[rules-dev] 3 issues with true modify, but PatientAdmissionScheduleApp does work
Geoffrey De Smet
ge0ffrey.spam at gmail.com
Sun Mar 7 13:01:50 EST 2010
Edson,
> I believe fixed the infinite loop
>> problem. Would you please take a look if it works for you?
Yes, it's no longer there, thanks :)
Examination now suffers from the same NPE as the NQueens smoke test.
On one PAS testdata I actually had 5%, but on another PAS testdata only 1%.
I believe the cloning in accumulates fix will allow me to do the score
calculation as it should be done (one rule instead of two rules)
and also further help performance.
I 'll run some experiments in that direction as soon as you give me the
sign that cloning in accumulates has been fixed.
With kind regards,
Geoffrey De Smet
Edson Tirelli schreef:
>
> Geoffrey,
>
> I just committed some code that I believe fixed the infinite loop
> problem. Would you please take a look if it works for you? I will take a
> look at the other problems now... thanks for reporting that.
>
> Regarding the performance, the main benefit of true modify is not a
> straight gain on performance, but the reduction of the garbage
> collection stress. Just to explain, the code now is actually more
> complex, but the idea is that the algorithm pays off by instantiating a
> lot less objects now, specially LeftTuples. I would not be surprised if
> the 4% gain you see is due to less object instantiation+garbage
> collection work. This is obviously dependent on the use case, but one
> really interesting statistic to see would be if you can profile how many
> instances of LeftTuple were created with the previous algorithm compared
> to now.
>
> Having said that, there are some performance improvements for
> specific cases, like on accumulate that before would frequently clone
> lists of tuples and iterate and now it avoids both the cloning and the
> iteration on specific cases. For those cases, we expect gains of 30% on
> the node, but that is just the node, not the whole kbase.
>
> Also, as you mentioned, true modify opens the door for several other
> optimizations that we will work on as soon as it is stable again.
>
> Thanks,
> Edson
>
> 2010/3/7 Geoffrey De Smet <ge0ffrey.spam at gmail.com
> <mailto:ge0ffrey.spam at gmail.com>>
>
> Found another issue with the true modify changes: 3 of my examples don't
> work (NQueens, Travelingtournament, Examination), due to distinct
> causes.
>
> PatientAdmissionScheduling does work and it's about 4% faster on first
> sight.
> It's a linear improvement, not an exponential improvement.
> Nothing world shocking (yet), but every little bit helps :)
> Edson and Mark believe that true modify allows for new optimization,
> which I believe might be an exponential improvement.
>
> Once examination works I'll test if I can remove the TMS workaround I
> needed before.
> This will be a huge performance boost, but only for that specific
> example.
>
>
> Here's the bug list:
>
> 1) Infinite loop in
> org.drools.core.util.LeftTupleList.get(LeftTupleList.java:102)
>
> 2) NPE in
> org.drools.core.util.RightTupleList.matches(RightTupleList.java:190)
> because this.first is null in
> return this.hashCode == tupleHashCode && this.index.equal(
> this.first.getFactHandle().getObject(),
>
> tuple );
>
> 3) NPE in
> org.drools.reteoo.SingleLeftTupleSinkAdapter#propagateModifyChildLeftTuple,
> line 198:
> the leftTuple or the sink is null
> childLeftTuple.getLeftTupleSink().modifyLeftTuple(
> childLeftTuple,
> context,
>
> workingMemory );
>
> Here are the stacktraces:
>
> 1)
>
> 1.
> To run:
> 2.
> module: drools-planner-examples
> 3.
> main class:
> org.drools.planner.examples.examination.app.ExaminationBenchmarkApp
> 4.
> VM params: -Xms256m -Xmx512m -server
> 5.
> program params: short
> 6.
> working dir:
>
> /home/ge0ffrey/projects/jboss/drools/drools-planner/drools-planner-examples
> 7.
>
> 8.
>
> 9.
> More logging? open log4j-benchmark.xml and turn on debug logging
> for org.drools.planner
> 10.
>
> 11.
>
> 12.
> 2010-03-03 19:47:34,625 [main] INFO Solving with random seed (0).
> 13.
> 2010-03-03 19:47:34,626 [main] DEBUG Solution is already
> initialized.
> 14.
> 2010-03-03 19:47:34,752 [main] INFO Initialization time spend
> (129) for score (0hard/-8154soft). Updating best solution and best
> score.
> 15.
> 2010-03-03 19:47:34,948 [main] DEBUG Move (489 {D180|S47} @ 17
> {D210} + 4 {C77} => 25 {D210}) with score (-47hard/-8190soft) and
> acceptChance (1.0).
> 16.
> 2010-03-03 19:47:34,972 [main] DEBUG Move (373 {D180|S157} @
> 23 {D210} + 0 {C260} => 16 {D210}) with score (-158hard/-8219soft)
> and acceptChance (1.0).
> 17.
> 2010-03-03 19:47:34,985 [main] DEBUG Move (530 {D120|S152} @
> 16 {D210} + 0 {C260} => 39 {D210}) with score (-153hard/-8481soft)
> and acceptChance (1.0).
> 18.
> 2010-03-03 19:47:35,001 [main] DEBUG Move (132 {D120|S61} @ 53
> {D210} + 5 {C65} => 24 {D210}) with score (-64hard/-8243soft) and
> acceptChance (1.0).
> 19.
> 2010-03-03 19:47:35,011 [main] DEBUG Move (553 {D180|S169} @
> 51 {D210} + 0 {C260} => 37 {D210}) with score (-168hard/-8323soft)
> and acceptChance (1.0).
> 20.
> 2010-03-03 19:52:03
> 21.
> Full thread dump Java HotSpot(TM) Server VM (14.2-b01 mixed mode):
> 22.
>
> 23.
> ...
> 24.
>
> 25.
> "main" prio=10 tid=0x08b2b800 nid=0x1104 runnable [0xb6aa6000]
> 26.
> java.lang.Thread.State: RUNNABLE
> 27.
> at
>
> org.drools.common.DefaultFactHandle.hashCode(DefaultFactHandle.java:136)
> 28.
> at org.drools.reteoo.LeftTuple.hashCode(LeftTuple.java:445)
> 29.
> at org.drools.reteoo.LeftTuple.equals(LeftTuple.java:461)
> 30.
> at
> org.drools.core.util.LeftTupleList.get(LeftTupleList.java:102)
> 31.
> at
> org.drools.core.util.LeftTupleList.contains(LeftTupleList.java:96)
> 32.
> at
>
> org.drools.reteoo.EvalConditionNode.modifyLeftTuple(EvalConditionNode.java:235)
> 33.
> at
>
> org.drools.reteoo.SingleLeftTupleSinkAdapter.propagateModifyChildLeftTuple(SingleLeftTupleSinkAdapter.java:209)
> 34.
> at
> org.drools.reteoo.JoinNode.modifyRightTuple(JoinNode.java:221)
> 35.
> at
> org.drools.reteoo.BetaNode.modifyObject(BetaNode.java:317)
> 36.
> at
>
> org.drools.reteoo.CompositeObjectSinkAdapter.doPropagateModifyObject(CompositeObjectSinkAdapter.java:444)
> 37.
> at
>
> org.drools.reteoo.CompositeObjectSinkAdapter.propagateModifyObject(CompositeObjectSinkAdapter.java:412)
> 38.
> at
>
> org.drools.reteoo.ObjectTypeNode.modifyObject(ObjectTypeNode.java:262)
> 39.
> at
>
> org.drools.reteoo.EntryPointNode.modifyObject(EntryPointNode.java:174)
> 40.
> at
>
> org.drools.common.AbstractWorkingMemory.update(AbstractWorkingMemory.java:1392)
> 41.
> at
>
> org.drools.common.AbstractWorkingMemory.update(AbstractWorkingMemory.java:1288)
> 42.
> at
>
> org.drools.planner.examples.examination.solver.move.ExaminationMoveHelper.movePeriod(ExaminationMoveHelper.java:17)
> 43.
> at
>
> org.drools.planner.examples.examination.solver.move.PeriodChangeMove.doMove(PeriodChangeMove.java:37)
> 44.
> at
>
> org.drools.planner.core.localsearch.decider.DefaultDecider.doMove(DefaultDecider.java:117)
> 45.
> at
>
> org.drools.planner.core.localsearch.decider.DefaultDecider.decideNextStep(DefaultDecider.java:95)
> 46.
> at
>
> org.drools.planner.core.localsearch.DefaultLocalSearchSolver.solveImplementation(DefaultLocalSearchSolver.java:105)
> 47.
> at
>
> org.drools.planner.core.AbstractSolver.solve(AbstractSolver.java:41)
> 48.
> at
>
> org.drools.planner.benchmark.SolverBenchmarkSuite.benchmark(SolverBenchmarkSuite.java:196)
> 49.
> at
>
> org.drools.planner.benchmark.XmlSolverBenchmarker.benchmark(XmlSolverBenchmarker.java:72)
> 50.
> at
>
> org.drools.planner.examples.common.app.CommonBenchmarkApp.process(CommonBenchmarkApp.java:25)
> 51.
> at
>
> org.drools.planner.examples.examination.app.ExaminationBenchmarkApp.main(ExaminationBenchmarkApp.java:27)
> 52.
> at sun.reflect.NativeMethodAccessorImpl.invoke0(Native
> Method)
> 53.
> at
>
> sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:39)
> 54.
> at
>
> sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
> 55.
> at java.lang.reflect.Method.invoke(Method.java:597)
> 56.
> at
>
> com.intellij.rt.execution.application.AppMain.main(AppMain.java:110)
>
>
>
> 2) Run NQueensSmokeTest
>
> java.lang.NullPointerException
> at
> org.drools.core.util.RightTupleList.matches(RightTupleList.java:190)
> at
> org.drools.core.util.RightTupleIndexHashTable.get(RightTupleIndexHashTable.java:301)
> at
> org.drools.core.util.RightTupleIndexHashTable.getFirst(RightTupleIndexHashTable.java:87)
> at org.drools.reteoo.JoinNode.modifyLeftTuple(JoinNode.java:260)
> at org.drools.reteoo.BetaNode.modifyLeftTuple(BetaNode.java:336)
> at
> org.drools.reteoo.CompositeLeftTupleSinkAdapter.doPropagateModifyObject(CompositeLeftTupleSinkAdapter.java:204)
> at
> org.drools.reteoo.CompositeLeftTupleSinkAdapter.propagateModifyObject(CompositeLeftTupleSinkAdapter.java:219)
> at
> org.drools.reteoo.LeftInputAdapterNode.modifyObject(LeftInputAdapterNode.java:154)
> at
> org.drools.reteoo.CompositeObjectSinkAdapter.doPropagateModifyObject(CompositeObjectSinkAdapter.java:444)
> at
> org.drools.reteoo.CompositeObjectSinkAdapter.propagateModifyObject(CompositeObjectSinkAdapter.java:412)
> at
> org.drools.reteoo.ObjectTypeNode.modifyObject(ObjectTypeNode.java:262)
> at
> org.drools.reteoo.EntryPointNode.modifyObject(EntryPointNode.java:174)
> at
> org.drools.common.AbstractWorkingMemory.update(AbstractWorkingMemory.java:1392)
> at
> org.drools.common.AbstractWorkingMemory.update(AbstractWorkingMemory.java:1288)
> at
> org.drools.planner.examples.nqueens.solver.move.YChangeMove.doMove(YChangeMove.java:38)
> at
> org.drools.planner.core.localsearch.decider.DefaultDecider.doMove(DefaultDecider.java:119)
> at
> org.drools.planner.core.localsearch.decider.DefaultDecider.decideNextStep(DefaultDecider.java:95)
> at
> org.drools.planner.core.localsearch.DefaultLocalSearchSolver.solveImplementation(DefaultLocalSearchSolver.java:105)
> at
> org.drools.planner.core.AbstractSolver.solve(AbstractSolver.java:41)
> at
> org.drools.planner.examples.nqueens.NQueensSmokeTest.testSmoke(NQueensSmokeTest.java:30)
> at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
> at
> sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:39)
> at
> sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
> at
> com.intellij.junit3.JUnit3IdeaTestRunner.doRun(JUnit3IdeaTestRunner.java:108)
> at
> com.intellij.junit3.JUnit3IdeaTestRunner.startRunnerWithArgs(JUnit3IdeaTestRunner.java:42)
> at
> com.intellij.rt.execution.junit.JUnitStarter.prepareStreamsAndStart(JUnitStarter.java:165)
> at
> com.intellij.rt.execution.junit.JUnitStarter.main(JUnitStarter.java:60)
> at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
> at
> sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:39)
> at
> com.intellij.rt.execution.application.AppMain.main(AppMain.java:110)
>
> 3) Run SmartTravelingTournamentBenchmarkApp
>
> Exception in thread "main" java.lang.NullPointerException
> at
> org.drools.reteoo.SingleLeftTupleSinkAdapter.propagateModifyChildLeftTuple(SingleLeftTupleSinkAdapter.java:198)
> at org.drools.reteoo.JoinNode.modifyLeftTuple(JoinNode.java:304)
> at org.drools.reteoo.BetaNode.modifyLeftTuple(BetaNode.java:336)
> at
> org.drools.reteoo.CompositeLeftTupleSinkAdapter.doPropagateModifyObject(CompositeLeftTupleSinkAdapter.java:204)
> at
> org.drools.reteoo.CompositeLeftTupleSinkAdapter.propagateModifyObject(CompositeLeftTupleSinkAdapter.java:219)
> at
> org.drools.reteoo.LeftInputAdapterNode.modifyObject(LeftInputAdapterNode.java:154)
> at
> org.drools.reteoo.CompositeObjectSinkAdapter.doPropagateModifyObject(CompositeObjectSinkAdapter.java:444)
> at
> org.drools.reteoo.CompositeObjectSinkAdapter.propagateModifyObject(CompositeObjectSinkAdapter.java:412)
> at
> org.drools.reteoo.ObjectTypeNode.modifyObject(ObjectTypeNode.java:262)
> at
> org.drools.reteoo.EntryPointNode.modifyObject(EntryPointNode.java:174)
> at
> org.drools.common.AbstractWorkingMemory.update(AbstractWorkingMemory.java:1392)
> at
> org.drools.common.AbstractWorkingMemory.update(AbstractWorkingMemory.java:1288)
> at
> org.drools.planner.examples.travelingtournament.solver.smart.move.MultipleMatchListRotateMove.rotateList(MultipleMatchListRotateMove.java:59)
> at
> org.drools.planner.examples.travelingtournament.solver.smart.move.MultipleMatchListRotateMove.doMove(MultipleMatchListRotateMove.java:44)
> at
> org.drools.planner.core.localsearch.decider.DefaultDecider.doMove(DefaultDecider.java:117)
> at
> org.drools.planner.core.localsearch.decider.DefaultDecider.decideNextStep(DefaultDecider.java:95)
> at
> org.drools.planner.core.localsearch.DefaultLocalSearchSolver.solveImplementation(DefaultLocalSearchSolver.java:105)
> at
> org.drools.planner.core.AbstractSolver.solve(AbstractSolver.java:41)
> at
> org.drools.planner.benchmark.SolverBenchmarkSuite.benchmark(SolverBenchmarkSuite.java:196)
> at
> org.drools.planner.benchmark.XmlSolverBenchmarker.benchmark(XmlSolverBenchmarker.java:72)
> at
> org.drools.planner.examples.common.app.CommonBenchmarkApp.process(CommonBenchmarkApp.java:25)
> at
> org.drools.planner.examples.travelingtournament.app.smart.SmartTravelingTournamentBenchmarkApp.main(SmartTravelingTournamentBenchmarkApp.java:28)
> at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
> at
> sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:39)
> at
> sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
> at java.lang.reflect.Method.invoke(Method.java:597)
> at
> com.intellij.rt.execution.application.AppMain.main(AppMain.java:110)
>
> --
> With kind regards,
> Geoffrey De Smet
>
> _______________________________________________
> rules-dev mailing list
> rules-dev at lists.jboss.org <mailto:rules-dev at lists.jboss.org>
> https://lists.jboss.org/mailman/listinfo/rules-dev
>
>
>
>
> --
> Edson Tirelli
> JBoss Drools Core Development
> JBoss by Red Hat @ www.jboss.com <http://www.jboss.com>
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> rules-dev mailing list
> rules-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-dev
More information about the rules-dev
mailing list