next up previous

4.1.4 SA Computational Considerations

As the procedures controlling the generation and acceptance of new solutions are so simple, the computational cost of implementing the SA algorithm is almost invariably dominated by that associated with the evaluation of the problem functions. It is essential that these evaluations should be performed as efficiently as possible. Sometimes, as, for example, in the Nuclear Fuel Management Case Study presented elsewhere in this project, it is possible to use a Generalized Perturbation Theory method to expedite these calculations. However, the viability of such an approach will obviously be problem-dependent. In general, any efforts to improve performance (reduce run times) should be directed towards exploiting the vectorization or parallelization capabilities of the intended computational platform to accelerate the problem function evaluations. Advice on how this can be achieved can be found in the chapter appropriate to the form of the system equations to be solved.