# Performance Comparisons of Eigensolutions Algorithms of a Symmetric Tridiagonal Matrix on GPU Accelerators

## Abstract

While finding the solutions of Hermitian matrix eigenproblem it is a key issue to find an efficient version of algorithms of symmetric tridiagonal solutions. In this paper these algorithms are compared for complex Hermitian matrices in hybrid systems. The methods were carried out on the Tesla C1060 and Tesla K40 GPU accelerators and the performances are presented as between the methods, as well as between the accelerators.

## References

“MAGMA Matrix Algebra on GPU and Multicore Architectures”, [Online]. Available:http://icl.cs.utk.edu/magma/, 2014.

E. Agullo, J. Demmel, J. Dongarra, B. Hadri, J. Kurzak, J. Langou, H. Ltaief, P. Luszczek, S. Tomov, “Numerical linear algebra on emerging architectures: The PLASMA and MAGMA projects ”, Journal of Physics: Conference Series, vol. 180, no. 1, 2009.

B. N. Parlett, The Symmetric Eigenvalue Problem. Englewood Cliffs, NJ: PrenticeHall, 1980.

G. H.Golub, C. F. V.Loan, Matrix Computations, 3rd ed. Baltimore: The Johns Hopkins University Press, 1996.

I. C. Ipsen, “Computing an eigenvector with inverse iteration”, SIAM review, vol. 39, no. 2, pp. 254–291, 1997.

I. S. Dhillon, “Current inverse iteration software can fail”, BIT Numerical Mathematics, vol. 38, no. 4, pp. 685–704, 1998.

J. J. M.Cuppen, “A divide and conquer method for the symmetric tridiagonaleigenproblem”, NumerischeMathematik, vol. 36, no. 2, pp. 177–195, 1980.

M. Gu and S. C. Eisenstat, “A divide-and-conquer algorithm for the symmetric tridiagonaleigenproblem”, SIAM Journal on Matrix Analysis and Applications, vol. 16,no. 1, pp. 172–191,1995.

I. S. Dhillon,B. N. Parlett, C. Vomel, “The design and Implementation of the MRRRalgorithm”, ACM Trans. Math. Soft.,vol. 32, no. 4, pp. 533–560, 2006.

I. S. Dhillon and B. N.Parlett, “Multiple representations to compute orthogonaleigenvectors of symmetric tridiagonalmatrices”, Linear Algebra and its Applications,vol. 387, pp. 1–28, 2004.

B. N. Parlett and I. S. Dhillon, “Relatively robust representations of symmetrictridiagonals”, Linear Algebra and its Applications, vol. 309, no. 1–3, pp.121–151, 2000.

J. W. Demmel, Applied Numerical Linear Algebra, SIAM, Philadelphia, 1997.

I.Dhillon, and B. Parlett,“Multiple representations to compute orthogonal eigenvectors of symmetric tridiagonal matrices”,Linear Algebra Appl., vol. 387, pp. 1– 28. 2004.

I. Dhillonand B. Parlett,“Orthogonal eigenvectors and relative gaps”,SIAM J. Matrix Anal. Appl., vol.25,pp. 858–899, 2004.

P. Willems, On MR3-type Algorithms for the Tridiagonal Symmetric Eigenproblem and Bidiagonal SVD. PhD thesis, University of Wuppertal, 2010.

P. Willems and B. Lang, A Framework for the MR3 Algorithm: Theory and Implementation. Tech. Rep. 11/21, BergischeUniversität Wuppertal, 2011.

I. Dhillon, B. Parlettand C. Vömel,“Glued matrices and the MRRR algorithm”, SIAM J. Sci. Comput., vol. 27, no. 2, pp. 496–510, 2005.

I. Dhillon, B. Parlett and C. Vömel, “The design and implementation of the MRRR algorithm”,ACM Trans. Math. Software, vol. 32, pp. 533–560, 2006.

O. Marques, B. Parlettand C. Vömel, “Computations of eigenpairsubsets with the MRRR algorithm”,Numerical Linear Algebra with Applications, vol. 13, no. 8,pp. 643–653, 2006.

O. Marques, E. Riedy and C. Vömel, “Benefits of IEEE-754 features in modern symmetric tridiagonaleigensolvers”,SIAM J. Sci. Comput., vol. 28, pp. 1613–1633, 2006.

M. Petschow and P. Bientinesi, “MR3-SMP: A symmetric tridiagonaleigensolver for multi-core architectures”,Parallel Computing, vol. 37, no. 12, pp. 795 – 805, 2011.

C. Vömel, “ScaLAPACK’s MRRR Algorithm”, ACM Trans. Math. Software, vol. 37, pp 1 , 2010.

D. Watkins, Fundamentals of Matrix Computations. Pure and applied mathematics. John Wiley & Sons, 2010.

I. S. Dhillon, B. N. Parlett, and C. Vömel, “Glued matrices and the MRRR algorithm”,SIAM J. Sci. Comput., vol27, pp. 496–510, 2005.

I. S. Dhillon, “Current inverse iteration software can fail”, BIT Numerical Mathematics, vol. 38 pp. 685–704, 1998.

## Downloads

## Published

## How to Cite

*Mathematical Problems of Computer Science*,

*44*, 93–100. Retrieved from http://mpcs.sci.am/index.php/mpcs/article/view/188

## Issue

## Section

## License

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.