 |
|
Chemistry: Gaussian '98The initial release of Gaussian 98 included parallelized versions of its Hartree-Fock and Density Functional Theory methods. Revision C of Gaussian 98 allows calculations including electron correlation using second-order Møller-Plesset perturbation theory (MP2) to be performed in parallel as well. Enabled via SCIENTIFIC's Linda® software, these Gaussian 98 features allow users to investigate new, more complex problems and to model even larger molecular systems with greater accuracy and efficiency. The program delivers these benefits to groups ranging from small departmental research teams to the largest national supercomputing centers. Performance GraphsThe following graph shows the parallel performance gains with respect
to running sequentially on a single processor for various numbers of Ethernet-connected
UNIX workstations. The diagonal line traversing the graph represents perfect
parallel speedup, which would scale exactly with the number of nodes. As
the graph indicates, Gaussian 98 performs very well on both calculations.
The solid line indicates the speedup when predicting the energy of guanine
(175 basis functions), a nucleic acid which is one of the components of
DNA, using DFT methods. The dashed line gives the speedup for a Hartree-Fock
calculation on bicyclopentanone (240 basis functions), a molecule which
is interesting for its highly-strained structure.
The following graph illustrates similar performance data for an MP2 single
point energy and force calculation on C20H42F (264 basis functions), running
on various numbers of processors on an IBM SP2 multiprocessor system. Once
again, Gaussian 98 performs well in parallel on this calculation, showing
excellent speedups across the entire range of number of processors applied
to the problem.
Gaussian, Inc. |
|
||||||||
Chemistry