fr.inria.aoste.kpassa.core

Class KPassa

java.lang.Object

fr.inria.aoste.kpassa.core.KPassa

public final class KPassa
extends java.lang.Object

This is the main class of the application KPassa. In order to use KPassa through its API, one should instantiate a KPassa object and use the available methods. For example:

KPassa kpassa = new KPassa("",DFPNType.SDF);

SDFGraph sdf = kpassa.getSdfGraph();

SDFVertex[] vertexTab = new SDFVertex[4];
SDFArc[] arcTab = new SDFArc[4];

vertexTab[0]=sdf.addVertex(0,1);
vertexTab[1]=sdf.addVertex(2,1);
vertexTab[2]=sdf.addVertex(2,0);
vertexTab[3]=sdf.addVertex(0,2);

arcTab[0]=sdf.addArc(vertexTab[0].getOutPorts().get(0), vertexTab[1].getInPorts().get(0), 0);
arcTab[1]=sdf.addArc(vertexTab[1].getOutPorts().get(0), vertexTab[2].getInPorts().get(0), 0);

vertexTab[1].getInPorts().get(1).setWeight(2);
arcTab[2]=sdf.addArc(vertexTab[3].getOutPorts().get(0), vertexTab[1].getInPorts().get(1), 0);

vertexTab[2].getInPorts().get(1).setWeight(2);
arcTab[3]=sdf.addArc(vertexTab[3].getOutPorts().get(1), vertexTab[2].getInPorts().get(1), 0);

kpassa.checkLiveness();

SDFGraph cfg = kpassa.buildControlFlowGraph().getSdfGraph();


Author:

AOSTE Team

See Also:

KPassaListener, MarkedGraphListener

Constructor Summary

Constructors

Constructor and Description
KPassa(DFPNType graphType) A KPassa object contains a single graph of type graphType.
KPassa(java.lang.String graphName, DFPNType graphType) A KPassa object contains a single graph of type graphType and a name graphName.

Method Summary

Methods

Modifier and Type	Method and Description
void	addKPassaListener(KPassaListener listener) This method adds a KPassaListener to the list of listeners
KPassa	buildControlFlowGraph() This method builds the control flow graph of the SDF graph.
java.util.Map<? extends SDFVertex,java.lang.Integer>	checkBalancedEquation() This method checks the balanced equation on the SDF graph
boolean	checkLiveness() Check wether the graph is live or not i.e.
java.util.Map<MGVertex,BinaryWord>	computeASAPExecution() To compute the ASAP schedule of a graph.
void	computeBalancedScheduling() Perform the balanced scheduling algorithm.
java.util.Map<MGArc,java.lang.Integer>	computeMaxBuffer(FireableStrategy _fireableStrategy, FiringStrategy _simuStrategy) This algorithm runs a single period of execution of a graph and returns the required size of the places for this execution.
void	computePeriodicity() Find the periodicity of the marked graph The result can be accessed from MarkedGraphImpl.getPeriod() and MarkedGraphImpl.getPeriodicity()
java.util.Set<StronglyConnectedComponent<MGVertex,MGArc>>	computeSSC() Find the strongly connected components in the marked graph
void	equalize() Perform the equalization algorithm
void	exportToLucyn() This method exports the current marked graph to the lucy-n format. Lucy-n is developed by Louis Mandel from the INRIA/ENS team PARKAS.
AsynchMGImpl	getAsynchronousMarkedGraph() To access to the graph as an AsynchMG It works only when type is DFPNType.MG
KRGGraph	getKRGGraph() To access to the graph as an KRG It works only when type is DFPNType.KRG
MarkedGraph<?>	getMarkedGraph() To access to the graph without knowing its exact type.
SDFGraph	getSdfGraph() To access to the graph as an SDF It works only when type is DFPNType.SDF
SynchronousMarkedGraph	getSynchronousMarkedGraph() This method returns the graph that is currently used.
void	isSteadyState() Check whether the current marking is reachable from itself by following an ASAP simulation.
void	listCycle() List the cycle in the marked graph The list can be accessed from MarkedGraphImpl.getCycleList()
void	loadBalancedMarking() Load the balanced marking computed by the balanced scheduling algorithm.
void	loadInitialMarking() Load the initial marking of the marked graph.
void	loadInitialMarkingAfterEqualization() Load the marking of the marked graph after the equalization algorithm.
void	runSimulation() Run an automatic ASAP simulation of the marked graph until it has reached a marking that has already been encountered.
void	validate() This method is used to validate the connections of the graphs.

Methods inherited from class java.lang.Object

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

Constructor Detail

KPassa

public KPassa(DFPNType graphType)

A KPassa object contains a single graph of type graphType.

Parameters:

graphType - the type of the graph to create

KPassa

public KPassa(java.lang.String graphName,
      DFPNType graphType)

A KPassa object contains a single graph of type graphType and a name graphName.

Parameters:

graphName - the name of the graph

graphType - the type of the graph

Method Detail

getKRGGraph

public KRGGraph getKRGGraph()

To access to the graph as an KRG It works only when type is DFPNType.KRG

Returns:

the graph as a KRG

getSdfGraph

public SDFGraph getSdfGraph()

To access to the graph as an SDF It works only when type is DFPNType.SDF

Returns:

the graph as an SDF

getAsynchronousMarkedGraph

public AsynchMGImpl getAsynchronousMarkedGraph()

To access to the graph as an AsynchMG It works only when type is DFPNType.MG

Returns:

the graph as an AsynchMG

getSynchronousMarkedGraph

@API
public SynchronousMarkedGraph getSynchronousMarkedGraph()

This method returns the graph that is currently used. The class SynchronousMarkedGraphImpl contains the methods required to build and modify the graph. It also contains the accessors to the marking and scheduling computed by the scheduling algorithm.

Returns:

The synchronous marked graph currently in use.

getMarkedGraph

public MarkedGraph<?> getMarkedGraph()

To access to the graph without knowing its exact type.

Returns:

the marked graph

addKPassaListener

@API
public void addKPassaListener(KPassaListener listener)

This method adds a KPassaListener to the list of listeners

Parameters:

listener - to add.

exportToLucyn

@API
public void exportToLucyn()

This method exports the current marked graph to the lucy-n format.
Lucy-n is developed by Louis Mandel from the INRIA/ENS team PARKAS.

See Also:

LucyNExporter

listCycle

@API
public void listCycle()

List the cycle in the marked graph The list can be accessed from MarkedGraphImpl.getCycleList()

computeSSC

@API
public java.util.Set<StronglyConnectedComponent<MGVertex,MGArc>> computeSSC()

Find the strongly connected components in the marked graph

Returns:

the set of strongly connected components

computePeriodicity

@API
public void computePeriodicity()

Find the periodicity of the marked graph The result can be accessed from MarkedGraphImpl.getPeriod() and MarkedGraphImpl.getPeriodicity()

equalize

@API
public void equalize()

Perform the equalization algorithm

computeBalancedScheduling

@API
public void computeBalancedScheduling()

Perform the balanced scheduling algorithm. It computes the full schedule of the marked graph. See SynchronousMarkedGraphImpl for details on how to access the computed markings and schedule.

loadInitialMarking

@API
public void loadInitialMarking()

Load the initial marking of the marked graph. Whatever is the current marking of the graph,
it replaces it with the initial one.

loadInitialMarkingAfterEqualization

@API
public void loadInitialMarkingAfterEqualization()

Load the marking of the marked graph after the equalization algorithm. The difference with the initial marking is that some arcs have a bigger latency and so the arc's marking is augmented with 0s.

loadBalancedMarking

@API
public void loadBalancedMarking()

Load the balanced marking computed by the balanced scheduling algorithm.

runSimulation

@API
public void runSimulation()

Run an automatic ASAP simulation of the marked graph until it has reached a marking that has already been encountered. The computed scheduling appears in every Vertex objects.

isSteadyState

@API
public void isSteadyState()

Check whether the current marking is reachable from itself by following an ASAP simulation.

validate

public void validate()

This method is used to validate the connections of the graphs. Every port of a computing vertex should be connected. In case of a KRG, the outputs of a select node or the inputs of a merge node can be free.

checkBalancedEquation

public java.util.Map<? extends SDFVertex,java.lang.Integer> checkBalancedEquation()
                                                                           throws java.lang.Exception

This method checks the balanced equation on the SDF graph

Returns:

The execution vector of the SDF agents

Throws:

java.lang.Exception

checkLiveness

public boolean checkLiveness()
                      throws KPassaException

Check wether the graph is live or not i.e. if the graph has enough token to start running.

Returns:

true when the graph is live, false otherwise

Throws:

KPassaException

computeMaxBuffer

public java.util.Map<MGArc,java.lang.Integer> computeMaxBuffer(FireableStrategy _fireableStrategy,
                                                      FiringStrategy _simuStrategy)

This algorithm runs a single period of execution of a graph and returns the required size of the places for this execution.

Parameters:

_fireableStrategy - the fireable strategy considered for this simulation

_simuStrategy - the firing strategy considered for this simulation

Returns:

The size for each place (SDFArc)

buildControlFlowGraph

public KPassa buildControlFlowGraph()
                             throws KPassaException

This method builds the control flow graph of the SDF graph.

Returns:

another KPAssa object that contains the CFG

Throws:

KPassaException

computeASAPExecution

public java.util.Map<MGVertex,BinaryWord> computeASAPExecution()
                                                        throws KPassaException

To compute the ASAP schedule of a graph. Concerning SDF and KRG, this method schedule a single period of execution.

Returns:

the schedules

Throws:

KPassaException