The integer labeling method is a simplicialtype homotopy method for solving systems of nonlinear equations with global convergence. Since this method does not require matrix operations, it is very simple and is suited to parallel computation on array processors. However, the computation time of the integer labeling method grows exponentially with the dimension n, because it uses simplicial subdivision and the number of simplices in an n-dimensional rectangle grows with n!. In this paper, we propose an efficient integer labeling method for solving systems of nonlinear equations with partially-separable mappings. Partially-separable mappings appear in various fields of science and engineering, such as nonlinear programming problems. In our method, the number of function evaluations is largely reduced by making use of the partial separability of nonlinear mappings. That is, function values of separable terms need not be evaluated as long as the labeled simplex is moving within the identical rectangle. Hence, as the number of separable terms increases, considerable improvement of the computational efficiency can be achieved.