Constraint solvers: An empirical evaluation of design decisions

Contemporary constraint solvers are very complex software systems. Each one of the many available today has its own characteristics, its own design decisions that the implementers made, and its own philosophy. The traits of a solver which will affect the performance for a particular problem class or instance often cannot be determined easily. Picking a particular solver is therefore a difficult task which requires specialist knowledge about each solver and is likely to have a significant impact on performance. On top of that, each solver has different ways of modelling problems. Not only do users need experience with a particular solver to model a problem in a way that enables it to be solved efficiently, but it is also hard to objectively compare solvers. This paper studies a small selection of constraint solvers and assesses their performance on problem models which were made as similar as possible.