The algebraic solution of large sparse systems of linear equations using REDUCE 2

  • Authors:
  • Martin L. Griss

  • Affiliations:
  • -

  • Venue:
  • ACM '74 Proceedings of the 1974 annual conference - Volume 1
  • Year:
  • 1974

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper discusses some of the problems encountered during the solution of a large system of sparse linear equations with algebraic coefficients, using REDUCE 2. Of particular importance is intermediate expression swell, which ultimately uses up all the available storage, and produces voluminous unreadable output. By optimally ordering the equations (optimal pivoting algorithms), and decomposing the intermediate expressions, so as to share common sub-expressions (“hash coded CONS”), a considerable saving in storage is achieved. By suitably renaming frequently used common sub-expressions, using the table built up above, and outputting these first, followed by the more complex expressions, a simplification in the output occurs. These techniques are general, and may be useful in any problem with large expressions to store and output.