TY - JOUR
T1 - Computational fluid dynamics on parallel processors
AU - Agarwal, R. K.
AU - Lewis, J. C.
N1 - Funding Information:
Acknowledgements--The authors wish to thank James D. Wilson of AFOSR, Michael J. Werle and Joseph R. Caspar of UTRC, and Martin H. Schultz of Yale University, for their comments and suggestions. This work was supported in part by the Air Force Office of Scientific Research under contract AFOSR-86-0098 and in part by United Technologies Research Center.
PY - 1992
Y1 - 1992
N2 - Greater computational power is needed for solving computational fluid dynamics problems of interest in engineering design. Parallel computers offer the promise of providing orders of magnitude increases in computational power compared with current uniprocessor vector supercomputers. This paper is mainly concerned with the implementation of a three-dimensional Navier-Stokes code MDNS3D on concurrent computers with grain sizes ranging from fine to coarse. An overview of commercially available parallel machines and the current state of the art in parallel algorithms is presented. The implementation of MDNS3D on machines such as the CRAY Y-MP/8, IBM 3090S, BBN Butterfly II, Intel iPSC/2, Symult 2010, MASPAR, and the Connection Machine CM-2, is described. Particular attention is paid to differences in implementation on SIMD and MIMD architectures. Factors affecting the performance of the code on different architectures are addressed. In addition, user interface and software portability issues are considered for various machines. Finally, future trends in parallel hardware and software development are assessed, and the factors important in determining the most suitable architecture for performing very large scale calculations are discussed.
AB - Greater computational power is needed for solving computational fluid dynamics problems of interest in engineering design. Parallel computers offer the promise of providing orders of magnitude increases in computational power compared with current uniprocessor vector supercomputers. This paper is mainly concerned with the implementation of a three-dimensional Navier-Stokes code MDNS3D on concurrent computers with grain sizes ranging from fine to coarse. An overview of commercially available parallel machines and the current state of the art in parallel algorithms is presented. The implementation of MDNS3D on machines such as the CRAY Y-MP/8, IBM 3090S, BBN Butterfly II, Intel iPSC/2, Symult 2010, MASPAR, and the Connection Machine CM-2, is described. Particular attention is paid to differences in implementation on SIMD and MIMD architectures. Factors affecting the performance of the code on different architectures are addressed. In addition, user interface and software portability issues are considered for various machines. Finally, future trends in parallel hardware and software development are assessed, and the factors important in determining the most suitable architecture for performing very large scale calculations are discussed.
UR - http://www.scopus.com/inward/record.url?scp=38249015591&partnerID=8YFLogxK
U2 - 10.1016/0956-0521(92)90110-5
DO - 10.1016/0956-0521(92)90110-5
M3 - Article
AN - SCOPUS:38249015591
SN - 0956-0521
VL - 3
SP - 251
EP - 259
JO - Computing Systems in Engineering
JF - Computing Systems in Engineering
IS - 1-4
ER -