BIT Numerical Mathematics

, Volume 51, Issue 4, pp 1009–1030 | Cite as

A GPU-based hyperbolic SVD algorithm

  • Vedran NovakovićEmail author
  • Sanja Singer


A one-sided Jacobi hyperbolic singular value decomposition (HSVD) algorithm, using a massively parallel graphics processing unit (GPU), is developed. The algorithm also serves as the final stage of solving a symmetric indefinite eigenvalue problem. Numerical testing demonstrates the gains in speed and accuracy over sequential and MPI-parallelized variants of similar Jacobi-type HSVD algorithms. Finally, possibilities of hybrid CPU–GPU parallelism are discussed.


One-sided Jacobi algorithm Hyperbolic singular value decomposition Symmetric indefinite eigenvalue problem GPU parallel programming 

Mathematics Subject Classification (2000)

65F15 65Y05 65Y10 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Brent, R.P., Luk, F.T.: The solution of singular–value and symmetric eigenvalue problems on multiprocessor arrays. SIAM J. Sci. Stat. Comput. 6(1), 69–84 (1985) MathSciNetzbMATHCrossRefGoogle Scholar
  2. 2.
    Bunch, J.R., Parlett, B.N.: Direct methods for solving symmetric indefinite systems of linear equations. SIAM J. Numer. Anal. 8(4), 639–655 (1971) MathSciNetCrossRefGoogle Scholar
  3. 3.
    Drmač, Z., Hari, V.: On quadratic convergence bounds for the J-symmetric Jacobi method. Numer. Math. 64(1), 147–180 (1993) MathSciNetzbMATHCrossRefGoogle Scholar
  4. 4.
    Hari, V., Veselić, K.: On Jacobi methods for singular value decompositions. SIAM J. Sci. Stat. Comput. 8(5), 741–754 (1987) zbMATHCrossRefGoogle Scholar
  5. 5.
    Hari, V., Singer, S., Singer, S.: Block-oriented J-Jacobi methods for Hermitian matrices. Linear Algebra Appl. 433(8–10), 1491–1512 (2010) MathSciNetzbMATHCrossRefGoogle Scholar
  6. 6.
    Hoberock, J., Bell, N.: Thrust: a parallel template library. Version 1.3.0 (2010). URL
  7. 7.
    IEEE: IEEE standard for floating–point arithmetic (2008). doi: 10.1109/IEEESTD.2008.4610935
  8. 8.
    The Khronos Group Inc.: The OpenCL specification 1.1 (2010) Google Scholar
  9. 9.
    Lahabar, S., Narayanan, P.J.: Singular value decomposition on GPU using CUDA. In: Proceedings of the 23rd IEEE International Symposium on Parallel & Distributed Processing (IPDPS 2009), Rome, Italy, May 23–29, 2009 Google Scholar
  10. 10.
    Luk, F.T., Park, H.: On parallel Jacobi orderings. SIAM J. Sci. Stat. Comput. 10(1), 18–26 (1987) MathSciNetGoogle Scholar
  11. 11.
    Luk, F.T., Park, H.: A proof of convergence for two parallel Jacobi SVD algorithms. IEEE Trans. Comput. C–38(6), 806–811 (1989) MathSciNetCrossRefGoogle Scholar
  12. 12.
    Mathias, R.: Analysis of algorithms for orthogonalizing products of unitary matrices. Numer. Linear Algebra Appl. 3(2), 125–145 (1996) MathSciNetzbMATHCrossRefGoogle Scholar
  13. 13.
    NVIDIA Corp.: NVIDIA CUDA C programming guide 3.1.1 (2010) Google Scholar
  14. 14.
    NVIDIA Corp.: PTX: parallel thread execution ISA version 2.1 (2010) Google Scholar
  15. 15.
    Sachdev, G.S., Vanjani, V., Hall, M.W.: Takagi factorization on GPU using CUDA. In: 2010 Symposium on Application Accelerators in High Performance Computing (SAAHPC’10), Knoxville, Tennessee, July 13–15, 2010. URL Google Scholar
  16. 16.
    Sameh, A.H.: On Jacobi and Jacobi-like algorithms for a parallel computer. Math. Comput. 25(118), 579–590 (1971) MathSciNetzbMATHGoogle Scholar
  17. 17.
    Singer, S.: Computing the eigendecomposition of a positive definite matrix. Master’s thesis, Dept. of Mathematics, University of Zagreb (1993) (in Croatian) Google Scholar
  18. 18.
    Singer, S.: Indefinite QR factorization. BIT 46(1), 141–161 (2006) MathSciNetzbMATHCrossRefGoogle Scholar
  19. 19.
    Singer, S., Singer, S., Hari, V., Bokulić, K., Davidović, D., Jurešić, M., Ušćumlić, A.: Advances in speedup of the indefinite one-sided block Jacobi method. In: Simos, T.E., Psihoyios, G., Tsitouras, C. (eds.) AIP Conf. Proc. Numerical Analysis and Applied Mathematics, vol. 936, pp. 519–522. AIP, New York (2007) CrossRefGoogle Scholar
  20. 20.
    Singer, S., Singer, S., Novaković, V., Davidović, D., Bokulić, K., Ušćumlić, A.: Three-level parallel J-Jacobi algorithms for Hermitian matrices. Technical report, University of Zagreb (2010). URL arXiv:1008.4166 [cs.NA]
  21. 21.
    Singer, S., Singer, S., Novaković, V., Ušćumlić, A., Dunjko, V.: Novel modifications of parallel Jacobi algorithms. Numer. Algorithms (2011). doi: 10.1007/s11075-011-9473-6 Google Scholar
  22. 22.
    Slapničar, I.: Componentwise analysis of direct factorization of real symmetric and Hermitian matrices. Linear Algebra Appl. 272, 227–275 (1998) MathSciNetzbMATHCrossRefGoogle Scholar
  23. 23.
    Slapničar, I.: Highly accurate symmetric eigenvalue decomposition and hyperbolic SVD. Linear Algebra Appl. 358, 387–424 (2003) MathSciNetzbMATHCrossRefGoogle Scholar
  24. 24.
    Slapničar, I., Veselić, K.: Perturbations of the eigenprojections of a factorized Hermitian matrix. Linear Algebra Appl. 218, 273–280 (1995) MathSciNetzbMATHCrossRefGoogle Scholar
  25. 25.
    Veselić, K.: A Jacobi eigenreduction algorithm for definite matrix pairs. Numer. Math. 64(1), 241–269 (1993) MathSciNetzbMATHCrossRefGoogle Scholar
  26. 26.
    Zha, H.: A note on the existence of the hyperbolic singular value decomposition. Linear Algebra Appl. 240, 199–205 (1996) MathSciNetzbMATHCrossRefGoogle Scholar
  27. 27.
    Zhang, S., Dou, H.: Matrix singular value decomposition based on computing unified device architecture. J. Appl. Res. Comput. 24(6), 4 (2007) (in Chinese) MathSciNetGoogle Scholar

Copyright information

© Springer Science + Business Media B.V. 2011

Authors and Affiliations

  1. 1.Faculty of Mechanical Engineering and Naval ArchitectureUniversity of ZagrebZagrebCroatia

Personalised recommendations