Measuring productive computational work

  • Authors:
  • Marvin Talbott

  • Affiliations:
  • Texas Instruments Incorporated, Dallas, Texas

  • Venue:
  • ACM '76 Proceedings of the 1976 annual conference
  • Year:
  • 1976

Quantified Score

Hi-index 0.00

Visualization

Abstract

The measurement of computational work in computer systems traditionally relied heavily on indicators of hardware. Unfortunately, these indicators have a low correlation to the user's perception of useful work. This paper defines computations in terms of fundamental bit logical manipulations and shows the user's “useful computational work” can be quantized from a job specification in FORTRAN source statements. The method can be automated yielding a specific quantity of work for a given method FORTRAN program with specified data. The resulting measure of work can be used to obtain ranges of computational power for any system on which the benchmarks are executed.