Significance of Problem Solving Parameters on the Performance of Combinatorial Algorithms on Multi-Computer Parallel Architectures.

Abstract

This paper describes the interaction between specific algorithm parameters and the parallel computer architecture. The classes of architectures we consider are those which have inherent distributed control and whose connection structure is regular. Combinatorial problems require solutions which do searching. This experiment has determined an optimum problem solving strategy for the consistent labeling problem. One combination of factors, depth first search strategy-transmit large problems-transmit 50% of a processor's work, was found to be statistically best, especially for large problem sizes or for architectures with restricted communications paths.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 03, 1982
Accession Number
ADA112487

Entities

People

  • F. Gail Gray
  • Robert M. Haralick
  • W. M. Mccormack

Organizations

  • Virginia Tech

Tags

Communities of Interest

  • Advanced Electronics
  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Algorithms
  • Analysis Of Variance
  • Computations
  • Computer Architecture
  • Computer Networks
  • Computer Programming
  • Computer Science
  • Computers
  • Computing System Architectures
  • Data Analysis
  • Information Science
  • Networks
  • Operations Research
  • Parallel Computing
  • Parallel Processing
  • Parallel Processors
  • Simulations

Fields of Study

  • Computer science

Readers

  • Operations Research
  • Parallel and Distributed Computing.
  • Regression Analysis.