N-Body Gravitational Simulations

Lead Author Major

Computer Science

Format

SOECS Senior Project Demonstration

Faculty Mentor Name

Jinzhu Gao

Faculty Mentor Department

Computer Science

Additional Faculty Mentor Name

Keisuke Juge

Additional Faculty Mentor Name

James Hetrick

Abstract/Artist Statement

N-Body simulations are a common method of modeling the interacting behavior between objects from the atomic scale to the astronomical scale. This simulation models the effects of the Newtonian gravitational forces between objects on a galactic scale. Simulations of this kind are important in understanding physical systems which cannot be solved analytically. In order to model a realistic system using an n-body simulation usually many bodies must be simulated which increases the computation time significantly. My project was designed around finding design methods of both software and hardware which could be used to increase the speed at which a large n-body system could be run. These methods included a parallel cluster system, a GPU architecture, a parallel application written in MPI and CUDA and a clustering algorithm design. The application will run in in a scalable format which will allow it to be expanded to run on larger clusters than the one I have put together. The simulated runs will be visualized. The architecture and software hold the potential for galaxy-scale simulations as well as galaxy collision simulations if it can be run on a sufficiently large machine.

Location

School of Engineering & Computer Science

Start Date

27-4-2013 2:00 PM

End Date

27-4-2013 3:30 PM

This document is currently not available here.

Share

COinS
 
Apr 27th, 2:00 PM Apr 27th, 3:30 PM

N-Body Gravitational Simulations

School of Engineering & Computer Science

N-Body simulations are a common method of modeling the interacting behavior between objects from the atomic scale to the astronomical scale. This simulation models the effects of the Newtonian gravitational forces between objects on a galactic scale. Simulations of this kind are important in understanding physical systems which cannot be solved analytically. In order to model a realistic system using an n-body simulation usually many bodies must be simulated which increases the computation time significantly. My project was designed around finding design methods of both software and hardware which could be used to increase the speed at which a large n-body system could be run. These methods included a parallel cluster system, a GPU architecture, a parallel application written in MPI and CUDA and a clustering algorithm design. The application will run in in a scalable format which will allow it to be expanded to run on larger clusters than the one I have put together. The simulated runs will be visualized. The architecture and software hold the potential for galaxy-scale simulations as well as galaxy collision simulations if it can be run on a sufficiently large machine.