On Aggressive Early Deflation in Parallel Variants of the QR Algorithm

The QR algorithm computes the Schur form of a matrix and is by far the most popular approach for solving dense nonsymmetric eigenvalue problems. Multishift and aggressive early deflation (AED) techniques have led to significantly more efficient sequential implementations of the QR algorithm during the last decade. More recently, these techniques have been incorporated in a novel parallel QR algorithm on hybrid distributed memory HPC systems. While leading to significant performance improvements, it has turned out that AED may become a computational bottleneck as the number of processors increases. In this paper, we discuss a two-level approach for performing AED in a parallel environment, where the lower level consists of a novel combination of AED with the pipelined QR algorithm implemented in the ScaLAPACK routine PDLAHQR. Numerical experiments demonstrate that this new implementation further improves the performance of the parallel QR algorithm.

[1]  W. Marsden.  I and J , 2012 .

[2]  David S. Watkins,et al.  Francis's Algorithm , 2011, Am. Math. Mon..

[3]  Jack J. Dongarra,et al.  A Parallel Implementation of the Nonsymmetric QR Algorithm for Distributed Memory Architectures , 2002, SIAM J. Sci. Comput..

[4]  Daniel Kressner,et al.  Parallel eigenvalue reordering in real Schur forms , 2009, Concurr. Comput. Pract. Exp..

[5]  Daniel Kressner,et al.  A Novel Parallel QR Algorithm for Hybrid Distributed Memory HPC Systems , 2010, SIAM J. Sci. Comput..

[6]  Gene H. Golub,et al.  The QR algorithm: 50 years later its genesis by John Francis and Vera Kublanovskaya and subsequent developments , 2009 .

[7]  Daniel B. Szyld,et al.  The matrix eigenvalue problem: GR and Krylov subspace methods , 2009, Math. Comput..

[8]  Jack Dongarra,et al.  ScaLAPACK Users' Guide , 1987 .

[9]  Daniel Kressner,et al.  Parallel Variants of the Multishift QZ Algorithm with Advanced Deflation Techniques , 2006, PARA.

[10]  Karen S. Braman,et al.  The Multishift QR Algorithm. Part I: Maintaining Well-Focused Shifts and Level 3 Performance , 2001, SIAM J. Matrix Anal. Appl..

[11]  Daniel Kressner,et al.  The Effect of Aggressive Early Deflation on the Convergence of the QR Algorithm , 2008, SIAM J. Matrix Anal. Appl..

[12]  R. W. Hamming.  State of the art in scientific computing , 1963, AFIPS '63 (Spring).

[13]  Robert A. van de Geijn,et al.  Parallelizing the QR Algorithm for the Unsymmetric Algebraic Eigenvalue Problem: Myths and Reality , 1996, SIAM J. Sci. Comput..

[14]  Daniel Kressner,et al.  Parallel eigenvalue reordering in real Schur forms , 2009 .

[15]  Karen S. Braman,et al.  The Multishift QR Algorithm. Part II: Aggressive Early Deflation , 2001, SIAM J. Matrix Anal. Appl..

[16]  James Demmel,et al.  On a Block Implementation of Hessenberg Multishift QR Iteration , 1989, Int. J. High Speed Comput..

[17]  Daniel Kressner,et al.  Numerical Methods for General and Structured Eigenvalue Problems , 2005, Lecture Notes in Computational Science and Engineering.

[18]  J. Demmel,et al.  On swapping diagonal blocks in real Schur form , 1993 .

[19]  Ralph Byers,et al.  Lapack 3 . 1 xHSEQR : Tuning and Implementation Notes on the Small Bulge Multi-shift QR Algorithm with Aggressive Early Deflation , 2007 .

[20]  Jack Dongarra,et al.  LAPACK Users' Guide, 3rd ed. , 1999 .