Scalability of Parallel Gridpoint Limited-Area Atmospheric Models. Part II: Semi-Implicit Time-Integration Schemes

Roar Skålin Department of Mathematical Sciences, Norwegian University of Science and Technology, Trondheim, Norway

Search for other papers by Roar Skålin in
Current site
Google Scholar
PubMed
Close
Restricted access

Abstract

A performance model of a semi-implicit limited-area atmospheric model is developed, and the parallel scalability is analyzed. Both a direct and an iterative method are studied for the implicit part of the atmospheric model. These methods have been implemented in a parallel message passing version of the limited-area model HIRLAM (High Resolution Limited Area Model), and numerical experiments are in good agreement with the performance model. The percentage of the total elapsed time for the full model taken by the implicit algorithms is not significantly higher on a parallel system in comparison with vector systems. The experiments indicate that when the number of processors increases, the problem size can increase at a rate that almost maintains a constant number of operations per processor and yet still results in an elapsed time below a prescribed upper limit.

Corresponding author address: Roar Skålin, Dept. of Mathematical Sciences, Norwegian University of Science and Technology, N-7034 Trondheim, Norway.

Email: Roar.Skaalin@itea.ntnu.no

Abstract

A performance model of a semi-implicit limited-area atmospheric model is developed, and the parallel scalability is analyzed. Both a direct and an iterative method are studied for the implicit part of the atmospheric model. These methods have been implemented in a parallel message passing version of the limited-area model HIRLAM (High Resolution Limited Area Model), and numerical experiments are in good agreement with the performance model. The percentage of the total elapsed time for the full model taken by the implicit algorithms is not significantly higher on a parallel system in comparison with vector systems. The experiments indicate that when the number of processors increases, the problem size can increase at a rate that almost maintains a constant number of operations per processor and yet still results in an elapsed time below a prescribed upper limit.

Corresponding author address: Roar Skålin, Dept. of Mathematical Sciences, Norwegian University of Science and Technology, N-7034 Trondheim, Norway.

Email: Roar.Skaalin@itea.ntnu.no

Save
  • Barros, S. R. M., and T. Kauranne, 1994: On the parallelization of global spectral weather models. Parallel Comput.,20, 1335–1356.

    • Crossref
    • Export Citation
  • Bjørge, D., 1995: Implementation of the semi-implicit scheme in a message passing version of HIRLAM. Coming of Age, G.-R. Hoffman and N. Kreitz, Eds., World Scientific, 75–90.

  • ———, and R. Skålin, 1995: PARLAM—The parallel HIRLAM version at DNMI. Research Rep. 27, 46 pp. [Available from Norwegian Meteorological Institute, P.O. Box 43 Blindern, N-0313 Oslo, Norway.].

  • Cray Research, Inc., 1994: SHMEM User’s Guide SN-2516 1.1, 30 pp. [Available from Cray Research, Inc., 2360 Pilot Knob Rd., Mendota Heights, MN 55120.].

  • Foster, I., 1995: Designing and Building Parallel Programs. Addison-Wesley, 381 pp.

  • ———, and P. H. Worley, 1996: Parallel algorithms for the spectral transform method. SIAM J. Sci. Comput., in press.

  • Gustafsson, N., Ed., 1992: The HIRLAM 2 final report. HIRLAM Tech. Rep. 9, 126 pp. [Available from SMHI, S-60176 Norrköping, Sweden.].

  • Intel Corporation, 1995: Paragon™ System User’s Guide, 412 pp. [Available from Intel Corporation, 15201 N. W. Greenbrier parkway, Beaverton, OR 97006.].

  • Källén, E., Ed., 1996: HIRLAM Documentation Manual. System 2.5, 241 pp. [Available from SMHI, S-60176 Norrköping, Sweden.].

  • Kauranne, T., J. Oinonen, S. Saarinen, O. Serimaa, and J. Hietaniemi, 1995: The operational HIRLAM 2 model on parallel computers. Coming of Age, G.-R. Hoffman and N. Kreitz, Eds., World Scientific, 63–74.

    • Crossref
    • Export Citation
  • Lie, I., 1994: Splitting of the Helmholtz equation in HIRLAM. FFI/Rep.T-94/04048, 10 pp. [Available from NDRE, N-2007 Kjeller, Norway.].

  • Skålin, R., 1997: Scalability of parallel gridpoint limited-area atmospheric models. Part I: Explicit time-integration schemes. J. Atmos. Oceanic Technol.,14, 427–441.

    • Crossref
    • Export Citation
  • ———, D. Bjørge, N. Gustafsson, I. Lie, and D. Salmond, 1994: Parallelization of HIRLAM on MIMD platforms. Parallel Scientific Computing, J. Dongarra and J. Waśniewski, Eds., Lecture Notes in Computer Science, Vol. 879, Springer-Verlag, 434–457.

    • Crossref
    • Export Citation
All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 286 165 2
PDF Downloads 37 12 3