Parallelization of Kinetic Theory Simulations
Numerical studies of shock waves in large scale systems via kinetic simulations with millions of particles are too computationally demanding to be processed in serial. In this work we focus on optimizing the parallel performance of a kinetic Monte Carlo code for astrophysical simulations such as core-collapse supernovae. Our goal is to attain a flexible program that scales well with the architecture of modern supercomputers. This approach requires a hybrid model of programming that combines a message passing interface (MPI) with a multithreading model (OpenMP) in C++. We report on our approach to implement the hybrid design into the kinetic code and show first results which demonstrate a significant gain in performance when many processors are applied.
KeywordsMessage Passing Interface Ghost Cell Ghost Particle Simulation Space Collision Partner
The authors would like to thank the Blue Water Undergraduate Petascale Education Program and Shodor for their financial and educational support. Furthermore, this work used the Extreme Science and Engineering Discovery Environment (XSEDE), which is supported by National Science Foundation grant number OCI-1053575. I.S. is thankful to the Alexander von Humboldt foundation and acknowledges the support of the High Performance Computer Center and the Institute for Cyber-Enabled Research at Michigan State University.
- 3.I. Bouras et al., Phys. Rev. Lett. 103, 032301 (2009); I. Bouras et al., Phys. Rev. C 82, 024910 (2010); I. Bouras et al., Phys. Lett. B 710, 641 (2012)Google Scholar
- 28.J.R. Wilson, Numerical Astrophysics (Jones and Bartlett, Boston, 1985)Google Scholar
- 37.T. Bollenbach, MSU, M.S. Thesis (2002)Google Scholar
- 39.T. Strother, W. Bauer, Prog. Part. Nucl. Phys. 62, 468 (2009)Google Scholar
- 40.T. Strother, MSU, Ph.D. Thesis (2009)Google Scholar