Abstract
A cluster-based spatial decomposition algorithm for solving large-scale Molecular Dynamics simulation of thermophysics is proposed. Firstly, three kinds of domain division strategies are provided and their efficiency and scalability are analyzed. Secondly, a method called FLNP (Fast Location of Neighboring Particles) to accelerate the location of neighboring particles is proposed, which greatly reduces the cost of calculation and communication of interaction. Additionally, a new memory management technique called AMM (Adaptive Memory Management) is applied to meet the large memory requirement. The parallel algorithm based on these above technologies was implemented on a cluster of SMPs and tested on a system of 6,912,000 particles and achieved an efficiency of 77.0%. © Springer-Verlag Berlin Heidelberg 2004.
Cite
CITATION STYLE
Shu, J., Wang, B., & Zheng, W. (2004). Cluster-based parallel simulation for large scale molecular dynamics in microscale thermophysics. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 3358, 200–211. https://doi.org/10.1007/978-3-540-30566-8_25
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
