Implementation and Evaluation of Interval Arithmetic Software. Report 5. The CDC CYBER 70 System.

Abstract

This is Report 5 of a series entitled 'Implementation and Evaluation of Interval Arithmetic Software. 'The potential user of this or any other interval analysis package should take into consideration the limitations of this technique. Some of the limitations are inherent in all interval analysis implementations and others are dependent upon one specific package. The user should be aware that software interval analysis tends to be very slow; cases run on one CDC CYBER 70 have required as much 100-fold increase in execution time over a noninterval version. Thus, large production codes involving significant amounts of floating-point computations are not viable candidates for a complete interval implementation; however, small portions of the computation might benefit from use of the package. The computed bounds using interval analysis can be overly pessimistic as well as being time-consuming to determine. The tightness of the bounds is dependent upon the arithmetic of the host computer and possible numerical anomalies introduced by compile-time arithmetic. Another factor affecting the performance of the interval package is the accuracy of the host computer's FORTRAN library routines.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Apr 01, 1979
Accession Number
ADA070375

Entities

People

  • David A. Cohn
  • J. Brian Potter
  • Myron Ginsberg

Organizations

  • Southern Methodist University

Tags

Communities of Interest

  • Air Platforms
  • Autonomy
  • Cyber
  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Artificial Intelligence
  • Central Processing Units
  • Computations
  • Computer Science
  • Computers
  • Contracts
  • Data Processing
  • Engineering
  • Engineers
  • Host Computers
  • Laser Diodes
  • Manufacturing
  • Production
  • Self Assembly
  • Sequences
  • Test And Evaluation
  • Universities

Fields of Study

  • Computer science

Readers

  • Computer Programming and Software Development.
  • Parallel and Distributed Computing.
  • Statistical inference.

Technology Areas

  • Cyber
  • Cyber - Cryptography