A GRAPH MODEL FOR PARALLEL COMPUTATIONS.

Abstract

The report presents a computational model called 'program graphs' which makes possible a precise description of parallel computations of arbitrary complexity on non-structured data. In the model, the computation steps are represented by the nodes of a directed graph whose links represent the elements of storage and transmission of data and/or control information. The activation of the computation represented by a node depends only on the control information residing in each of the links incident into and out of the node. At any given time any number of nodes may be active, and there are no assumptions in the model regarding either the length of time required to perform the computation represented by a node or the length of time required to transmit data or control information from one node to another. Data dependent decisions are incorporated in the model in a novel way which makes a sharp distinction between the local sequencing requirements arising from the data dependency of the computation steps and the global sequencing requirements determined by the logical structure of the algorithm. (Author)

Document Details

Document Type
Technical Report
Publication Date
Sep 01, 1969
Accession Number
AD0697759

Entities

People

  • Jorge E. Rodriquez

Organizations

  • Massachusetts Institute of Technology

Tags

DTIC Thesaurus Topics

  • Algorithms
  • Computational Complexity
  • Computations
  • Mathematical Analysis
  • Mathematics
  • Parallel Computing

Fields of Study

  • Computer science

Readers

  • Atmospheric Science / Meteorology, specifically Wind Wave Turbulence.
  • Neural Network Machine Learning.
  • Parallel and Distributed Computing.