PHYS 234 |
| In the last two years a new class of very high-end computational machines has emerged to take over the lead in total system performance. Machines such as the IBM Bluegene/L and Cray XT3 systems offer applications new levels of performance, but also significant new challenges. Many applications are already taking good advantage of these systems. However, the requirements for extreme scalability and the restrictions imposed by the very limited compute node operating system are posing significant challenges for other applications. The GAMESS computational chemistry application illustrates both of these extremes. While some methods have scaled very well on Bluegene, there are significant challenges remaining. This paper will discuss the current GAMESS implementation on Bluegene, ongoing development efforts and planned future work aimed at improving overall application throughput on ultra-scale machines. |
|
Cyber Science, Chemistry
1:20 PM-5:20 PM, Tuesday, 12 September 2006 Grand Hyatt San Francisco -- San Miguel, Oral
Division of Physical Chemistry |