikova, D.G., Kochurov, A.V., Malysheva, S.A.: Solving finite-difference equations for diffractive optics problems using graphics processing units. Opt. Eng. 52 (9) (2013). doi: 10.1117/1.OE.52.9.091719
  2. 2.
    Krotkiewski, M.: Physics of Geological Processes (PGP), University of Oslo. Efficient solution of Poisson’s equation on modern GPUs using structured grids. Proceedings of NOTUR2011, Oslo (2011)Google Scholar
  3. 3.
    Micikevicius, P.: 3D Finite Difference Computation on GPUs using CUDA. In Proceedings of 2nd Workshop on General Purpose Processing on Graphics Processing Units (2009)Google Scholar
  4. 4.
    Pavelyev, V.S., Karpeev, S.V., Dyachenko, P.N., Miklyaev, Y.V.: Fabrication of three-dimensional photonics crystals by interference lithography with low light absorption. J. Modern Optics. 9 56, 1133–1136 (2009)CrossRefGoogle Scholar
  5. 5.
    Stefanski, T.P., Chavannes, N., Kuster, N.: Multi-GPU Accelerated Finite-difference Time-domain Solver in Open Computing Language. PIERS Online, 1 7, 71–74 (2011)Google Scholar
  6. 6.
    Micikevicius, P.: Multi-GPU Programming for Finite Difference Codes on Regular Grids Stanford AHPCRC/iCME Colloquium Series. Accessed 21 June 2014 (2012).
  7. 7.
    Jin, G., Endo, T., Matsuoka, S.: A parallel optimization method for stencil computation on the domain that is bigger than memory capacity of GPUs. Cluster Computing (CLUSTER), 2013 IEEE International Conference on. 1–8 (2013)Google Scholar
  8. 8.
    Golovashkin, D.L., Kochurov, A.V.: Pyramid method for GPU-aided finite difference method. Proceedings of CMMSE (2013)Google Scholar
  9. 9.
    Bethune, I., Bull, J.M., Dingle, N.J., Higham, N.J.: Investigating the Performance of Asynchronous Jacobis Method for Solving Systems of Linear Equations. University of Manchester EPrint (2011)Google Scholar
  10. 10.
    Anzt, H., Tomov, S., Dongarra, J., Heuveline, V.: A Block-Asynchronous Relaxation Method for Graphics Processing Units. J. Parallel Distrib. Comput. 73(12), 1613–1626 (2013)CrossRefGoogle Scholar
  11. 11.
    Samarskii, A.A., Vabishchevich, P.N.: Computational heat transfer: Volume 1, mathematical modelling, Wiley (1995)Google Scholar
  12. 12.
    Ortega, J.M.: Introduction to Parallel & Vector Solution of Linear Systems. Plenum Press, New York (1988)CrossRefMATHGoogle Scholar
  13. 13.
    Courtecuisse, H., Allard, J.: Parallel Dense Gauss-Seidel Algorithm on Many-Core Processors. High Performance Computing and Communications, 2009 HPCC ’09. 11th IEEE International Conference on (2013)Google Scholar
  14. 14.
    Bruaset, A.M., Tveito, A.: Numerical solution of partial differential equations on parallel computers. Springer, New York (2006)CrossRefMATHGoogle Scholar
  15. 15.
    EIMagbrbay, M., Ammar, R., Rajasekaran, R.: Fast GPU Algorithms for Implementing the Red-Black Gauss-Seidel Method for Solving Partial Differential Equations. Computers and Communications (ISCC), 2013 IEEE Symposium on (2013)Google Scholar

Copyright information

© Springer Science+Business Media Dordrecht 2015

Authors and Affiliations

  1. 1.Image Processing Systems Institute of Russian Academy of ScienceSamaraRussian
  2. 2.Samara State Aerospace UniversitySamaraRussian

Personalised recommendations