A parallel implementation of the Car-Parrinello method by orbital decomposition

James Wiggs, Hannes Jónsson

Rannsóknarafurð: Framlag til fræðitímaritsGreinritrýni

20 Tilvitnanir (Scopus)

Útdráttur

We have implemented the first-principles molecular dynamics algorithm of Car and Parrinello to take advantage of the currently available and emerging class of parallel supercomputers. Computations involving different orbitals are largely carried out in parallel. The communications intensive orthonormalization has been implemented in a way that maintains equal load balancing and takes advantage of nearly the full communications bandwidth, even on computers with a low degree of connectivity, such as the mesh communications architecture. Performance statistics on the Touchstone Delta computer and comparison with the performance of an optimized code running on the Cray Y-MP are presented.

Upprunalegt tungumálEnska
Síður (frá-til)1-18
Síðufjöldi18
FræðitímaritComputer Physics Communications
Bindi81
Númer tölublaðs1-2
DOI
ÚtgáfustaðaÚtgefið - jún. 1994

Athugasemd

Funding Information:
This research has been supported by the Department of Energy's Computational Science Graduate Fellowship Program (JW) and by NSF-CARM grant CHE-9217294. Computer time was made available on the Delta by PNL-Battelle in Richland, WA, and on the CRAY Y-MPg/684 by the San Diego Supercomputer Center.

Fingerprint

Sökktu þér í rannsóknarefni „A parallel implementation of the Car-Parrinello method by orbital decomposition“. Saman myndar þetta einstakt fingrafar.

Vitna í þetta