TOWARDS A METHODOLOGY FOR THE DESIGN OF ABSTRACT MACHINES FOR LOGIC PROGRAMMING-LANGUAGES

A number of constituents of a methodology for the systematic design of abstract machines for logic programming languages are described. By means of partial deduction and other program transformation techniques, an interpreter and a source program are ''compiled'' into a new residual program consisting of: 1) ''machine code'' for the source program, and 2) an abstract machine for the machine code. Based upon the appearance of these, the user may choose to refine the original interpreter and repeat the process until the resulting ''machine code'' and the abstract machine satisfy the user's expectations of an abstract machine. We illustrate these principles by reconstructing several of the control instructions of Warren's Abstract Machine. The paper complements previous work of Kursawe, who reconstructed several of the unification instructions using similar techniques.