In traditional compiler design the work of a compiler is divided into several phases: lexical, syntactical and semantical analysis, optimizations and code generation. For several of these phases generators exist -- most prominently LEX and YACC for generating lexical and syntactical analyzers. A common feature of all generators is that the phase in the compiler is described using a meta-language (e.g. regular expressions or context-free grammars) and that the generator produces the related compiler module. There exist several good textbooks on compiler design. However, all of these books present ready made mappings from source language constructs to target language constructs, the so called translation schemes, instead of deriving them. Hence, the reader is expected to learn how to design code generators by analysing translation schemes as opposed to from first principles. The same is true for abstract machines. Abstract machines are virtual target architectures which support the concepts of the source language. Typically abstract machines are presented together with translation schemes from the source language to the abstract machine language. There is only little work on how translation schemes and abstract machines are designed. The aim of our work is to detect underlying principles that relate abstract machines to programming language semantics, and to automate part of the design process for abstract machines. Thus, we need to ensure that the behaviour of a source program will be maintained by translating it into the abstract machine language, and then applying the abstract machine. The behaviour of a program will depend on itu27s semantics. Often this aspect of a programming language is only described in natural language which is both ambiguous and vague. We shall use formal techniques to describe the meaning of programs in a particular language and to prove that our transformations are correct. In this thesis we concentrate on natural semantics, but we also address action semantics.
展开▼
