An Executable Intermediate Representation for Incremental Global Compilation of Prolog

The programming language Prolog has built-in predicates which enable the modiication of a program at runtime. This makes the global compilation of Pro-log programs a complex task. This paper presents an intermediate representation which enables fast incre-mental global compilation of Prolog. This intermediate representation is based on the Vienna Abstract Machine (VAM) with its two versions, the VAM 2P and the VAM 1P. The VAM 2P has two instruction pointers and is well suited for intermediate code interpretation. Since the VAM 1P has only one instruction pointer it can be expanded to machine code. Abstract interpretation can be used to collect information to generate specialized machine code which reduces the code size and allows faster execution. The VAM 2P has been modiied to an abstract machine for abstract interpretation, the VAM AI. This VAM AI is an intermediate representation which can be used to reconstruct the source program, which can be executed for dataaow analysis and which can be translated into VAM 1P instructions and expanded further into machine code for RISC processors.