This paper presents a methodology for developing the requirements for general purpose scientific computing software. The first step in the methodology is to determine the general purpose scientific software of interest. The second step consists of a commonality analysis on this identified family of general purpose tools to document the terminology, commonalities and variabilities. The commonality analysis is then refined in the third step into a family of specific requirements documents. Besides fixing the variabilities and their binding times, each specific requirements document also shows the relative importance of the different nonfunctional requirements, for instance using the Analytic Hierarchy Process (AHP). The new methodology addresses the challenge of writing validatable requirements by including solution validation strategies as part of the requirements documention. To illustrate the methodology an example is shown of a solver for a linear system of equations.
展开▼