We show how both the tridiagonal and bidiagonal qr algorithms can be restructured so that they become rich in operations that can achieve near-peak performance on a modern processor. The key is a novel, cache-friendly...
详细信息
We show how both the tridiagonal and bidiagonal qr algorithms can be restructured so that they become rich in operations that can achieve near-peak performance on a modern processor. The key is a novel, cache-friendly algorithm for applying multiple sets of Givens rotations to the eigenvector/singular vector matrix. This algorithm is then implemented with optimizations that: (1) leverage vector instruction units to increase floating-point throughput, and (2) fuse multiple rotations to decrease the total number of memory operations. We demonstrate the merits of these new qr algorithms for computing the Hermitian eigenvalue decomposition (EVD) and singular value decomposition (SVD) of dense matrices when all eigenvectors/ singular vectors are computed. The approach yields vastly improved performance relative to traditional qr algorithms for these problems and is competitive with two commonly used alternatives-Cuppen's Divide-and-Conquer algorithm and the method of Multiple Relatively Robust Representations-while inheriting the more modest O(n) workspace requirements of the original qr algorithms. Since the computations performed by the restructured algorithms remain essentially identical to those performed by the original methods, robust numerical properties are preserved.
In the year 2000 the dominant method for solving matrix eigenvalue problems is still the qr algorithm. This paper discusses the family of GR algorithms, with emphasis on the qr algorithm. Included are historical remar...
详细信息
In the year 2000 the dominant method for solving matrix eigenvalue problems is still the qr algorithm. This paper discusses the family of GR algorithms, with emphasis on the qr algorithm. Included are historical remarks, an outline of what GR algorithms are and why they work, and descriptions of the latest, highly parallelizable, versions of the qr algorithm. Now that we know how to parallelize it, the qr algorithm seems likely to retain its dominance for many years to come. (C) 2000 Elsevier Science B.V. All rights reserved. MSC: 65F15.
The BR algorithm is a novel and efficient method to find all eigenvalues of upper Hessenberg matrices and has never been applied to eigenanalysis for power system small signal stability. This paper analyzes difference...
详细信息
The BR algorithm is a novel and efficient method to find all eigenvalues of upper Hessenberg matrices and has never been applied to eigenanalysis for power system small signal stability. This paper analyzes differences between the BR and the qr algorithms with performance comparison in terms of CPU time based on stopping criteria and storage requirement. The BR algorithm utilizes accelerating strategies to improve its performance when computing eigenvalues of narrowly banded, nearly tridiagonal upper Hessenberg matrices. These strategies significantly reduce the computation time at a reasonable level of precision. Compared with the qr algorithm, the BR algorithm requires fewer iteration steps and less storage space without depriving of appropriate precision in solving eigenvalue problems of large-scale power systems. Numerical examples demonstrate the efficiency of the BR algorithm in pursuing eigenanalysis tasks of 39-, 68-, 115-, 300-, and 600-bus systems. Experiment results suggest that the BR algorithm is a more efficient algorithm for large-scale power system small signal stability eigenanalysis.
One of the most interesting dynamical systems used in numerical analysis is the qr algorithm. An added maneuver to improve the convergence behavior is the qr iteration with shift which is of fundamental importance in ...
详细信息
One of the most interesting dynamical systems used in numerical analysis is the qr algorithm. An added maneuver to improve the convergence behavior is the qr iteration with shift which is of fundamental importance in eigenvalue computation. This paper is a theoretical study of the set of all isospectral matrices "reachable" by the dynamics of the qr algorithm with shift. A matrix B is said to be reachable by A if B = RQ + mu I, where A - mu I = qr is the qr decomposition for some mu is an element of R. It is proved that in general the qr algorithm with shift is neither reflexive nor symmetric. Examples are given to demonstrate that this relation is neither transitive nor antisymmetric. It is further discovered that the reachable set from a given n x n matrix A forms 2(n-1) disjoint open loops if n is even and 2(n-2) disjoint components each of which is no longer a loop when n is odd.
We propose a new qr-like algorithm, symmetric squared qr (SSqr) method, that can be readily parallelized using commonly available parallel computational primitives such as matrix-matrix multiplication and qr decomposi...
详细信息
We propose a new qr-like algorithm, symmetric squared qr (SSqr) method, that can be readily parallelized using commonly available parallel computational primitives such as matrix-matrix multiplication and qr decomposition. The algorithm converges quadratically and the quadratic convergence is achieved through a squaring technique without utilizing any kind of shifts. We provide a rigorous convergence analysis of SSqr and derive structures for several of the important quantities generated by the algorithm. We also discuss various practical implementation issues such as stopping criteria and deflation techniques. We demonstrate the convergence behavior of SSqr using several numerical examples. (c) 2005 Elsevier Inc. All rights reserved.
For an Hermitian matrix the qr transform is diagonally similar to two steps of the LR transforms. Even for non-Hermitian matrices the qr transform may be written in rational form.
For an Hermitian matrix the qr transform is diagonally similar to two steps of the LR transforms. Even for non-Hermitian matrices the qr transform may be written in rational form.
The qr algorithm and its variants are among the most popular methods for calculating eigenvalues of matrices. Typical implementations chase bulges from top to bottom of an upper Hessenberg matrix. It is also possible ...
详细信息
The qr algorithm and its variants are among the most popular methods for calculating eigenvalues of matrices. Typical implementations chase bulges from top to bottom of an upper Hessenberg matrix. It is also possible to chase bulges from bottom to top. There are some situations in which it may be advantageous to chase bulges in both directions at once, in which case one needs a procedure for passing bulges through each other without mixing up the information that the bulges convey. This paper derives a procedure for passing bulges of arbitrary degree through each other. Experiments with a Fortran 90 program show that the procedure works well in practice for bulges of degree two.
The full symmetric Toda system is a generalization of the open Toda chain, for which the Lax operator is a symmetric matrix of general form. This system is Liouville integrable and even superintegrable. Deift, Lee, Na...
详细信息
The full symmetric Toda system is a generalization of the open Toda chain, for which the Lax operator is a symmetric matrix of general form. This system is Liouville integrable and even superintegrable. Deift, Lee, Nando, and Tomei (DLNT) proposed the chopping method for constructing integrals of such a system. In the paper, a solution of Hamiltonian equations for the entire family of DLNT integrals is constructed by using the generalized qr factorization method. For this purpose, certain tensor operations on the space of Lax operators and special differential operators on the Lie algebra are introduced. Both tools can be interpreted in terms of the representation theory of the Lie algebra sl(n) and are expected to generalize to arbitrary real semisimple Lie algebras. As is known, the full Toda system can be interpreted in terms of a compact Lie group and a flag space. Hopefully, the results on the trajectories of this system obtained in the paper will be useful in studying the geometry of flag spaces.
In this paper, we propose and investigate numerical methods based on qr factorization for computing all or some Lyapunov or Sacker-Sell spectral intervals for linear differential-algebraic equations. Furthermore, a pe...
详细信息
In this paper, we propose and investigate numerical methods based on qr factorization for computing all or some Lyapunov or Sacker-Sell spectral intervals for linear differential-algebraic equations. Furthermore, a perturbation and error analysis for these methods is presented. We investigate how errors in the data and in the numerical integration affect the accuracy of the approximate spectral intervals. Although we need to integrate numerically some differential-algebraic systems on usually very long time-intervals, under certain assumptions, it is shown that the error of the computed spectral intervals can be controlled by the local error of numerical integration and the error in solving the algebraic constraint. Some numerical examples are presented to illustrate the theoretical results.
Parallel implementation of the qr algorithm for solving the symmetric eigenvalue problem requires more than a straightforward transcription of sequential code to parallel code. Experimental adjustments include new shi...
详细信息
Parallel implementation of the qr algorithm for solving the symmetric eigenvalue problem requires more than a straightforward transcription of sequential code to parallel code. Experimental adjustments include new shifting techniques and omission of the initial reducing to upper Hessenberg form. In this paper, the theory of the convergence of algorithms of decomposition type for the algebraic eigenvalue problem developed by Watkins and Elsner is generalized. The results are extended to deferred shifting schemes, which allow pipelining of iterations in parallel implementations, and analyzing the deterioration of the convergence rate. Furthermore, it is shown that eigenvalues need not be simple to obtain quadratic convergence for nondefective nonsymmetric matrices and cubic convergence for symmetric matrices.
暂无评论