Driving in the Jungle

  • Authors:
  • Jens P. Secher

  • Affiliations:
  • -

  • Venue:
  • PADO '01 Proceedings of the Second Symposium on Programs as Data Objects
  • Year:
  • 2001

Quantified Score

Hi-index 0.00

Visualization

Abstract

Collapsed jungle evaluation is an evaluation strategy for functional programs that can give super-linear speedups compared to conventional evaluation strategies such as call-by-need. However, the former strategy may incur administrative evaluation overhead. We demonstrate how this overhead can be eliminated by transforming the program using a variation of positive supercompilation in which the transformation strategy is based on collapsed jungle evaluation. In penetrating the constantfactor barrier, we seem to be close to establishing a transformation technique that guarantees the efficiency of the transformed program. As a spin-off, we clarify the relationship between call-by-name, call-by-need and collapsed-jungle evaluation, showing that all three can be expressed as instances of a common semantics in which the variations -- differing only in efficiency -- are obtained by varying the degree of sharing in a DAG representation.