Advertisement

On the Performance Enhancements of a CFD Algorithm in High Performance Computing

  • Michael Lenke
  • Arndt Bode
  • Thomas Michl
  • Siegfried Wagner
Part of the Notes on Numerical Fluid Mechanics (NNFM) book series (NONUFM, volume 48)

Summary

This report deals with aspects concerning HPC with respect to the industrial CFD solver NSFLEX for aerodynamic problems. Since 1992, this solver has been parallelized and tested on different parallel computer architectures and different communication paradigms. Good performance results could be achieved. Based on that parallel solver, some more aspect are considered in this report to improve its performance behavior. Multigrid techniques and cg-like methods are implemented as well as the use of new programming paradigms like Fortran 90 and HPF (High Performance Fortran). Again, performance improvement can be achieved. Furthermore, one very important aspect concerning the application cycle of high parallel solvers is introduced and an interactive tool concept is proposed which enables online monitoring, online analyzing and online steering of parallel simulations to the user.

Keywords

Computational Fluid Dynamics High Performance Computing Speedup Factor Multigrid Method Implicit Solver 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. [1]
    Baldwin, B. S., UND Lomax, H. Thin Layer Approximation and Algebraic Model for Separated Turbulent Flow. Report AIAA 78-0257 (1978).Google Scholar
  2. [2]
    Bode, A., Lenke, M., Wagner, S., UND Michl, T. Implicit Euler Solver on Alliant FX/2800 and Intel iPSC/860 Multiprocessors. In Flow Simulation with High-Performance Computers I, E. H. Hirschel, Ed., vol. 38 of Notes on Numerical Fluid Mechanics (NNFM). Vieweg, Braunschweig, 1993, Seiten 41–55.Google Scholar
  3. [3]
    Brandt, A.Multigrid Techniques: 1984 Guide with Applications to Fluid Dynamics. GMD-Studien 85, 1984.Google Scholar
  4. [4]
    Dick, E. A multigrid method for steady incompressible navier-stokes equations based on partial flux splitting. In International journal for numerical methods in fluids (1989), vol. 9, Seiten 113–120.zbMATHCrossRefGoogle Scholar
  5. [5]
    Dick, E. Multigrid methods for steady Euler-and Navier-Stokes equations vased on polynomial flux-difference splitting. In International Series of Numerical Mathematics (1991), vol. 98.Google Scholar
  6. [6]
    Fairley, R. E.Software Engineering Concepts. McGraw-Hill series in software engineering and technology. McGraw-Hill, 1985.Google Scholar
  7. [7]
    Flechter, R. Conjugate Gradient Methods for Indefinite Systems. In Lecture Notes in Math. 506, H. Watson, Ed. Springer, Berlin, 1976.Google Scholar
  8. [8]
    Hackbusch, W.Multi-Grid Methods and Applications. Springer-Verlag, 1985.Google Scholar
  9. [9]
    Hänel, D., UND Schwane, R. An Implicit Flux Vector Splitting Scheme for the Computation of Viscous Hypersonic Flows. Report AIAA 89-0274 (1989).Google Scholar
  10. [10]
    Hold, R. K., UND Ritzdorf, H. Portable Parallelization of the Navier-Stokes Code NSFLEX. POPINDA project report, 1995.Google Scholar
  11. [11]
    Hortmann, M., Peric, M., UND Scheuerer, G. Finite volume multigrid prediction of laminar natural convection: bench-mark solutions. In International journal for numerical methods in fluids (1990), vol. 11, Seiten 189–207.zbMATHCrossRefGoogle Scholar
  12. [12]
    Howard, D., Connolley, W. M., UND Rollett, J. S. Solution of Finite Element Fluid Matrix Systems by Unsymmetric Conjugate Gradient and Direct Methods on the Cray-XMP and Cray 2. In Conference Proceedings on Application of Supercomputers in Engineering: Fluid Flow and Stress Analysis Applications (Amsterdam, 1989), C. A. Breb-bia, Ed., Elsevier.Google Scholar
  13. [13]
    Kanarachos, A., UND Vournas, I. Multigrid solutions for the compressible euler equations by an implicit characterictic-flux-averaging. In Computational Fluid Dynamics’ 92 (1992), vol. 2, Elsevier Science Publishers.Google Scholar
  14. [14]
    Koren, B. Multigrid and Defect Correction for the Steady Navier-Stokes Equations. In Journal of Computational Physics (1990), vol. 87, Seiten 25–46.zbMATHCrossRefGoogle Scholar
  15. [15]
    Lenke, M. An interactive engineering tool for parallel HPC applications. Advances in Engenineering Software (AES). (accepted 1995).Google Scholar
  16. [16]
    Lenke, M. VIPER — A tool concept to control parallel CFD applications. In Applications of High-Performance Computing in Engineering IV (June 1995), H. Power, Ed., Wessex Institute of Technology, UK, Computational Mechanics Publications, Southampton Boston, Seiten 179–186. ASE’95.Google Scholar
  17. [17]
    Lenke, M. VIPER — Konzept eines interaktionsvitalen Werkzeuges zur Leistungssteigerung im Anwendungszyklus paralleler Simulationsalgorithmen. In Luft-und Raumfahrt — Schlüsseltechnologien für das 21. Jahrhundert (1995). Annual Meeting 1995, Bonn — Bad Godesberg, 26.–29. September 1995.Google Scholar
  18. [18]
    Lenke, M., UND Bode, A. Methoden zur Leistungssteigerung von “real-world” Simulationsanwendungen in der Luft-und Raumfahrt für das technisch-wissenschaftliche Hochleistungsrechnen. In Basistechnologien für neue Herausforderungen in der Luft-und Raumfahrt (1994), Seiten 465–474. Annual Meeting 1994, Erlangen, Oktober 4–7.Google Scholar
  19. [19]
    Lenke, M., Bode, A., Michl, T., Maier, S., UND Wagner, S. Dataparallel Navier-Stokes Solutions on Different Multiprocessors. In Applications of Supercomputers in Engineering III, C. Brebbia und H. Power, Eds. Computational Mechanics Publications, Sept. 1993, Seiten 263–277. ASE’93 in Bath (UK).Google Scholar
  20. [20]
    Lenke, M., Bode, A., Michl, T., UND Wagner, S. Nsflex90 — A 3d Euler and Navier-Sokes solver in Fortran 90. In Computational Fluid Dynamics on Parallel System (1995), S. Wagner, Ed., vol. 50 of Notes on Numerical Fluid Mechanics (NNFM), Vieweg Braunschweig, Seiten 125–134.Google Scholar
  21. [21]
    Lenke, M., UND Oliva, L. Implementation of a direct FAS-similar methode in an implicit solver for the Navier-Stokes equations, (to be submitted 1996).Google Scholar
  22. [22]
    Lenke, M., Rathmayer, S., Bode, A., Michl, M., UND Wagner, S. Parallelizations with a real-world CFD application on different parallel architectures. In High-performance Computing in Engineering, C. B. H. Power, Ed., vol. 2. Computational Mechanics Publications, Southhampton, UK, 1995, ch. 4, Seiten 119–166.Google Scholar
  23. [23]
    Michl, T.Effiziente Euler-und Navier-Stokes-Löser für den Einsatz auf Vektor-Hochleistungsrechnern und massiv-parallelen Systemen. Dissertation, Institut für Aerodynamik und Gasdynamik der Universität Stuttgart, 1995.Google Scholar
  24. [24]
    Michl, T., Maier, S., Wagner, S., Lenke, M., UND Bode, A. The Parallel Implementation of a 3d Navier-Stokes Solver on Intel Multiprocessor Systems. In Applications on Massively Parallel Systems (Sept. 1993), vol. 7 of SPEEDUP, Proceedings of 14th Workshop; Zurich, Seiten 19–23.Google Scholar
  25. [25]
    Michl, T., UND Wagner, S. A Parallel Navier-Stokes Solver Based on BI-CGSTAB on Different Multiprocessors. In International Symposium on Computational Fluid Dynamics (Sept. 1995), vol. II, Seiten 827–832.Google Scholar
  26. [26]
    Michl, T., Wagner, S., Lenke, M., UND Bode, A. Big Computation with a 3-D Parallel Navier-Stokes Solver on Different Multiprocessors. In Computational Fluid Dynamics on Parallel Systems (1995), S. Wagner, Ed., vol. 50 of Notes on Numerical Fluid Mechanics (NNFM), Vieweg Braunschweig, Seiten 157–166.Google Scholar
  27. [27]
    Michl, T., Wagner, S., Lenke, M., UND Bode, A. Dataparallel Implicit 3-D Navier-Stokes Solver on Different Multiprocessors. In Parallel Computational Fluid Dynamics-New Trends and Advances (May 1995), P. L. A. Ecer, J. Hauser und J. Periaux, Eds., Elsevier, North-Holland, Seiten 133–140.CrossRefGoogle Scholar
  28. [28]
    Paillere, H., Deconinck, H., UND Bonfiglioli, A. A Linearity Preserving Wave-Model for the Solution of the Euler Equations on Unstructured Meshes. In Computational Fluid Dynamics’ 94 (Chisester, 1994), S. Wagner, E. Hirschel, J. Periaux, und R. Piva, Eds., John Wiley & Sons.Google Scholar
  29. [29]
    Peters, A., Babovsky, H., Daniels, H., Michl, T., Maier, S., Wagner, S., Bode, A., UND Lenke, M. Parallel Implementation of Nsflex on an IBM RS/6000 Cluster. In Supercomputer Applications in the Automotive Industries (Sept. 1993), Seiten 289–296.Google Scholar
  30. [30]
    Rathmayer, S., Oliva, L., Krammer, J., UND Lenke, M. Experiences with HPF compiler systems with respect to the real world simulation algorithm NsflexHPF. (submitted).Google Scholar
  31. [31]
    Saad, Y., UND Schultz, M. H. GMRES: A Generalized Minimal Residual Algorithm for Solving Nonsymmetric Linear Systems. SIAM J. Sci. Stat. Comput.7 (1986), 856–869.MathSciNetzbMATHCrossRefGoogle Scholar
  32. [32]
    Sonnefeld, P. CGS, a fast Lanczos-type Solver for Nonsymmetric Linear Systems. SIAM J. Sci. Stat. Comput. 10 (1989), 36–52.CrossRefGoogle Scholar
  33. [33]
    STüben, K., UND Trottenberg, U.Multigrid methods. GMD-Studien 96, 1985.Google Scholar
  34. [34]
    van der Vorst, H. A. Iterative Solution Methods for Certain Sparse Linear Systems with a Non-Symmetric Matrix Arising from PDE Problems. Journal of Computational Physics 44 (1981), 1–19.MathSciNetzbMATHCrossRefGoogle Scholar
  35. [35]
    van der Vorst, H. A. BI-CGSTAB: A Fast and Smoothly Converging Variant of BI-CG for the Solution of Nonsymmetric Linear Systems. SIAM J. Sci. Stat. Comput. 13, 2 (1992), 631–644.zbMATHCrossRefGoogle Scholar
  36. [36]
    Wagner, S. Computational Fluid Dynamics on Parallel Systems, vol. 50. Vieweg und Sohn, Braunschweig/Wiesbaden, 1995.Google Scholar
  37. [37]
    Wirtz, R.Eine Methodik für die aerodynamische Vorentwurfsrechnung im Hyperschall. Dissertation, Institut für Aerodynamik und Gasdynamik der Universität Stuttgart, 1993.Google Scholar
  38. [38]
    Ziegler, H.-J. Einsatz von Mehrgittermethoden zur Konvergenzbeschleunigung bei Euler-Lösern. Diplomarbeit, Universität der Bundeswehr, Munich, 1992.Google Scholar

Copyright information

© Friedr. Vieweg & Sohn Verlagsgesellschaft mbH, Braunschweig/Wiesbaden 1996

Authors and Affiliations

  • Michael Lenke
    • 1
  • Arndt Bode
    • 1
  • Thomas Michl
    • 2
  • Siegfried Wagner
    • 2
  1. 1.Lehrstuhl für Rechnertechnik und RechnerorganisationTU MünchenMünchenGermany
  2. 2.Institut für Aerodynamik und GasdynamikUniversität StuttgartStuttgartGermany

Personalised recommendations