3-D discrete ordinates calculations with parallel-vector processors
Although three-dimensional (3-D) discrete ordinates calculations are of practical value now, they require that large blocks of computer memory be dedicated for the entire run duration. Depending on system parameters, charges for this memory may comprise most of the cost of solving a problem. One way of alleviating this problem is to assign several processors to work in parallel on the job, thus reducing the time during which the memory must be reserved. When system resources are shared with many other users, the wait time required to obtain results may be several days, and the use of several processors can also reduce that wait. In this study, we apply a small number, i.e., 2 or 4, Cray vector processors sharing a large central memory. A typical discrete ordinates calculation at Oak Ridge uses almost all of its CPU time in three major sections: the discrete ordinates flux sweep, a diffusion-like acceleration sweep, and improvement of the source to eliminate negatives. The present work concentrates on the TORT 3-D code, and on the concrete building problems for which it was designed. 3 refs.
- Research Organization:
- Oak Ridge National Lab., TN (USA)
- DOE Contract Number:
- AC05-84OR21400
- OSTI ID:
- 6178476
- Report Number(s):
- CONF-890408-2; ON: DE88016795
- Resource Relation:
- Conference: Advances in nuclear engineering computation and radiation shielding, Santa Fe, NM, USA, 9 Apr 1989; Other Information: Portions of this document are illegible in microfiche products
- Country of Publication:
- United States
- Language:
- English
Similar Records
Solution of the first-order form of the 3-D discrete ordinates equations on a massively parallel machine
Solution of the first-order form of the 3-D discrete ordinates equations on a massively parallel machine