Advertisement

MeDLey: An abstract approach to message passing

  • E. Dillon
  • J. Guyard
  • G. Wantz
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 1184)

Abstract

Message Passing has become one of the most popular parallel programming paradigms because of its ease of use. Conversely, using it efficiently often requires lots of efforts, when designing a distributed application. Finally, such an application often requires optimization, since performances are not always as good as expected.

With MeDLey, we propose an abstract approach to message passing. The basic idea is to provide a language that might be used to specify all communications within a distributed application. With this specification, the user does not have to care about the implementation any more, and leaves it to the compiler's charge. Finally, the aim of this approach is to achieve more efficiency by generating specific communication primitives rather than using generic message passing libraries.

MeDLey has followed two main steps: MeDLey-0 was the first version of the compiler. It was able to generate C++ and MPI communications primitives, mainly to help the user in the design process. MeDLey-1 now tries to gain more efficiency by implementing communications with dedicated devices like shared-memory or AAL5 for ATM-based clusters of workstations.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. [Dillon 95]
    Eric Dillon, Carlos Gamboa Dos Santos and Jacques Guyard. Homogeneous and Heterogeneous Networks of Workstations: Message Passing Overhead. In MPI Developers Conference '95, June 1995.Google Scholar
  2. [Dillon 96]
    E. Dillon and G. Wantz. Medley: Efficient communications for distributed applications. Technical report, INRIA-Lorraine, 1996. To appear.Google Scholar
  3. [MPI Forum 95]
    MPI Forum. MPI: A Message Passing Interface Standard. University of Tennessee, June 1995.Google Scholar
  4. [Stricker 95a]
    T. Stricker and T. Gross. Optimizing memory system performance for communication in parallel computers. In Proceedings of the 22nd International Symposium on Computer Architecture, Santa Marguerita di Ligure, Italy, June 1995.Google Scholar
  5. [Stricker 95b]
    T. Stricker, J. Stichnoth, D. O'Hallaron, S. Hinrichs and T. Gross. Decoupling synchronization and data transfer in message passing systems of parallel computers. In Proceedings of the 9th International Conference on Supercomputing, Barcelona, Spain, July 1995.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 1996

Authors and Affiliations

  • E. Dillon
    • 1
  • J. Guyard
    • 1
  • G. Wantz
    • 2
  1. 1.CRIN/INRIAVillers-les-NancyFrance
  2. 2.CRP-CULuxembourg

Personalised recommendations