Advertisement

Preconditioned Jacobi SVD Algorithm Outperforms PDGESVD

  • Martin Bečka
  • Gabriel OkšaEmail author
Conference paper
  • 112 Downloads
Part of the Lecture Notes in Computer Science book series (LNCS, volume 12043)

Abstract

Recently, we have introduced a new preconditioner for the one-sided block-Jacobi SVD algorithm. In the serial case it outperformed the simple driver routine DGESVD from LAPACK. In this contribution, we provide the numerical analysis of applying the preconditioner in finite arithmetic and compare the performance of our parallel preconditioned algorithm with the procedure PDGESVD, the ScaLAPACK counterpart of DGESVD. Our Jacobi based routine remains faster also in the parallel case, especially for well-conditioned matrices.

Keywords

Singular value decomposition Parallel computation Dynamic ordering One-sided block-Jacobi algorithm Preconditioning 

Notes

Acknowledgment

Authors were supported by the VEGA grant no. 2/0004/17.

References

  1. 1.
    Anderson, A., et al.: LAPACK Users’ Guide, 3rd edn. SIAM, Philadelphia (1999)CrossRefGoogle Scholar
  2. 2.
    Bečka, M., Okša, G., Vidličková, E.: New preconditioning for the one-sided block-Jacobi SVD algorithm. In: Wyrzykowski, R., Dongarra, J., Deelman, E., Karczewski, K. (eds.) PPAM 2017. LNCS, vol. 10777, pp. 590–599. Springer, Cham (2018).  https://doi.org/10.1007/978-3-319-78024-5_51CrossRefGoogle Scholar
  3. 3.
    Bečka, M., Okša, G., Vajteršic, M.: Dynamic ordering for a parallel block Jacobi SVD algorithm. Parallel Comput. 28, 243–262 (2002).  https://doi.org/10.1016/S0167-8191(01)00138-7CrossRefzbMATHGoogle Scholar
  4. 4.
    Bečka, M., Okša, G., Vajteršic, M.: New dynamic orderings for the parallel one-sided block-Jacobi SVD algorithm. Parallel Proc. Lett. 25, 1–19 (2015).  https://doi.org/10.1142/S0129626415500036MathSciNetCrossRefzbMATHGoogle Scholar
  5. 5.
    Bečka, M., Okša, G.: Parallel one-sided Jacobi SVD algorithm with variable blocking factor. In: Wyrzykowski, R., Dongarra, J., Karczewski, K., Waśniewski, J. (eds.) PPAM 2013. LNCS, vol. 8384, pp. 57–66. Springer, Heidelberg (2014).  https://doi.org/10.1007/978-3-642-55224-3_6CrossRefGoogle Scholar
  6. 6.
    Dongarra, J., et al.: The singular value decomposition: anatomy of optimizing an algorithm for extreme scale. SIAM Rev. 60, 808–865 (2018).  https://doi.org/10.1137/17M1117732MathSciNetCrossRefzbMATHGoogle Scholar
  7. 7.
    Golub, G.H., van Loan, C.F.: Matrix Computations, 4th edn. The John Hopkins University Press, Baltimore (2013)zbMATHGoogle Scholar
  8. 8.
    Higham, N.J.: Accuracy and Stability of Numerical Algorithms, 2nd edn. SIAM, Philadelphia (2002)CrossRefGoogle Scholar
  9. 9.
    Jia, Z.: Using cross-product matrices to compute the SVD. Numer. Algorithms 42, 31–61 (2006).  https://doi.org/10.1007/s11075-006-9022-xMathSciNetCrossRefzbMATHGoogle Scholar
  10. 10.
    Kudo, S., Yamamoto, Y., Bečka, M., Vajteršic, M.: Performance analysis and optimization of the parallel one-sided block-Jacobi algorithm with dynamic ordering and variable blocking. Concurr. Comput. Pract. Exp. 29, 1–24 (2017).  https://doi.org/10.1002/cpe.4059CrossRefGoogle Scholar
  11. 11.
    Nakatsukasa, Y., Higham, N.J.: Stable and efficient spectral divide and conquer algorithms for the symmetric eigenvalue decomposition and the SVD. SIAM J. Sci. Comput. 35, 1325–1349 (2013).  https://doi.org/10.1137/120876605MathSciNetCrossRefzbMATHGoogle Scholar
  12. 12.
    Okša, G., Yamamoto, Y., Vajteršic, M.: Asymptotic quadratic convergence of the block-Jacobi EVD algorithm for Hermitian matrices. Numer. Math. 136, 1071–1095 (2017).  https://doi.org/10.1007/s00211-016-0863-5MathSciNetCrossRefzbMATHGoogle Scholar
  13. 13.

Copyright information

© Springer Nature Switzerland AG 2020

Authors and Affiliations

  1. 1.Institute of MathematicsSlovak Academy of SciencesBratislavaSlovak Republic

Personalised recommendations