GPU Acceleration of the Generalized Interpolation Material Point Method
School of Computing, University of Utah, Salt Lake City, UT 84112
Symposium on Application Accelerators in High Performance Computing, 2009 (SAAHPC’09)
@article{chiang2009gpu,
title={GPU Acceleration of the Generalized Interpolation Material Point Method},
author={Chiang, Wei-Fan and DeLisi, Michael and Hummel, Todd and Prete, Tyler and Tew, Kevin and Hall, Mary and Wallstedt, Phil and Guilkey, James},
booktitle={Symposium on Application Accelerators in High Performance Computing, SAAHPC},
year={2009}
}
This paper describes our experience rewriting a sequential particle-in-cell code so that its key computations are executed on a GPU. This code is well-suited to GPU acceleration, as it performs data-parallel operations on a regular grid. Key performance challenges are the need for global synchronization in mapping particles to grid nodes, and managing memory bandwidth to global memory. Performance results show overall speedups of 3.3x including the time to display the results of simulation, or 10.9x without the display I/O overhead.
February 22, 2011 by hgpu
