A Performance Analysis of Communications between Two Massively Parallel Computers: The Connection Machine CM-200a and the 64-Cell iWarp

Abstract

Two massively parallel computers consisting of Thinking Machines Corp's Connection Machine CM-200a and Intel Corp's 64-cell iWarp are being used within an experimental sonar subsystem to evaluate their impact on various signal processing algorithms. Each machine performs specific tasks within the sonar processing thread and are required to communicate with one another in a tightly coupled manner. This interface is in the process of being defined. A number of design options exist; however, it is not obvious which of the options would provide optimum performance. As a result, several studies are being conducted to help resolve this issue. This paper documents two such studies, one which focuses on the performance of various I/O command suites available on the Connection Machine, the other on the performance of reformatting data (corner turning) so that data generated on one machine (the iWarp) is usable by the other (the Connection Machine).

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 12, 1994
Accession Number
ADA638529

Entities

People

  • Matthew J. Krzych

Organizations

  • Naval Undersea Warfare Center

Tags

Communities of Interest

  • Materials and Manufacturing Processes
  • Sensors

DTIC Thesaurus Topics

  • Access Time
  • Algorithms
  • Communication Channels
  • Computers
  • Data Transmission
  • Ethernet
  • Frequency
  • Host Computers
  • Instructions
  • Military Research
  • Signal Processing
  • Software Development
  • Standards
  • Thinking
  • Three Dimensional
  • Throughput
  • Undersea Warfare

Fields of Study

  • Computer science
  • Engineering

Readers

  • Computational Fluid Dynamics (CFD)
  • Computer Science/Computer Engineering/Data Science/Digital Signal Processing.
  • Electrical Engineering