A performance study of the PLAPACK and ScaLAPACK Eigensolvers on HPCx for the standard problem
暂无分享,去创建一个
This report compares the performance of two packages available on HPCx, PLAPACK and ScaLAPACK, for the solution of the symmetric, standard eigenvalue problem. Both PLAPACK and ScaLAPACK provide parallel eigensolvers based on the QR algorithm for dense real symmetric matrices. These packages also provide newly developed algorithms: PLAPACK contains an eigensolver based upon the Multiple Relatively Robust Representations algorithm (MR) which has much reduced memory requirements as compared to other solvers and is therefore particularly suited for large matrices. ScaLAPACK v1.7 contains an eigensolver based on the Divide and Conquer (D&C) method (PDSYEVD). This is a Technical Report from the HPCx Consortium. Report available from http://www.hpcx.ac.uk/research/hpc/HPCxTR0406.pdf © HPCx UoE Ltd 2003 Neither HPCx UoE Ltd nor its members separately accept any responsibility for loss or damage arising from the use of information contained in any of their reports or in any communication about their tests or investigations. A performance study of the PLAPACK and ScaLAPACK Eigensolvers ii
[1] A. G. Sunderland,et al. An overview of eigensolvers for hpcx , 2003 .
[2] Robert A. van de Geijn,et al. A Parallel Eigensolver for Dense Symmetric Matrices Based on Multiple Relatively Robust Representations , 2005, SIAM J. Sci. Comput..
[3] Jack J. Dongarra,et al. A Parallel Divide and Conquer Algorithm for the Symmetric Eigenvalue Problem on Distributed Memory Architectures , 1999, SIAM J. Sci. Comput..
[4] J. Cuppen. A divide and conquer method for the symmetric tridiagonal eigenproblem , 1980 .