Advertisement

An Implementation of Parallel Eigenvalue Computation Using Dual-Level Hybrid Parallelism

  • Yonghua Zhao
  • Xuebin Chi
  • Qiang Cheng
Part of the Lecture Notes in Computer Science book series (LNCS, volume 4494)

Abstract

This paper describes a hybrid two-level parallel method with MPI/OpenMP for computing the eigenvalues of dense symmetric matrices on cluster of SMP’s environments. The eigenvalue computation is Based on both the Householder tridiagonalization method and a divide-and-conquer algorithm of tridiagonal eigenproblem. In hybrid parallel design, We take a coarse-grain approach to OpenMP shared-memory parallelization, which keeps BLAS-3 operations in tridiagonalization. Moreover, dynamic work sharing is used in the divide-and-conquer algorithm of tridiagonal eigenproblem. So the amount of synchronization has also been reduced, and these could have an effect on the load balance. In addition, we analyze the communication overhead between hybrid MPI/ OpenMP and pure MPI. An experimental analysis on the Deepcomp6800 shows the hybrid algorithm performs good scalability.

Keywords

MPI/OpenMP hybrid parallel algorithm parallel solver matrix eigenvalue 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Bova, S.W., Breshears, C., Cuicchi, C., Demirbilek, Z., Gabb, H.A.: Dual-level parallel analysis of harbor wave response using MPI and OpenMP. Int. J. Perform Comput. Appl. 14, 49–64 (2000)CrossRefGoogle Scholar
  2. 2.
    Rabense, R., Wellein, G.: Communication and optimization aspects of parallel programming models on hybrid architecture. International Journal of High performance Computing Applications 17(1), 49–62 (2003)CrossRefGoogle Scholar
  3. 3.
    Dong, S., Karniadakis, G.E.: Dual_level parallelism for high-order CFD methods. Parallell Computing 30, 1–20 (2004)CrossRefGoogle Scholar
  4. 4.
    Cappello, F., Etiemble, D.: MPI versus MPI+OpenMP on IBM SP for the NAS benchmarks. In: Proceedings of the 2000 ACM/IEEE conference on Supercomputing, Dallas, Texas, USA (2000)Google Scholar
  5. 5.
    Blackford, L.S., Choi, J., Cleary, A., D’Azevedo, E., Demmel, J., Dhillon, I.S., et al.: ScaLAPACK Users’s Guide. SIAM, USA (1997)Google Scholar
  6. 6.
    Luong, P., Breshears, C.P., Ly, L.N.: Coastal ocean modeling of the U.S. west coast with multiblock grid and dual-level parallelism. In: Supercomputing 2001: High Performance Networking and Computing (SC2001)Google Scholar
  7. 7.
    Loft, R.D., Thomas, S.J., Dennis, J.M.: Terascale spectral element dynamical core for atmospheric general circulation models. Supercomputing 2001: High Performance Networking and Computing (SC2001)Google Scholar
  8. 8.
    Hendrickson, B., Jessup, E., Smith, C.: Toward an efficent parallel eigensolver for dense symmetric matrices. SIAM J.Sci. Compute. 20(3), 1132–1154 (1999)zbMATHCrossRefMathSciNetGoogle Scholar
  9. 9.
    Demmel, J.: In: Applied Numerical Linear Algebra. SIAM, USA (1997)Google Scholar
  10. 10.
    Zhao, Y., Chen, J., Chi, X.B.: Solving the symmetric tridiagonal eigenproblem using MPI/OpenMP hybrid parallelization. In: Cao, J., Nejdl, W., Xu, M. (eds.) APPT 2005. LNCS, vol. 3756, pp. 164–173. Springer, Heidelberg (2005)CrossRefGoogle Scholar
  11. 11.
    Cuppen, J.J.M.: A divide and conquer method for symmetric tridiagonal eigenproblem. Numer. Mathematik. 2(36), 177–195 (1981)MathSciNetGoogle Scholar
  12. 12.
    Sameh, A., Kuck, D.: A parallel QR algorithm for symmetric tridiagonal matrices. IEEE tans. Comput. C-26, 81–91 (1977)Google Scholar
  13. 13.
    LI, T.Y., Zeng, Z.: Lagurre’s iteration in solving the symmetric tridiagonal eigenproblem. SIAM J. Scientific Comput. 15, 1145–1173 (1994)zbMATHCrossRefMathSciNetGoogle Scholar
  14. 14.
    Gu, M., Eisenstat, S.C.: A divide-and conquer algorithm for symmetric tridiagonal eigenproblem. SIAM J. Mat. Anal. Appl. 16(1), 172–191 (1995)zbMATHCrossRefMathSciNetGoogle Scholar
  15. 15.
    Dhillon, I.S., Parlett, B.N.: Multiple representations compute orthogonal eigenvertors of symmetric tridiagonal matrices. Lin. Alg. Appl. 387, 1–28 (2004)zbMATHCrossRefMathSciNetGoogle Scholar
  16. 16.
  17. 17.
    Philadelphia, 1997 Bischof, C. Gerorge, W. Huss-Lederman, S. Sun, X., Tsao, A., Turnbull, T.: Prism software (1997), http://www.mcs.anl.gov/Projects/PRISM/lib/software.html
  18. 18.
    Alpatov, P., Baker, G., Edwards, C., Gunnels, J., Morrow, G., Overfelt, J., van de Gejin, R., Wu, Y.-J.J.: PLAPACK: Parallel linear algebra package - design overview. In: Processing of SC 1997 (1997)Google Scholar
  19. 19.
    Fann, G., Littlefield, R.: Performance of a fully parallel dense real symmetric eigensolver in quantum chemistry applications. In: Proceedings of the Sixth SIAM Conference on Parallel Processing for Scientific Computation. SIAM (1994)Google Scholar
  20. 20.
    Treffz, C., Huanf, C.C.: A scalable eigenvalue solver for symmetric tridiagonal matrices. Parallel Computing 21, 1213–1240 (1995)CrossRefMathSciNetGoogle Scholar
  21. 21.
    Liu, W., Zheng, W.M., Zheng, X.W.: The concept of node-oriented speedup on SMP cluster. Computer engineering and design, vol. 21(5) (October 2000)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2007

Authors and Affiliations

  • Yonghua Zhao
    • 1
    • 2
    • 3
  • Xuebin Chi
    • 1
  • Qiang Cheng
    • 1
  1. 1.Supercomputing Center, Computer Network Information Center, Chinese Academy of Sciences, 100080, BeijingChina
  2. 2.Institute of Software, Chinese Academy of Sciences, 100080, BeijingChina
  3. 3.Department of Computer Science, Dezhou University, 253000, DezhouChina

Personalised recommendations