Mix: A self-applicable partial evaluator for experiments in compiler generation

Abstract

Abstract, The program transformation principle called partial evaluation has interesting applications in compilation and compiler generation. Self-applicable partial evaluators may be used for transforming interpreters into corresponding compilers and even for the generation of compiler generators. This is useful because interpreters are significantly easier to write than compilers, but run much slower than compiled code. A major difficulty in writing compilers (and compiler generators) is the thinking in terms of distinct binding times: run time and compile time (and compiler generation time). The paper gives an introduction to partial evaluation and describes a fully automatic though experimental partial evaluator, called mix, able to generate stand-alone compilers as well as a compiler generator. Mix partially evaluates programs written in Mixwell, essentially a first-order subset of statically scoped pure Lisp. For compiler generation purposes it is necessary that the partial evaluator be self-applicable. Even though the potential utility of a self-applicable partial evaluator has been recognized since 1971, a 1984 version of mix appears to be the first successful implementation. The overall structure of mix and the basic ideas behind its way of working are sketched. Finally, some results of using a version of mix are reported.Since the early 1970s it has been k n o w n that in theory, the p r o g r a m transformation principle called partial evaluation can be used for compiling and compiler generation, and even for the a u t o m a t i c generation o f a compiler generator. A partial evaluator able to generate stand-alone compilers and compiler generators had not, however, been successfully implemented before 1984 when the first mix system was brought to work at the University o f Copenhagen.In this paper we discuss partial evaluation and its applications to compiler generation and sketch the partial evaluator we developed, called the mix system. The results we report are sufficiently r e m a r k a b l e to justify further research into using partial evaluation for compiler generation purposes. We also mention other applications. The description here is essentially a snapshot of the mix system and its applications as of early 1987.A partial evaluator m a y be thought of as a " s m a r t interpreter." If an ordinary interpreter is given a p r o g r a m and only part o f this p r o g r a m ' s input data, it will leave * Current address:

Occam's razor in metacomputation: the notion of a perfect process tree

“…This implements what is known as constant propagation , and corresponds to first-order partial evaluators based on constant prop agation (e.g. [12]). …”

“…It was noticed [12] that the test const? does not require the values proper and may be approximated in a separate pre-processing phase, called binding-time analysis .…”

Sharing of Computations

“…Af teknikker kan naevnes memoisering [Kho90]; udfold/fold transformationer [PP91b]; graf-baseret implementation af "lazy" evaluering [Jon87] og partiel evaluering [JSS89]. At disse metoder forbedrer effektiviteten skyldes at nogle beregninger deles, så de kun skal gøresén gang.…”

“…These include memoization [Kho90]; unfold/fold transformations [PP91b]; graphbased implementation of lazy evaluation [Jon87] and partial evaluation [JSS89]. The efficiency improvement caused by these techniques all are due to the fact that some computations are shared, i.e.…”

Formal Aspects of Polyvariant Specialization

“…These annotations are then blindly followed in the subsequent specialization phase. The utility of this division was demonstrated by Jones's group [34] in connection with compiler-generation by selfapplication of the specializer, an idea originally conceived by Futamura [24].…”