Advertisement

A Run-Time System for Dynamic Grain Packing

  • JoãoLuís Sobral
  • AlbertoJosé Proença
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 1685)

Abstract

The SCOOPP (Scalable Object Oriented Parallel Programming) system is an hybrid compile and run-time system, that extracts parallelism, supports explicit parallelism and dynamically serialises parallel tasks in excess, to dynamically scale applications through a wide range of target platforms. This paper describes the run-time system of the current SCOOPP prototype - the ParC++ - and its mechanism to serialise parallelism. Low level performance results are presented, which indicate that the proposed methodology is effective and provides an high reduction in parallelism overheads. These features can improve the scalability of parallel applications with excessive parallelism.

References

  1. 1.
    Kruatrachue, B., Lewis, T.: Grain Size Determination for Parallel Processing, IEEE Software, Vol. 5(1), January (1988)Google Scholar
  2. 2.
    Gresoulis, A., Yang., T.: On the Granularity and Clustering of Direct Acyclic Graphs, IEEE Transactions on Parallel and Distributed Systems, Vol. 4(6), June (1993)Google Scholar
  3. 3.
    Sobral, J., Proença., A.: Dynamic Grain-Size Adaptation on Object-Oriented Parallel Programming-The SCOOPP Approach, Proc. of the 2nd Merged IPPS/SPDP 1999, Puerto Rico, April (1999)Google Scholar
  4. 4.
    Mohr, E., Kranz, A., Halstead, R.: Lazy Task Creation: A Technique for Increasing the Granularity of Parallel Programs, IEEE Transactions on Parallel and Distributed Processing, V2(3), July (1991)Google Scholar
  5. 5.
    Lopez, P., Hermenegildo, M., Debray, S.: A Methodology for Granularity Based Control of Parallelism in Logic Programs, Journal of Symbolic Computation, Vol. 22, (1998)Google Scholar
  6. 6.
    Sobral, J., Proença, A.: ParC++: A Simple Extension of C++ to Parallel Systems, Proceedings of the 6th Euromicro Workshop on Parallel and Distributed Applications (PDP’98), Madrid, Spain, January (1998)Google Scholar
  7. 7.
    Santos, L., Chalmers, A., Proença, A.: A messages density monitoring strategy for distributed memory parallel system, 2nd Int. Conf. on Software for Multiprocessors and Supercomputers: theory, practice and experience, Moscow, September (1994)Google Scholar
  8. 8.
    Sobral, J., Proença, A.: Overheads on the dynamical removal of excess of parallelism on OO irregular applications, 1st Work. Parallel Computing for Irregular Appl., 5th Int. Symp. HPC Arch.(HPCA-5), http://www-apache.imag.fr/manifestations/PCIA, Orlando, January (1999)

Copyright information

© Springer-Verlag Berlin Heidelberg 1999

Authors and Affiliations

  • JoãoLuís Sobral
    • 1
  • AlbertoJosé Proença
  1. 1.Departamento de InformáticaUniversidade do MinhoBraga CodexPortugal

Personalised recommendations