Molecular Simulation

The Challenge
Researchers have been working to identify the thermodynamic properties of HOONO, a molecule that is crucial to understanding the dynamics of smog. They use Gaussian to calculate the molecule's energy in a variety of configurations, and then use MATLAB® to determine the vibrational wave function for the configurations based on quantum mechanics principles.

In order to see how the molecule would interact within a larger context such as a weather system, scientists must first obtain an accurate model of the thermodynamic properties. To achieve this, they must employ larger matrices in computing the molecule's properties - larger than their desktop systems' practical limit of several million elements. The lab's high-performance cluster would be an ideal platform for efficiently performing the calculations, but manually reprogramming the model for a parallel system would take months, putting a stop to ongoing research.

Thermodynamic Calculations for Molecules  with Asymmetric Internal Rotors - Wong et al

Star-P Solution
Using Star-P, researchers made minor modifications to the original MATLAB® algorithm, identifying the variables and calls that would benefit most from parallelization. Star-P's global array syntax enabled the server to hold entire data sets in memory, allowing for faster and more interactive algorithm development and analysis of the results.

Summary & Metrics   

• MATLAB®-based algorithm parallelized in seconds
• No need to learn and implement C and MPI
• Used a 16-processor SGI Altix server for calculations beyond the range of a serial desktop
• Powerful new insights into molecular behavior

"Our molecular models involve computations with 20,000 x 20,000 matrices, so we quickly reached the limits of desktop memory and computing power. We used Star-P to let us continue working interactively within our MATLAB® environment, extending our computations to larger and more complicated models."