Date: March 1977
Creator: Byrne, G. D.; Hindmarsh, A. C.; Jackson, Kenneth R. & Brown, H. Gordon
Description: This is a sequel to the paper ''A comparison of two ODE codes: GEAR and EPISODE,'' and is concerned with the testing of two superficially similar ODE packages, GEAR and EPISODE. Fourteen basic test problems, some with several cases, are the basis for the testing. These problems represent several types-nonlinear systems with real and complex eigenvalues, linear systems with varied diagonal dominance, linear scalar problems, stiff and non-stiff problems, chemical kinetics with and without diurnal effect, and systems arising from the use of the numerical method of lines. Some problems are included in order to examine the options and error returns. The test results are presented in two forms: raw output and a comparative display of operation counts and of timings for the best method in the GEAR package and the best method in the EPISODE package. This approach allows a comparison of the consequences of the fixed-step interpolate strategy (GEAR) for changing step size against the truly variable step size strategy (EPISODE). It is concluded that EPISODE is generally faster than GEAR for problems involving wave fronts or transients on the interior of the interval of integration. For linear or simply decaying problems, these roles are usually reversed.
Contributing Partner: UNT Libraries Government Documents Department