A Parallel Processor for the Solution of Large, Sparse Symmetric Linear Systems.

Abstract

The need for rapid solution of large, sparse linear systems in certain applications is described. Conventional single processor computers are nearing fundamental speed limits, and will not be able to attain sufficient speeds. Parallel processing allows the application of a large number of processors to these problems, and thus has the potential to achieve faster operation. This dissertation describes a processor designed specifically for this type of problem. This system uses the technique of successive overrelaxation to obtain a solution to the linear system by iteration. The architecture features a separate processor for each variable in the linear system. The processors are arranged in a two dimensional grid, with connections to the four nearest neighbors. Each processor is partitioned into an Arithmetic Unit, which performs the actual computations, and a Communications Unit, to provide the necessary interchange of data between the processors. The design of the processor is described in detail. A detailed simulation is described which shows the performance and efficiency of the proposed system to be very high.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 01, 1986
Accession Number
ADA171256

Entities

People

  • David A. Arpin

Organizations

  • Air Force Institute of Technology

Tags

Communities of Interest

  • Biomedical
  • Energy and Power Technologies
  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Arithmetic Units
  • Central Processing Units
  • Computer Programming
  • Computers
  • Content Addressable Memory
  • Differential Equations
  • Electrical Engineering
  • Engineering
  • Equations
  • Finite Element Analysis
  • Floating Point Operations
  • Image Processing
  • Image Reconstruction
  • Parallel Computing
  • Parallel Processing
  • Parallel Processors
  • Two Dimensional

Readers

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