Spatial and angular domain decomposition algorithms for the curvilinear SNtransport theory method
- 1 April 1993
- journal article
- research article
- Published by Taylor & Francis in Transport Theory and Statistical Physics
- Vol. 22 (2) , 391-417
- https://doi.org/10.1080/00411459308203820
Abstract
This paper surveys several domain decomposition algorithms developed for the 1-D and 2-D curvilinear SN transport theory methods. The angular and spatial domain decomposition algorithms are incorporated into TWOTRAN-II (an SN production code) and the parallel performances of these algorithms are measured on the IBM 3090 multiprocessors, mainly in the ‘BATCH’ mode. The measured parallel efficiencies for most test cases are greater than 60%. Further, it is demonstrated that the combined spatial and angular domain decomposition algorithms yield tens of independent tasks without significantly affecting solution convergence.Keywords
This publication has 5 references indexed in Scilit:
- Domain Decomposition of a Two-DimensionalSnMethodNuclear Science and Engineering, 1992
- Angular Parallelization of a CurvilinearSnTransport Theory MethodNuclear Science and Engineering, 1991
- Multigrid MethodsPublished by Society for Industrial & Applied Mathematics (SIAM) ,1987
- Generalized Rebalance: A Common Framework for Transport Acceleration MethodsNuclear Science and Engineering, 1978
- Diffusion Synthetic Acceleration Methods for the Diamond-Differenced Discrete-Ordinates EquationsNuclear Science and Engineering, 1977