Comparison of Vectorization Methods Used in a Monte Carlo Code
- 1 January 1991
- journal article
- research article
- Published by Taylor & Francis in Nuclear Science and Engineering
- Vol. 107 (1) , 58-66
- https://doi.org/10.13182/nse91-a23780
Abstract
Vectorization methods used in Monte Carlo codes for particle transport calculations are examined. Event and zone selection methods developed from conventional all-zone and one-zone algorithms have been implemented in a general-purpose vectorized code, GMVP. Moreover, a vectorization procedure to treat multiple-lattice geometry has been developed using these methods. Use of lattice geometry can reduce the computation cost for a typical pressurized water reactor fuel subassembly calculation, especially when the zone selection method is used. Sample calculations for external and fission source problems are used to compare the performances of both methods with the results of conventional scalar codes. Though the speedup resulting from vectorization depends on the problem solved, a factor of 7 to 10 is obtained for practical problems on the FACOM VP-100 computer compared with the conventional scalar code, MORSE-CG.Keywords
This publication has 6 references indexed in Scilit:
- Status of Vectorized Monte Carlo for Particle Transport AnalysisThe International Journal of Supercomputing Applications, 1987
- Summary of Recent Results from the Jaeri/U.S. Fusion Neutronics Phase I ExperimentsFusion Technology, 1986
- Monte Carlo photon transport on a vector supercomputerIBM Journal of Research and Development, 1986
- Vectorization of the KENO-IV CodeNuclear Science and Engineering, 1986
- Vectorized Monte Carlo photon transportParallel Computing, 1984
- Monte Carlo methods for radiation transport analysis on vector computersProgress in Nuclear Energy, 1984