| Issue |
2014
SNA + MC 2013 - Joint International Conference on Supercomputing in Nuclear Applications + Monte Carlo
|
|
|---|---|---|
| Article Number | 04207 | |
| Number of page(s) | 7 | |
| Section | 4. Advanced Parallelism and HPC Strategies: b. Monte Carlo Methods, Parallelism and HPC | |
| DOI | https://doi.org/10.1051/snamc/201404207 | |
| Published online | 06 June 2014 | |
The energy band memory server algorithm for parallel Monte Carlo transport calculations
1 Argonne National Laboratory, MCS Division, 9700 S. Cass Ave, Building 240, Argonne, IL 60439
2 Argonne National Laboratory, MCS and NE Divisions, 9700 S. Cass Ave, Building 240, Argonne, IL 60439
3 Massachusetts Institute of Technology, Department of NSE, 77 Massachusetts Ave, Cambridge, MA 02139
An algorithm is developed to significantly reduce the on-node footprint of cross section memory in Monte Carlo particle tracking algorithms. The classic method of per-node replication of cross section data is replaced by a memory server model, in which the read-only lookup tables reside on a remote set of disjoint processors. The main particle tracking algorithm is then modified in such a way as to enable efficient use of the remotely stored data in the particle tracking algorithm. Results of a prototype code on a Blue Gene/Q installation reveal that the penalty for remote storage is reasonable in the context of time scales for real-world applications, thus yielding a path forward for a broad range of applications that are memory bound using current techniques.
Key words: Monte Carlo / memory server / cross section / energy band
© Owned by the authors, published by EDP Sciences, 2014