Local Møller-Plesset Perturbation Theory:  A Massively Parallel Algorithm

A massively parallel algorithm is presented for computation of energies with local second-order Møller-Plesset (LMP2) perturbation theory. Both the storage requirement and the computational time scale linearly with the molecular size. The parallel algorithm is designed to be scalable, employing a di...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of chemical theory and computation Jg. 3; H. 1; S. 71
Hauptverfasser: Nielsen, Ida M B, Janssen, Curtis L
Format: Journal Article
Sprache:Englisch
Veröffentlicht: United States 01.01.2007
ISSN:1549-9618
Online-Zugang:Weitere Angaben
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:A massively parallel algorithm is presented for computation of energies with local second-order Møller-Plesset (LMP2) perturbation theory. Both the storage requirement and the computational time scale linearly with the molecular size. The parallel algorithm is designed to be scalable, employing a distributed data scheme for the two-electron integrals, avoiding communication bottlenecks, and distributing tasks in all computationally significant steps. A sparse data representation and a set of generalized contraction routines have been developed to allow efficient massively parallel implementation using distributed sparse multidimensional arrays. High parallel efficiency of the algorithm is demonstrated for applications employing up to 100 processors.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISSN:1549-9618
DOI:10.1021/ct600188k