The Large-Scale GRG Solver maintains a sparse representation of the Jacobian matrix (i.e. the matrix of partial derivatives of the objective and constraints with respect to the decision variables, used to determine search directions), in which zero partial derivatives are omitted. If the sparsity pattern of the Jacobian matrix is not known when the solution process starts, the Solver constructs the matrix based on the partial derivatives at the starting point you supply. Some partial derivatives that are zero at the starting point may become nonzero as the Solver moves to new trial points. The Large-Scale GRG Solver can accommodate a large, but limited number of these new nonzero partial derivatives; if the limit is exceeded, this message or result code is returned.
You may be able to work around this problem by restarting the Solver at a different point – for example at the ending point reached when this message appears. A better approach is to specify the sparsity pattern of the Jacobian matrix at the beginning of the solution process. In Analytic Solver Comprehensive and Analytic Solver Optimization, the PSI Interpreter can determine the sparsity pattern automatically. In Solver SDK Platform, if you’ve written your own Evaluator for the objective and constraints, you can specify the sparsity pattern by setting the Model object AllGradDepend property, or by calling the SolverModAllGradDependSet procedural API function. This will ensure that the Solver will not encounter the limit on additional nonzero Jacobian entries.