A Comparison Study of the Behavior of Equivalent Algorithms in Fault Injection Experiments in Parallel Superscalar Architectures

  • Authors:
  • Ioannis Vakalis

  • Affiliations:
  • -

  • Venue:
  • SAFECOMP '01 Proceedings of the 20th International Conference on Computer Safety, Reliability and Security
  • Year:
  • 2001

Quantified Score

Hi-index 0.00

Visualization

Abstract

This work describes an experimental comparison study of the behaviour of a set of algorithms in the presence of faults. The algorithms have the characteristic that they belong to the same problem class and a number of comparison studies exist in bibliography, with respect to their numerical, conversion and time and space complexity.The class of the used algorithms solve the matrix exponentiation problem. This is a well studied numerical problem encountered in the area of linear differential equation, with a number of solving algorithms. For this study we use Fault Injection techniques at the compile time, namely software based script fault injection method based on random bit inversion assumption. The experiments are performed on a fully operating, parallel architecture machine with two superscalar processors.The algorithms are studied for their fault tolerance in the presence of an elementary fault detection method based on command duplication and exploiting the parallel architecture.