Design Tools for Evaluating Multiprocessor Programs

Abstract

An approach to designing programs for implementation in a multiple instruction stream-multiple data stream processing environment is presented. A program is modeled as a directed graph consisting of two types of nodes: processing nodes and linking nodes. Communication among nodes in the model is represented by message tokens. Each processing node is similar in form to a semi-Markov process. A simulation of the operation of the model is nondeterministic, but is based on prescribed probabilistic choice functions. A system, called STEPPS, has been built in which a model can be described and evaluation tools can be used to manipulate and act upon a model to predict performance of a program decomposition. The design approach is to describe a multiprocessing program in terms of the modeling system. The model is examined to determine some analytic attributes of the model. The analysis available determines (a) whether the model is well formed, (b) whether the model contains deadlocks, (c) predictions of steady state properties of each process. In addition, without much difficulty, analysis functions external to STEPPS may be included as needed by a program designer.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jul 01, 1976
Accession Number
ADA034856

Entities

People

  • Philip H. Mason

Organizations

  • Carnegie Mellon University

Tags

DTIC Thesaurus Topics

  • Algorithms
  • Artificial Intelligence
  • Automated Speech Recognition
  • Compilers
  • Computer Programming
  • Computer Science
  • Computers
  • Data Analysis
  • Information Processing
  • Language
  • Machine Languages
  • Markov Processes
  • Organizational Structure
  • Parallel Computing
  • Probability
  • Programming Languages
  • Simulators

Fields of Study

  • Computer science
  • Engineering

Readers

  • Mathematical Modeling and Probability Theory.
  • Parallel and Distributed Computing.
  • Theoretical Analysis.