Object-oriented Programming Strategies for Numerical Solvers Applied to Continuous Simulation

The study of problems and inconveniences appearing in the Object-oriented (OO) implementation of continuous simulation systems based in Differential-algebraic Equations (DAE) is presented. It was found that the numerical treatment of the equations is crucial to find a design compatible with OO programming practices which lead naturally to implicit schemes. The family of Backward Differential Formulas (BDF) was found particularly appropriate to achieve high levels of software flexibility and reusability. A series of numerical studies were carried out comparing numerical performances with software quality metrics. It was found that BDF implementations improve substantially the software quality, although the computer costs also increase significatively which ultimately calls for pondering the importance of each software characteristic (i.e., modifiability-extensibility vs. calculation time). The equilibrium of this balance is determined by the size of the problem to solve. A utility function is proposed which can be used to determine the optimum choice.