On solving systems of ordinary differential equations on MIMD-computers (Q2458058)

Two aspects of solving systems of ordinary differential equations (ODEs) with initial condition are considered. The first aspect is the solution of the ODEs with approximate initial data and a theorem is given which shows the dependence of the solution on the exact data. For the second aspect of parallelism, a parallel computer with distributed memory is assumed and a parallel implementation of solution methods for ODEs in \(C\) and MPI (message passing interface) with a blockwise distribution of the system is described. Runtime experiments are presented for an Adams method, a Gear method and a Runge-Kutta-method for up to 16 processors.