Performance of a fully parallel dense real symmetric eigensolver in quantum chemistry applications

Conference · Sat Apr 01 04:00:00 EST 1995

OSTI ID:80385

Fann, G I; Littlefield, R J; Elwood, D M

The parallel performance of a dense, standard and generalized, real, symmetric eigensolver based on bisection for eigenvalues and repeated inverse iteration and reorthogonalization for eigenvectors is described. The performance of this solver, called PeIGS, is given for two test problems and for three ``real-world`` quantum chemistry applications: SCF-Hartree-Fock, density functional theory,and Moeller-Plesset theory. The distinguishing feature of the repeated inverse iteration and orthogonalization method used by PEIGS is that orthogonalization may be performed across multiple processors as dictated by the spectrum. For each problem we describe the spectrum and the clustering of the eigenvalues, the most important factor in determining the execution time. For a spectrum that is well spaced, there is essentially no orthogonalization time. Most of the time is consumed in the Householder reduction to tridiagonal form. For large clusters, almost all of the time is consumed in the Householder reduction and in orthogonalization. Performance results from the Intel Paragon, and Kendall Square Research KSR-2 are reported.

• Research Organization: Pacific Northwest Lab., Richland, WA (United States)
• Sponsoring Organization: USDOE, Washington, DC (United States)
• DOE Contract Number: AC06-76RL01830
• OSTI ID: 80385
• Report Number(s): PNL-SA--25658; CONF-950439--11; ON: DE95011496
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
ALGORITHMS
CHEMISTRY
EIGENVALUES
EIGENVECTORS
ITERATIVE METHODS
MATRICES
PARALLEL PROCESSING
PERFORMANCE