High-performance algorithms can be found at the heart of many software systems; they often provide the key to effectively solving the computationally difficult problems encountered in the application areas in which these systems are deployed. Examples of such problems include planning, scheduling, timetabling, resource allocation, computer-aided design and software verification. Many of these problems are NP-hard and considered computationally intractable; nevertheless, these `intractable' problems arise in practice, and finding good solutions to them in many cases tends to become more difficult as economic constraints tighten. In most (if not all) cases, the key to solving such computationally challenging problems lies in the use of high-performance heuristic algorithms, that is, algorithms that make use of mechanisms whose efficacy can be demonstrated empirically, yet remains inaccessible to the analytical techniques used for proving theoretical complexity results. High-performance heuristic algorithms are typically constructed in an iterative, manual process in which the designer gradually introduces or modifies components or mechanisms whose performance is then tested by empirical evaluation on one or more sets of benchmark problems. During this iterative design process, the algorithm designer has to make many decisions, ranging from choices of the heuristic mechanisms to be used and the details of these mechanisms to lower-level implementation details, such as data structures. Some of these choices take the form of parameters, whose values are guessed or determined based on limited experimentation. This traditional approach for designing high-performance algorithms can and often does lead to satisfactory results. However, it tends to be tedious and labour-intensive; furthermore, the resulting algorithms are often unnecessarily complicated, yet fail to realise the full performance potential present in the space of designs that can be built using the same und...