T2: Gaussian 03

Time: 10:00-17:00
Location:
Morning session: B-huset, entrance 23, room BL33, Linköping University.
Afternoon session: B-huset, entrance 27, , rooms SU10 and SU11, Linköping University.


Speaker:
Dr Roberto Gomperts,
Silicon Graphics Inc.,
Hudson, MA, U.S.A.

Purpose and Topics:
NSC are glad to present an expert on the Gaussian program at LCSC 2006 which thereby continues the line of application software presentations that we have had in the past. Gaussian is the predominant software by measure of allocated time at the Swedish HPC centra, and in-depth information concerned with efficient use of the program is highly relevant from both academic and economic view points.

The tutorial will be focused at algorithms and performance in order to get the most out of the program. Dr R. Gomperts has a long experience from the optimization of kernel routines in Gaussian and will share his experiences at this tutorial.

The morning session will be given in the form of a lecture and the afternoon session will be held in a terminal room with a possibility for hands-on use of the program. Participants are welcomed to bring their own examples from their research and discuss these with Roberto, but discussions must be focused at the more technical aspects and not the chemical.

Abstract:
Effective use of Gaussian 03 on the SGI-Altix

After a short architectural introduction of the SGI Altix Computer system some aspects of the structure of the Gaussian software will be discussed, in particular memory layout and memory utilization. Using Amdahl's law as a basis, efficiency considerations for parallel execution will be presented. A case study of "growing dendrimers" will be discussed to illustrate efficient uses of the computer system. Recent experiments using Linda on SGI's Altix XE (cluster) platform will show the differences between the distributed memory parallel programming model and the shared memory parallel programming model in Gaussian 03. Miscellaneous examples of peculiarities of some post-SCF algorithms (MP2, MP4) will be shown.

Prerequisites:
A basic knowledge of Gaussian is required.


info@nsc.liu.se