Towards Native Code Generation in CLP

Native code generation and global analysis gave substantial speed-ups in the case of Prolog. In order to generate native code in CLP we must first generate native code for the various constraint solving algorithms. The paper presents a simple scheme for the native code compilation of the incremental linear equation solver used in most CLP systems. Starting from a high level description of this constraint solving algorithm we decompose the algorithm to the level of a RISC abstract machine. Finally, we present preliminary measurements of the impact of abstract interpretation on the efficiency of the generated code. 1 Introduction Constraint Logic Programming [15, 26, 16] provides an elegant and powerful framework for the cooperation of various constraint solving algorithms. In the future, a major requirement from a CLP system will be the ability to easily incorporate a new constraint solving algorithm. This trend is underlined by the increasing number of algorithms (besides stan..