Two new algorithms for discrete boundary value problems (Q916328)

For the solution of linear multipoint discrete boundary value problems the authors introduce an adjoint system of equations. Using some identities involving solutions of these two sets of equations, backward solution of the originally given problem becomes possible. With further adjustments the storage of large amounts of data can be avoided: forward and backward computations are alternatively used in subintervals given by the gridpoints of the multipoint boundary conditions. Examples illustrating the results show that at least one of the two possible ways of computation gives good results in cases where other methods were numerically unstable.