Output details
11 - Computer Science and Informatics
University of St Andrews
An Automated Approach to Generating Efficient Constraint Solvers
<11>Modern constraint solvers are able to solve efficiently a broad range of problems.
However, this convenience comes at the price of a complex, monolithic architecture that typically requires significant manual tuning by an expert to achieve best performance. This paper describes an entirely novel constraint solver architecture based on generative programming. An input problem is analysed and a constraint solver tailored exactly to its features is synthesised from a library containing a wide variety of components, many unsuitable for inclusion in a generic solver. This removes the need for expert tuning and leads to a dramatic increase in solving efficiency.