org.chocosolver.solver

Class MasterSolver

java.lang.Object

org.chocosolver.solver.thread.AbstractParallelMaster<SlaveSolver>

org.chocosolver.solver.MasterSolver

public class MasterSolver
extends AbstractParallelMaster<SlaveSolver>

A MasterSolver which enables multi-thread resolution. The main idea of that class is to solve the same problem with various search strategies, and to share few possible information. A problem declared in a solver is duplicated into solvers. Each of them is then configured and run into a single thread. On satisfaction problem, the first solver who finds a solution advises the others. On optimisation problem, the best value found so far is shared among all the solvers.

The expected ways to solve a problem using MasterSolver is:

     Solver solver = new Solver();
     // declare the variables and constraints
     // and an optional search strategy
     //...
     // Then create the master-solver
     MasterSolver ms = new MasterSolver();
     // duplicate the solver into 4 solvers (1+3)
     ms.populate(solver, 3);
     // configure the search strategies (optional, but recommended)
     ms.configureSearches();
     // Finally, solve the problem
     ms.findSolution();
 

Since:

24/10/14

Version:

choco

Author:

Charles Prud'homme

Field Summary

Fields

Modifier and Type	Field and Description
protected Solver[]	solvers Pool of solvers to drive

Fields inherited from class org.chocosolver.solver.thread.AbstractParallelMaster

nbWorkingSlaves, slaves, wait

Constructor Summary

Constructors

Constructor and Description
MasterSolver()

Method Summary

Methods

Modifier and Type	Method and Description
void	declare(Solver... solvers) An alternative to populate(Solver, int) where the initial model has already been duplicated.
void	declareSettings(Settings... settings) Declare a specific Settings to each solver.
void	findOptimalSolution(ResolutionPolicy policy, IntVar objective) Attempts optimize the value of the objective variable w.r.t. to the optimization policy.
boolean	findSolution() Attempts to find the first solution of the declared problem, running all solvers declared.
Solver[]	getSolvers() Return the solvers to drive
boolean	hasReachedLimit() Returns information on the completeness of the search process.
ESat	isFeasible() Returns information on the feasibility of the current problem defined by the solver.
void	populate(Solver model, int n) Make n-1 copies of the current model, the first solver of the array is the one given in parameter.
void	wishGranted() A slave notify the master that he fulfilled his task

Methods inherited from class org.chocosolver.solver.thread.AbstractParallelMaster

distributedSlavery, sequentialSlavery

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

solvers

protected Solver[] solvers

Pool of solvers to drive

Constructor Detail

MasterSolver

public MasterSolver()

Method Detail

populate

public void populate(Solver model,
            int n)

Make n-1 copies of the current model, the first solver of the array is the one given in parameter. The solvers created are available thanks to the Thus, there is n solvers in the returned array Note that only the variables and the constraints are duplicated.

Parameters:

model - the model to duplicate

n - number of of copies to make.

declare

public void declare(Solver... solvers)

An alternative to populate(Solver, int) where the initial model has already been duplicated. The first solver in the array needs to be the original one (required for optimization problem)

Parameters:

solvers - the set of solvers to drive.

getSolvers

public Solver[] getSolvers()

Return the solvers to drive

Returns:

the array of solvers

declareSettings

public void declareSettings(Settings... settings)

Declare a specific Settings to each solver. Calling this method is highly recommended to, at least, configure the search strategies for each solver.

Parameters:

settings - array of settings

isFeasible

public ESat isFeasible()

Returns information on the feasibility of the current problem defined by the solver.

Possible back values are:
- ESat.TRUE: a solution has been found,
- ESat.FALSE: the CSP has been proven to have no solution,
- ESat.UNDEFINED: no solution has been found so far (within given limits) without proving the unfeasibility, though.

Returns:

an ESat.

hasReachedLimit

public boolean hasReachedLimit()

Returns information on the completeness of the search process.

A call to isFeasible() may provide complementary information.

Possible back values are:

- false : the resolution is complete and
* findSolution(): a solution has been found or the CSP has been proven to be unsatisfiable.
* findOptimalSolution(ResolutionPolicy, IntVar) : the optimal solution has been found and proven to be optimal, or the CSP has been proven to be unsatisfiable.
- true: the resolution stopped after reaching a limit.

findSolution

public boolean findSolution()

Attempts to find the first solution of the declared problem, running all solvers declared. The first one which finds a solution stops the process.

Returns:

true if and only if a solution has been found.

findOptimalSolution

public void findOptimalSolution(ResolutionPolicy policy,
                       IntVar objective)

Attempts optimize the value of the objective variable w.r.t. to the optimization policy. Restores the best solution found so far (if any)

Parameters:

policy - optimization policy, among ResolutionPolicy.MINIMIZE and ResolutionPolicy.MAXIMIZE

objective - the variable to optimize, the variable must be declared in solvers[0].

wishGranted

public void wishGranted()

Description copied from class: AbstractParallelMaster

A slave notify the master that he fulfilled his task

Overrides:

wishGranted in class AbstractParallelMaster<SlaveSolver>