From operational semantics to abstract machines: preliminary results

  • Authors:
  • John Hannan;Dale Miller

  • Affiliations:
  • Department of Computer and Information Science, University of Pennsylvania, Philadelphia, PA;Department of Computer and Information Science, University of Pennsylvania, Philadelphia, PA

  • Venue:
  • LFP '90 Proceedings of the 1990 ACM conference on LISP and functional programming
  • Year:
  • 1990

Quantified Score

Hi-index 0.00

Visualization

Abstract

The operational semantics of functional programming languages is frequently presented using inference rules within simple meta-logics. Such presentations of semantics can be high-level and perspicuous since meta-logics often handle numerous syntactic details in a declarative fashion. This is particularly true of the meta-logic we consider here, which includes simply typed &lgr;-terms, quantification at higher types, and &bgr;-conversion. Evaluation of functional programming languages is also often presented using low-level descriptions based on abstract machines: simple term rewriting systems in which few high-level features are present. In this paper, we illustrate how a high-level description of evaluation using inference rules can be systematically transformed into a low-level abstract machine by removing dependencies on high-level features of the meta-logic until the resulting inference rules are so simple that they can be immediately identified as specifying an abstract machine. In particular, we present in detail the transformation of two inference rules specifying call-by-name evaluation of the untyped &lgr;-calculus into the Krivine machine, a stack-based abstract machine that implements such evaluation. The initial specification uses the meta-logic's &bgr;-conversion to perform substitutions. The resulting machine uses de Bruijn numerals and closures instead of formal substitution. We also comment on a similar construction of a simplified SECD machine implementing call-by-value evaluation. This approach to abstract machine construction provides a semantics-directed method for motivating, proving correct, and extending such abstract machines.