API

multiSolve::pr_multiSolve()

Repeatedly solves small variations of a GMP, when possible in parallel.

Arguments

  • ep_baseGMP –

    Input. The GMP to base variations on.

    To use the 'generate' provide mode, provide the empty element.

    To use the 'modify' provide mode, use the GMP on which modifications can be made. This GMP should contain a feasible solution on position 1 of its solution repository.

  • ep_onProvideGMP –

    Input. A callback procedure which is called when a new instance of the mathematical program needs to be provided.

    This callback procedure implements:

    1. Business logic to update the model identifiers for the next instance.

    2. Logic to either:

      • translate these changes to changes in the base GMP ('modify' provide mode), or

      • create a GMP based on the model identifiers.

    This callback has the following arguments:

    1. ep_gmp: InOut. The GMP to generate or make modifications to.

    2. ep_handle: Output. An element parameter with range Integers. The value of this element parameter will be supplied to the callback procedure ep_onRetrieveSolution when this instance is solved.

    Return value This procedure returns:

    • 1: if the GMP is to be solved.

    • 0: if there is no more work to be started.

  • ep_onRetrieveSolution –

    Input. A callback procedure which is called when a step is solved.

    This callback procedure implements:

    1. Reading the solution from the finished solver session to the model parameters. The procedure multiSolve::pr_storeSolutionInModelVariables can be called for this purpose.

    2. Logic to copy the relevant part of the solution to the application parameters.

    3. Optionally, logic to copy relevant statistics about the solution obtained.

    4. When a next solve step is to be executed, logic to 'modify' or 'generate' a GMP for that next step.

    This callback has the following arguments:

    1. ep_gmp: InOut. If a next step is to be executed, upon return of this callback, this GMP should contain the matrix to be solved.

    2. ep_finishedSolverSession: Input. Optionally, the modeler can use this solver session to obtain solution and statistics about the solve.

      The solution in the solver session can be copied to the model variables using the procedure multiSolve::pr_storeSolutionInModelVariables. Prior to calling this procedure, you can restore index sets used in the model variables when the matrix was generated.

    3. ep_handle: Input. An element parameter with range Integers. The value of this element parameter identifies the instance solved to the business logic.

    4. ep_step: Input. An element parameter with range Integers. Upon completion of the first step, this element parameter has the value '1'.

    Return value This procedure returns:

    • 1: if there is a next step for the OR problem for the data instance at hand.

    • 0: if there are no more steps to be executed for the OR problem at hand.

  • p_maxParallelGMPs –

    Optional, range { 1 .. }, default 1.

    The maximum number of GMPs that are solved in parallel.

  • p_maxThreadsPerSolve –

    Optional, range { 1 .. }, default 1.

    The maximum number of threads used to solve a single GMP.

    Please ensure that p_maxParallelGMPs X p_maxThreadsPerSolve <= the number of logical cores on your computer.

  • p_startingSolutionMethod –

    Optional, range { 0 .. 3 }, default 2. Interpretation:

    1. Do not tamper with the starting solution. This might be a good strategy, because it permits the solver to crash a good corner point to use as starting point when using the simplex method.

    2. Use solution of worker as starting solution. In other words, the starting solution will not be overwritten.

    3. Use solution of base GMP as starting solution.

    4. Use values of model variables as starting solution. This starting solution method permits the ep_onProvideGMP callback to provide a tailored starting solution.

multiSolve::pr_storeSolutionInModelVariables()

Procedure to retrieve the solution from a finished solver session and store that solution in the model variables. To be called in a ep_onRetrieveSolution callback.

Arguments

  • ep_finishedSolverSession –

multiSolve::pr_cleanup()

Procedure to cleanup a GMP and solver session prior to regenerating one. To be called in a ep_onRetrieveSolution callback only when the provide mode is 'Generate'.

Arguments

  • ep_finishedGmp – The GMP that is no longer needed.

  • ep_finishedSolverSession – The solver session that is no longer needed.