ptolemy.domains.sdf.optimize

Class OptimalScheduleFinder

java.lang.Object

ptolemy.domains.sdf.optimize.OptimalScheduleFinder

public class OptimalScheduleFinder
extends java.lang.Object

Class comments

An OptimalScheduleFinder encapsulates an algorithm to find an optimized schedule. In particular it implements a simple state space exploration algorithm to find a minimum buffer size schedule.

See OptimizingSDFDirector, OptimizingSDFScheduler and BufferingProfile for more information.

Since:

Ptolemy II 10.0

Version:

$Id$

Author:

Marc Geilen

See Also:

OptimizingSDFDirector, OptimizingSDFScheduler, BufferingProfile

Pt.AcceptedRating:

Red ()

Pt.ProposedRating:

Red (mgeilen)

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
protected static class	OptimalScheduleFinder._Actor A model of an actor.
protected static class	OptimalScheduleFinder._Channel the state of the channel in the channel array has one integer at stateIndex indicating the number of tokens in the channel and another integer at stateIndex+1 indicating how many consumer are still to read the token I need to remember per receiver how many tokens there are for that receiver.
protected static class	OptimalScheduleFinder._ListOfActors A list of actors, derived from LinkedList.
protected static class	OptimalScheduleFinder._ListOfChannels A list of channels, based on LinkedList.
protected static class	OptimalScheduleFinder._ListOfPorts A list of ports, based on LinkedList.
protected static class	OptimalScheduleFinder._Port A port of an actor, connected to a channel.
protected static class	OptimalScheduleFinder._SetOfStates A set of states, based on HashSet.
protected static class	OptimalScheduleFinder._SortedSetOfStates A sorted set of states.
protected static class	OptimalScheduleFinder._State State models a global state of the SDF graph and remembers the actor that was fired to reach it.
protected static class	OptimalScheduleFinder._StateComparator An abstract super class for Comparators to maintain a sorted list of states.
protected static class	OptimalScheduleFinder._StateComparatorLowestValue A Comparator to maintain a sorted list of states, sorted on their value.
protected static class	OptimalScheduleFinder._StateComparatorMaximumProgress A Comparator to maintain a sorted list of states, sorted on their progress to the final state.
protected static class	OptimalScheduleFinder._TwoWayHashMap A two-way hash map provides fast lookup in both directions of a bijective function between objects.

Constructor Summary

Constructors

Constructor and Description
OptimalScheduleFinder(OptimizingSDFScheduler scheduler, OptimizingSDFDirector.OptimizationCriteria criterion) Construct an instance of the OptimalScheduleFinder.

Method Summary

Modifier and Type	Method and Description
protected void	_instantiateAnalysisModel(java.util.Map firingVector) Instantiate the analysis model from the core model.
Schedule	makeSchedule(java.util.Map firingVector) Make a schedule using an exhaustive BFS-like optimizing algorithm.
Schedule	makeScheduleGreedy(java.util.Map firingVector) Make a schedule using a greedy (non-optimizing algorithm).

Methods inherited from class java.lang.Object

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

Constructor Detail

OptimalScheduleFinder

public OptimalScheduleFinder(OptimizingSDFScheduler scheduler,
                             OptimizingSDFDirector.OptimizationCriteria criterion)

Construct an instance of the OptimalScheduleFinder. Creates an object associated with the OptimizingSDFSchedule scheduler and using the optimization criterion criterion to find an optimized schedule.

Parameters:

scheduler - scheduler

criterion - optimization criterion

Method Detail

makeScheduleGreedy

public Schedule makeScheduleGreedy(java.util.Map firingVector)

Make a schedule using a greedy (non-optimizing algorithm).

Parameters:

firingVector - repetition vector

Returns:

the computed schedule

makeSchedule

public Schedule makeSchedule(java.util.Map firingVector)

Make a schedule using an exhaustive BFS-like optimizing algorithm.

Parameters:

firingVector - repetition vector

Returns:

the computed schedule

_instantiateAnalysisModel

protected void _instantiateAnalysisModel(java.util.Map firingVector)
                                  throws IllegalActionException

Instantiate the analysis model from the core model.

Parameters:

firingVector - contains repetition vector information

Throws:

IllegalActionException - if model information inconsistent