The goal of this project is to facilitate the use of partial evaluation in broader applications and by a larger class of users. To do so, a comprehensive plan to refine and extend the design and implementation of partial evaluation systems is proposed. The anticipated contributions of this work include: Improving the quality of residual programs by exploring and designing new tools for a programming environment of a partial evaluation system, and by developing new techniques to improve the accuracy of the binding-time analysis. Extending partial evaluation to handle side-effects by developing an automatic binding-time analysis and a program specializer. The first part of this work will focus on the notion of references a la ML. The improvements described above will allow further applications of partial evaluation in some areas as well as the attack upon new ones. Examples from the area of compiler generation are sub-typing, generic functions, exceptions, and task operations. A novel use of partial evaluation in operating systems involves using partial evaluation to derive automatically implementations of operating system components from generic specifications.
