• DocumentCode
    282998
  • Title

    DDC: Delta Driven Computer

  • Author

    Gonzalez-Rubio, R. ; Bradier, A. ; Bergsten, B.

  • Author_Institution
    Bull Sa centre de Recherche, Louveciennes, France
  • fYear
    1988
  • fDate
    32177
  • Firstpage
    42491
  • Lastpage
    42497
  • Abstract
    Presents the `Delta Driven Computer´ (DDC), a system that can be specialized to support efficient deductive and database processing. DDC is a parallel inference multiprocessor computer, and is partially supported by ESPRIT-415. The architecture of DDC can be viewed as a multiprocessor system composed of a set of interconnected PCM (processor, communication and memory) nodes, in which there is no need to have a shared memory. From a conceptual point of view, DDC executes a language based on production rules, called VIM (virtual inference machine). This execution is performed following a forward chaining strategy. Given a set of rules and a set of initial facts, the only mode of operation of the machine is saturation (all conclusions are found). VIM is an intermediate language; so the authors are studying how to translate from a high level language to this intermediate language. The execution if VIM is possible with the DDEM (delta driven execution model). In this model the execution is driven by the facts deduced from the rules. The parallelism of the machine is achieved by distributing the facts among PCM nodes and by firing rules independently in each processor
  • Keywords
    database management systems; expert systems; parallel machines; DDC; Delta Driven Computer; ESPRIT-415; PCM nodes; communication nodes; database processing; deductive processing; delta driven execution model; firing rules; forward chaining strategy; intermediate language; memory nodes; parallel inference multiprocessor computer; processor nodes; production rules; saturation; virtual inference machine;
  • fLanguage
    English
  • Publisher
    iet
  • Conference_Titel
    Knowledge Manipulation Engines, IEE Colloquium on
  • Conference_Location
    London
  • Type

    conf

  • Filename
    208785