The Expressive Power and Complexity of Dynamic Process Graphs

  • Andreas Jakoby
  • Maciej LiŚkiewicz
  • Rüdiger Reischuk
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 1928)


A model for parallel and distributed programs, the dynamic process graph, is investigated under graph-theoretic and complexity aspects. Such graphs are capable of representing all possible executions of a parallel or distributed program in a very compact way. The size of this representation is small — in many cases only logarithmic with respect to the size of any execution of the program. An important feature of this model is that the encoded executions are directed acyclic graphs with a regular structure, which is typical of parallel programs, and that it embeds constructors for parallel programs, synchronization mechanisms as well as conditional branches.

In a previous paper we have analysed the expressive power of the general model and various restrictions. Furthermore, from an algorithmic point of view it is important to decide whether a given dynamic process graph can be executed correctly and to estimate the minimal deadline given enough parallelism. Our model takes into account communication delays between processors when exchanging data.

In this paper we study a variant with output restriction. It is appropriate in many situations, but its expressive power has not been known exactly. First, we investigate structural properties of the executions of such dynamic process graph s \( \mathcal{G} \). A natural graph-theoretic conjecture that executions must always split into components isomorphic to subgraphs of \( \mathcal{G} \) turns out to be wrong. We are able to establish a weaker property. This implies a quadratic bound on the maximal deadline in contrast to the general case, where the execution time may be exponential. However, we show that the problem to determine the minimal deadline is still intractable, namely this problem is \( \mathcal{N} \) EXPTIME-complete as is the general case. The lower bound is obtained by showing that this kind of dynamic process graph s can represent certain Boolean formulas in a highly succint way.


Execution Time Schedule Problem Sink Node Parallel Program Expressive Power 
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.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    H. Galperin, A. Wigderson, Succinct Representations of Graphs, Information and Control 56, 1983, 183–198.zbMATHMathSciNetGoogle Scholar
  2. 2.
    S. Ha, E. Lee, Compile-time Scheduling and Assignment of Data-flow Program Graphs with Data-dependent Iteration, IEEE Trans. Computers 40, 1991, 1225–1238.Google Scholar
  3. 3.
    A. Jakoby, M. LiŚkiewicz, R. Reischuk, Scheduling Dynamic Graphs, in Proc. 16th STACS’99, LNCS 1563, Springer-Verlag, 1999, 383–392; for a complete version see TR A-00-02, Universität Lübeck, 2000.Google Scholar
  4. 4.
    T. Lengauer, K. Wagner, The Correlation between the Complexities of the Non-hierarchical and Hierarchical Versions of Graph Problems, J. CSS 44, 1992, 63–93.zbMATHMathSciNetGoogle Scholar
  5. 5.
    C. Papadimitriou and M. Yannakakis, A Note on Succinct Representations of Graphs, Information and Control 71, 1986, 181–185.zbMATHMathSciNetGoogle Scholar
  6. 6.
    C. Papadimitriou and M. Yannakakis, Towards an Architecture-Independent Analysis of Parallel Algorithms, Proc. 20. STOC, 1988, 510–513, see also SIAM J. Comput. 19, 1990, 322-328.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2000

Authors and Affiliations

  • Andreas Jakoby
    • 1
    • 2
  • Maciej LiŚkiewicz
    • 1
    • 3
  • Rüdiger Reischuk
    • 1
  1. 1.Institut für Theoretische InformatikMedizinische Universität zu LübeckGermany
  2. 2.Department of Computer ScienceUniversity of TorontoCanada
  3. 3.Instytut InformatykiUniwersytet Wroc lawskiPoland

Personalised recommendations