A Large Grain Parallel Navier-Stokes Solver
Project team
Research Objectives
The simulation of a flow is highly computational intensive. For
each gridpoint, 10^{7} floating point operations have to be performed
if the Navier-Stokes equations have to be solved. In CFD a
grid can easily consist of 10^{6} gridpoints (or more), which results
in 10^{13} floating point operations to be performed in one run of a
flow simulation program.
SOLEQS is a three-dimensional multi-block multi-zone compressible
Navier-Stokes solver, produced by NLR.
The objective of this project is to parallelize SOLEQS for a wide range of parallel
platforms: clusters of IBM RISC 6000/Systems, clusters of SGI Indigo's and Indy's,
IBM SP1, NEC Cenju-3, SGI Challenge, Intel Paragon, and Cray C98. The actual parallelization
methodology consisted of a number of well-defined phases which enabled to analyze high performance
computing aspects during parallelization and allowed to exploit this analysis by restructuring the
parallel system into a system which is more efficient. The software development is according to
the quality assurance ISO-9001 industry standard. Computing time on the Cray C98 is granted by
NCF/Cray Research Inc. Grant SC-374.
Collaborations
The parallel solver is developed in close collaboration with the
National Aerospace Laboratory
(Projectleaders: P.A. van Mourik and G. van Beek).
Intended results, deliverables
Design and efficient implementations of the parallel version of SOLEQS specifically
targeted for the systems mentioned above. Transfer of parallelization
expert knowledge to NLR.
Timetable
Starting date: May 1993
Ending date: April 1995
Recent publications
- G. van Beek, J.P. Geschiere and E.H. van Hardeveld.
Parallelization of two CFD simulation codes on a Distributed Memory
machine by Domain Decomposition.
To appear in Proceedings of Parallel CFD '94, Kyoto, Japan, June 1994.
- G. van Beek, J.P. Geschiere and A.R. Sukul.
Parallelization of a 3D multi-block Navier-Stokes flow solver
on a distributed memory MIMD machine.
Proceedings of High-Performance Computing and Networking
Conference 1994, Munich, Germany,
Proceedings Volume I, pp.163-170, April 1994.
- J.P. Geschiere.
Using a Modular Approach to Parallelizing a Large-Scale CFD System.
Technical Report 95-28, Department of Computer Science, Leiden University, September 1995.
- J.P. Geschiere.
Installation Guide for the Parallel Navier-Stokes Solver SOLEQS.
Memorandum IN-95-023, National Aerospace Laboratory NLR, July 1995.
- J.P. Geschiere.
Verification Report of the Parallel Navier-Stokes Solver SOLEQS.
Technical Report National Aerospace Laboratory NLR TR 95255 L, June 1995.
- J.P. Geschiere.
A Large Grain Parallel Navier-Stokes Solver, Phase 6 of PARASOL:
Data-Decomposition for a distributed memory MIMD machine,
Detailed Design.
Technical Report National Aerospace Laboratory NLR TR 95163 C, April 1995.
- J.P. Geschiere and P.A. van Mourik.
Solving the Navier-Stokes equations on SGI and IBM workstation clusters.
Technical Report National Aerospace Laboratory NLR TR 94471 L,
November 1994.
- J.P. Geschiere.
Evaluation of the parallelization of a 3D multi-block
Navier-Stokes solver for the RISC-based IBM parallel systems.
Technical Report National Aerospace Laboratory NLR TP 94366 L, September 1994
(IBM deliverable).
- J.P. Geschiere.
Structured parallelization of a multi-block
Navier-Stokes solver targeting distributed memory platforms.
Technical Report National Aerospace Laboratory NLR TP 94364 L, August 1994
(IBM deliverable).
- J.P. Geschiere and P.A. van Mourik.
Parallelizing a large scale 3D multi-block Navier-Stokes solver.
Technical Report National Aerospace Laboratory NLR TP 94239 L, June 1994.
- G. van Beek, J.P. Geschiere and A.R. Sukul.
Parallelization of a 3D multi-block Navier-Stokes flow solver
on a distributed memory MIMD machine.
Technical Report National Aerospace Laboratory NLR TP 94158 L, June 1994
(This report is published at HPCN94, cf. above).
- J.P. Geschiere.
A Large Grain Parallel Navier-Stokes Solver, Phase 6 of PARASOL:
Data-Decomposition for a distributed memory MIMD machine,
Technical Concept.
Technical Report National Aerospace Laboratory NLR TR 94032 L, February 1994.
- J.P. Geschiere.
A Large Grain Parallel Navier-Stokes Solver, Phase 6 of PARASOL:
Data-Decomposition for a distributed memory MIMD machine,
Predesign.
Technical Report National Aerospace Laboratory NLR TR 93456 L, October 1993.
Last modified on July 2, 1996 by Lex Wolters.