This award supports research in extracting properties from action semantics definitions and in improving programs by partial evaluation. Results from the two areas will be combined into a methodology for compiler construction based on staging semantic definitions automatically. Such a methodology will ultimately be implemented as a part of a programming language design and analysis workbench." Property extraction from action semantics builds on earlier research on models and type inference for programs coded in action semantics notation. Type inference algorithms will be extended so that analysis can be undertaken on language definitions per si to elicit key properties regarding block-structuredess and strong typing. Such information is crucial to the automated construction of correct and efficient compilers. Research on partial evaluation has centered on effective program improvement based on partial knowledge of the nature of a program's input, for example, its type. Application of partial evaluation to compiler generation from interpreter definitions has proved fruitful. The work is to be extended so that it formalizes and automates the "staging" of and interpretive specification into a static semantics processor and a dynamic semantics processor.
