Warm Start for the Primal and Dual Simplex Solvers |
You can warm start the primal and dual simplex solvers by specifying the option BASIS=WARMSTART. Additionally you need to specify the PRIMALIN= and DUALIN= data sets. The primal and dual simplex solvers start with the basis thus provided. If the given basis cannot form a valid basis, the solvers use the basis generated using their crash techniques.
After an LP model is solved using the primal and dual simplex solvers, the BASIS=WARMSTART option enables you to perform sensitivity analysis such as modifying the objective function, changing the right-hand sides of the constraints, adding or deleting constraints or decision variables, and combinations of these cases. A faster solution to such a modified LP model can be obtained by starting with the basis in the optimal solution to the original LP model. This can be done by using the BASIS=WARMSTART option, modifying the DATA= input data set, and specifying the PRIMALIN= and DUALIN= data sets. Example 10.4 and Example 10.5 illustrate how to reoptimize an LP problem with a modified objective function and a modified right-hand side by using this technique. Example 10.6 shows how to reoptimize an LP problem after adding a new constraint.
The network simplex solver ignores the option BASIS=WARMSTART.
Caution: Since the presolver uses the objective function and/or right-hand-side information, the basis provided by you might not be valid for the presolved model. It is therefore recommended that you turn the PRESOLVER= option off when using BASIS=WARMSTART.