Estimating the cost of output in ODE codes
The author is concerned with estimating the algorithmic cost of output in explicit, fixed order Runge-Kutta codes and Rosenbrock codes for solving the initial value problem. This cost consists of the additional computation resulting from the requirement to return a solution value at a particular point forcing the code to use a smaller and therefore less efficient step length than it would otherwise have selected. Any cost associated with returns from the subroutine to report results and with checks of input on subsequent calls is, however, not considered. A method of assessing the algorithmic cost is devised, and shown by numerical examples to be a reasonable one. The author suggests that codes of the type considered would be significantly enhanced if they were to advise the user when the method of use is resulting in high output cost.
This page was built for publication: Estimating the cost of output in ODE codes
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3316489)