Computational simulation plays a vital role in nanotechnology. Molecular dynamics (MD) is an important computational method to understand the fundamental behavior of nanoscale systems, and to transform that understanding into useful products. MD computations, however, are severely restricted by the spatial and temporal scales of simulations. This paper describes the methods used to achieve effective spatial parallelization of a MD code that is based on a multi-body bond order potential. The material system studied here is a carbon nanotube (CNT). We discuss the scientific and computational issues in the development and implementation of parallel algorithms, when the domain needs to be discretized with fine granularity. Specific issues in terms of neighbor-list computation, communication reduction, and cache awareness are delineated, with corresponding benefit in terms of speed up. Important practical problems relevant to CNT based composites are studied, and the effectiveness of various strategies reported. Our implementation achieves efficient parallelization at a finer granularity compared with published works on CNTs with complex configurations.
|Original language||English (US)|
|Number of pages||11|
|Journal||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|State||Published - 2004|
All Science Journal Classification (ASJC) codes
- Theoretical Computer Science
- Computer Science(all)