Solution of Nonlinear Structural Problems Using Array Processors.

Abstract

The paper describes ongoing efforts to investigate the usefulness of so called 'attached processors' or 'array processors', combined with mini- and superminicomputers, in the analysis of computationally demanding engineering problems; such as, in finite element nonlinear applications. A close examination of some basic algorithms, which play an important part in such analysis, is presented. Furthermore, a methodology is developed, which forms a useful guide to further efforts of this nature, including the more general case of multiprocessing. An emphasis is made on the ability to predict the performance of algorithms from time measurements of certain basic operations, coupled with an understanding of the characteristics of the hardware in question and the interplay between its various components. Four hardware alternatives are considered, two of which have attached array processors. Of the algorithms considered, it appears that the decomposition process is the one which benefits most from the presence of an array processor. The speed advantage gained in stiffness assembly, and forward and backward substitution may both be obtained by purchasing a more powerful superminicomputer. (Author)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Aug 20, 1981
Accession Number
ADA104987

Entities

People

  • Hussein Kamel
  • Jacek Maitan
  • Nesrin Sarigul

Organizations

  • University of Arizona

Tags

Communities of Interest

  • C4I
  • Materials and Manufacturing Processes
  • Space

DTIC Thesaurus Topics

  • Algorithms
  • Computations
  • Computer Programming
  • Computers
  • Data Transmission
  • Engineering
  • Finite Element Analysis
  • Host Computers
  • Measurement
  • Mechanical Engineering
  • Midrange Computers
  • Operating Systems
  • Parallel Computing
  • Parallel Processing
  • Simulations
  • Structural Mechanics
  • Urban Areas

Fields of Study

  • Engineering

Readers

  • Finite Element Method (FEM) for solving Partial Differential Equations (PDEs)
  • Parallel and Distributed Computing.
  • Theoretical Analysis.