Skip to main content

(EMSL21) Computational Methods for Atmospheric Chemistry


EMSL Project ID
1961

Abstract

With the advent and ease of accessibility to massively parallel computing architecture has come a corresponding need for programming techniques tailored to Eulerian air chemistry models. These problems focus on how to optimize the use of hundreds of processors on a single system relative to the thousands of nodes and multitude of physical processes used in modern air chemistry models (Peters et al., 1995). Support for this proposal will allow us to complete a task that, to date, has taken a model that once took several weeks per simulation and modified it to run on a massively parallel system so that the same simulations can now be done in 2 hours. The modifications involved reformulating calculations of local chemistry, transport and data exchange and were carried out under the constraints of maintaining code portability to other machines. However, additional testing is necessary before completion of this work. CPU cycles for this testing form the basis of our request for sup! port. In this 'Description of Proposed Project', we give an overview of work done to date, followed by a description of how we would use additional cycles on the EMSL system. We conclude with a brief overview of the physics being simulated.

Project Details

Project type
Capability Research
Start Date
2000-06-02
End Date
2001-10-01
Status
Closed

Team

Principal Investigator

Carl Berkowitz
Institution
Pacific Northwest National Laboratory

Team Members

Donald Jones
Institution
Pacific Northwest National Laboratory

Rahul Zaveri
Institution
Pacific Northwest National Laboratory

Richard Easter
Institution
Pacific Northwest National Laboratory

Elaine Chapman
Institution
Pacific Northwest National Laboratory