Solving Navier- Stokes Equations On a Massively Parallel Processor: Beyond the 1 Gflop Performance
- 1 March 1990
- journal article
- Published by SAGE Publications in The International Journal of Supercomputing Applications
- Vol. 4 (1), 72-80
- https://doi.org/10.1177/109434209000400106
Abstract
This paper reports on our experience in solving large- scale fluid dynamics problems on the Connection Ma chine model CM_2. We have implemented a parallel version of the MacCormack scheme for the solution of the Navier-Stokes equations. By using triad floating point operations and reducing the number of interprocessor communications, we have achieved a sustained perfor mance rate of 1.42 GFLOPS.Keywords
This publication has 6 references indexed in Scilit:
- Comparison of several finite-difference methodsJournal of Aircraft, 1990
- New architectures: Performance highlights and new algorithmsParallel Computing, 1988
- On a Class of High Resolution Total-Variation-Stable Finite-Difference SchemesSIAM Journal on Numerical Analysis, 1984
- A nonreflecting outflow boundary condition for subsonic navier-stokes calculationsJournal of Computational Physics, 1980
- Towards the ultimate conservative difference scheme. V. A second-order sequel to Godunov's methodJournal of Computational Physics, 1979
- Dissipative two-four methods for time-dependent problemsMathematics of Computation, 1976