ck-core/ckmessagelogging.h File Reference

Go to the source code of this file.

Data Structures

class RSSN

Class that stores all received-sender-sequence-numbers (rssn) from another object. More...

class ChareMlogData

This file includes the definition of the class for storing the meta data associdated with the message logging protocol. More...

class MlogEntry

Entry in a message log. More...

class StoredCheckpoint

struct CheckPointDataMsg

struct DistributeObjectMsg

struct RestartRequest

Struct to request a particular action during restart. More...

struct RestartProcessorData

struct ResendRequest

struct ReceivedTNData

struct ReceivedDetData

struct ResendData

struct MigrationRecord

struct MigrationNotice

struct MigrationNoticeAck

struct RetainedMigratedObject

struct VerifyAckMsg

struct CheckpointBarrierMsg

struct CurrentLocationMsg

struct LBStepMsg

struct DummyMigrationMsg

Typedefs

typedef CheckPointDataMsg CheckPointAck

typedef RestartRequest CkPingMsg

typedef RestartRequest CheckpointRequest

typedef ResendRequest RemoveLogRequest

typedef void(* MlogFn )(void *, ChareMlogData *)

Functions

CpvExtern (Chare *, _currentObj)

CpvExtern (int, _numImmigrantRecObjs)

void _messageLoggingInit ()

Initialize message logging data structures and register handlers.

void sendGroupMsg (envelope *env, int destPE, int _infoIdx)

Sends a group message that might be a broadcast.

void sendArrayMsg (envelope *env, int destPE, int _infoIdx)

Sends a message to an array element.

void sendChareMsg (envelope *env, int destPE, int _infoIdx, const CkChareID *pCid)

Sends a message to a singleton chare.

void sendNodeGroupMsg (envelope *env, int destNode, int _infoIdx)

Sends a nodegroup message that might be a broadcast.

void sendCommonMsg (CkObjID &recver, envelope *env, int destPE, int _infoIdx)

A method to generate the actual ticket requests for groups, nodegroups or arrays.

void sendRemoteMsg (CkObjID &sender, CkObjID &recver, int destPE, MlogEntry *entry, MCount SN, int resend)

Method that does the actual send by creating a ticket request filling it up and sending it.

void sendLocalMsg (envelope *env, int _infoIdx)

Function to send a local message.

void _pingHandler (CkPingMsg *msg)

void _skipCldEnqueue (int pe, envelope *env, int infoFn)

void _noCldNodeEnqueue (int node, envelope *env)

void generalCldEnqueue (int destPE, envelope *env, int _infoIdx)

int preProcessReceivedMessage (envelope *env, Chare **objPointer, MlogEntry **localLogEntry)

void postProcessReceivedMessage (Chare *obj, CkObjID &sender, MCount SN, MlogEntry *entry)

Updates a few variables once a message has been processed.

CpvExtern (StoredCheckpoint *, _storedCheckpointData)

void CkStartMlogCheckpoint (CkCallback &cb)

Starts checkpoint phase at PE 0.

void checkpointAlarm (void *_dummy, double curWallTime)

void startMlogCheckpoint (void *_dummy, double curWallTime)

Starts the checkpoint phase after migration.

void pupArrayElementsSkip (PUP::er &p, bool create, MigrationRecord *listToSkip, int listSize=0)

Pups all the array elements in this processor.

void _checkpointRequestHandler (CheckpointRequest *request)

void _storeCheckpointHandler (char *msg)

void _checkpointAckHandler (CheckPointAck *ackMsg)

void _removeProcessedLogHandler (char *requestMsg)

Removes messages in the log according to the received ticket numbers.

void garbageCollectMlog ()

Garbage collects the message log and other data structures.

void _startCheckpointHandler (CheckpointBarrierMsg *msg)

Starts checkpoint: send its checkpoint to its partner.

void _endCheckpointHandler (char *msg)

Finishes checkpoint process by making the callback.

void CkMlogRestart (const char *dummy, CkArgMsg *dummyMsg)

Function for restarting the crashed processor.

void CkMlogRestartDouble (void *, double)

void initializeRestart (void *data, ChareMlogData *mlogData)

Initializes variables and flags for restarting procedure.

void distributeRestartedObjects ()

Distributes objects to accelerate recovery after a failure.

void sendDummyMigration (int restartPE, CkGroupID lbID, CkGroupID locMgrID, CkArrayIndexMax &idx, int locationPE)

this method is used to send messages to a restarted processor to tell it that a particular expected object is not going to get to it

void CkMlogRestartLocal ()

void _getCheckpointHandler (RestartRequest *restartMsg)

Gets the stored checkpoint for its buddy processor.

void _recvCheckpointHandler (char *_restartData)

Receives the checkpoint data from its buddy, restores the state of all the objects and asks everyone else to update its home.

void _resendMessagesHandler (char *msg)

Resends messages since last checkpoint to the list of objects included in the request.

void _sendDetsHandler (char *msg)

Send all remote determinants to a particular failed PE.

void _sendDetsReplyHandler (char *msg)

Receives determinants stored on remote nodes.

void _receivedTNDataHandler (ReceivedTNData *msg)

Receives a list of TNs coming from the home PE of a migrated object (parallel restart).

void _receivedDetDataHandler (ReceivedDetData *msg)

Receives a list of determinants coming from the home PE of a migrated object (parallel restart).

void _distributedLocationHandler (char *receivedMsg)

Handler to update information about an object just received.

void _sendBackLocationHandler (char *receivedMsg)

Handler to receive back a location.

void _updateHomeRequestHandler (RestartRequest *updateRequest)

Updates the homePe for all chares in this processor.

void _updateHomeAckHandler (RestartRequest *updateHomeAck)

Receives the updateHome ACKs from all other processors.

void _verifyAckRequestHandler (VerifyAckMsg *verifyRequest)

void _verifyAckHandler (VerifyAckMsg *verifyReply)

void _dummyMigrationHandler (DummyMigrationMsg *msg)

this handler is used to process a dummy migration msg.

void _restartHandler (RestartRequest *restartMsg)

Function to restart this processor.

void _getRestartCheckpointHandler (RestartRequest *restartMsg)

Gets the stored checkpoint but calls another function in the sender.

void _recvRestartCheckpointHandler (char *_restartData)

Receives the checkpoint coming from its buddy.

void startLoadBalancingMlog (void(*fnPtr)(void *), void *_centralLb)

Load Balancing.

void finishedCheckpointLoadBalancing ()

void sendMlogLocation (int targetPE, envelope *env)

void resumeFromSyncRestart (void *data, ChareMlogData *mlogData)

void restoreParallelRecovery (void(*fnPtr)(void *), void *_centralLb)

Restores objects after parallel recovery, either by sending back the immigrant objects or by waiting for all emigrant objects to be back.

void _receiveMlogLocationHandler (void *buf)

void _receiveMigrationNoticeHandler (MigrationNotice *msg)

void _receiveMigrationNoticeAckHandler (MigrationNoticeAck *msg)

void _getGlobalStepHandler (LBStepMsg *msg)

void _recvGlobalStepHandler (LBStepMsg *msg)

Receives the global step handler from PE 0.

void _checkpointBarrierHandler (CheckpointBarrierMsg *msg)

Processor 0 receives a contribution from every other processor after checkpoint.

void _checkpointBarrierAckHandler (CheckpointBarrierMsg *msg)

int getCheckPointPE ()

Getting the pe number of the current processor's buddy.

void forAllCharesDo (MlogFn fnPointer, void *data)

Map function pointed by fnPointer over all the chares living in this processor.

envelope * copyEnvelope (envelope *env)

void _initDone (void)

This function (not a handler) is called once and only once per processor.

void _resetNodeBocInitVec (void)

void informLocationHome (CkGroupID mgrID, CkArrayIndexMax idx, int homePE, int currentPE)

method that informs an array elements home processor of its current location It is a converse method to bypass the charm++ message logging framework

void _receiveLocationHandler (CurrentLocationMsg *data)

void CmiDeliverRemoteMsgHandlerRange (int lowerHandler, int higherHandler)

Variables

char objString [100]

int _checkpointRequestHandlerIdx

int _storeCheckpointHandlerIdx

int _checkpointAckHandlerIdx

int _removeProcessedLogHandlerIdx

int _getCheckpointHandlerIdx

int _recvCheckpointHandlerIdx

int _resendMessagesHandlerIdx

int _sendDetsHandlerIdx

int _sendDetsReplyHandlerIdx

int _receivedTNDataHandlerIdx

int _receivedDetDataHandlerIdx

int _distributedLocationHandlerIdx

int _updateHomeRequestHandlerIdx

int _updateHomeAckHandlerIdx

int _verifyAckRequestHandlerIdx

int _verifyAckHandlerIdx

int _dummyMigrationHandlerIdx

int onGoingLoadBalancing

For testing on clusters we might carry out restarts on a porcessor without actually starting it 1 -> false restart 0 -> restart after an actual crash.

void * centralLb

void(* resumeLbFnPtr )(void *)

int _receiveMlogLocationHandlerIdx

int _receiveMigrationNoticeHandlerIdx

int _receiveMigrationNoticeAckHandlerIdx

int _getGlobalStepHandlerIdx

int _recvGlobalStepHandlerIdx

int _checkpointBarrierHandlerIdx

int _checkpointBarrierAckHandlerIdx

std::vector< MigrationRecord > migratedNoticeList

std::vector
< RetainedMigratedObject * > retainedObjectList

int _receiveLocationHandlerIdx

Typedef Documentation

typedef CheckPointDataMsg CheckPointAck

Definition at line 259 of file ckmessagelogging.h.

typedef RestartRequest CkPingMsg

Definition at line 270 of file ckmessagelogging.h.

typedef RestartRequest CheckpointRequest

Definition at line 271 of file ckmessagelogging.h.

typedef ResendRequest RemoveLogRequest

Definition at line 291 of file ckmessagelogging.h.

typedef void(* MlogFn)(void *, ChareMlogData *)

Definition at line 386 of file ckmessagelogging.h.

Function Documentation

CpvExtern	(	Chare *	,
		_currentObj
	)

CpvExtern	(	int	,
		_numImmigrantRecObjs
	)

void _messageLoggingInit ( )

Initialize message logging data structures and register handlers.

Definition at line 277 of file ckcausalmlog.C.

References _checkpointAckHandler(), _checkpointAckHandlerIdx, _checkpointBarrierAckHandler(), _checkpointBarrierAckHandlerIdx, _checkpointBarrierHandler(), _checkpointBarrierHandlerIdx, _checkpointRequestHandler(), _checkpointRequestHandlerIdx, _distributedLocationHandler(), _distributedLocationHandlerIdx, _dummyMigrationHandler(), _dummyMigrationHandlerIdx, _endCheckpointHandler(), _endCheckpointIdx, _getCheckpointHandler(), _getCheckpointHandlerIdx, _getGlobalStepHandler(), _getGlobalStepHandlerIdx, _getRestartCheckpointHandler(), _getRestartCheckpointHandlerIdx, _indexBufferedDets, _maxBufferedDets, _numBufferedDets, _phaseBufferedDets, _pingHandler(), _pingHandlerIdx, _receivedDetDataHandler(), _receivedDetDataHandlerIdx, _receivedTNDataHandler(), _receivedTNDataHandlerIdx, _receiveLocationHandler(), _receiveLocationHandlerIdx, _receiveMigrationNoticeAckHandler(), _receiveMigrationNoticeAckHandlerIdx, _receiveMigrationNoticeHandler(), _receiveMigrationNoticeHandlerIdx, _receiveMlogLocationHandler(), _receiveMlogLocationHandlerIdx, _recvCheckpointHandler(), _recvCheckpointHandlerIdx, _recvGlobalStepHandler(), _recvGlobalStepHandlerIdx, _recvRestartCheckpointHandler(), _recvRestartCheckpointHandlerIdx, _removeDeterminantsHandler(), _removeDeterminantsHandlerIdx, _removeProcessedLogHandler(), _removeProcessedLogHandlerIdx, _resendMessagesHandler(), _resendMessagesHandlerIdx, _restartHandler(), _restartHandlerIdx, _sendBackLocationHandler(), _sendBackLocationHandlerIdx, _sendDetsHandler(), _sendDetsHandlerIdx, _sendDetsReplyHandler(), _sendDetsReplyHandlerIdx, _startCheckpointHandler(), _startCheckpointIdx, _storeCheckpointHandler(), _storeCheckpointHandlerIdx, _storeDeterminantsHandler(), _storeDeterminantsHandlerIdx, _updateHomeAckHandler(), _updateHomeAckHandlerIdx, _updateHomeRequestHandler(), _updateHomeRequestHandlerIdx, _verifyAckHandler(), _verifyAckHandlerIdx, _verifyAckRequestHandler(), _verifyAckRequestHandlerIdx, bufferedDetsSize, CcdCallOnCondition(), Converse::CkMyPe(), Converse::CkNumPes(), CmiAlloc(), CmiWallTimer(), CqsCreate(), diskCkptFlag, fName, heartBeatCheckHandler(), heartBeatCheckHandlerIdx, heartBeatHandler(), heartBeatHandlerIdx, heartBeatPartner(), int, lastCompletedAlarm, lastRestart, msgLogSize, numDets, numDupDets, numMsgsTarget, numPiggyDets, partnerFailureHandler(), partnerFailureHandlerIdx, sizeMsgsTarget, storedDetsSize, totalMsgsSize, totalMsgsTarget, and traceRegisterUserEvent().

Referenced by _initCharm().

Here is the call graph for this function:

Here is the caller graph for this function:

void sendGroupMsg	(	envelope *	env,
		int	destPE,
		int	_infoIdx
	)

Sends a group message that might be a broadcast.

Definition at line 628 of file ckcausalmlog.C.

References CkCopyMsg(), Converse::CkMyPe(), CmiMyPe(), CkObjID::data, EnvToUsr(), envelope::getGroupNum(), _ObjectID::group, _ObjectID::id, _ObjectID::onPE, sendCommonMsg(), sendGroupMsg(), CkObjID::type, TypeGroup, TypeInvalid, and UsrToEnv().

Referenced by _sendMsgBranch(), and sendGroupMsg().

Here is the call graph for this function:

Here is the caller graph for this function:

void sendArrayMsg	(	envelope *	env,
		int	destPE,
		int	_infoIdx
	)

Sends a message to an array element.

Definition at line 693 of file ckcausalmlog.C.

References _ObjectID::array, CkArrayIndexBase::asChild(), Converse::CkMyPe(), CkObjID::data, envelope::getArrayMgr(), _ObjectID::s_array::id, _ObjectID::s_array::idx, sendCommonMsg(), CkObjID::toString(), CkObjID::type, and TypeArray.

Referenced by CkArrayManagerDeliver().

Here is the call graph for this function:

Here is the caller graph for this function:

void sendChareMsg	(	envelope *	env,
		int	destPE,
		int	_infoIdx,
		const CkChareID *	pCid
	)

Sends a message to a singleton chare.

Definition at line 715 of file ckcausalmlog.C.

References _ObjectID::chare, Converse::CkMyPe(), CkObjID::data, _ObjectID::id, sendCommonMsg(), CkObjID::toString(), CkObjID::type, TypeArray, and TypeChare.

Referenced by CkSendMsg().

Here is the call graph for this function:

Here is the caller graph for this function:

void sendNodeGroupMsg	(	envelope *	env,
		int	destNode,
		int	_infoIdx
	)

Sends a nodegroup message that might be a broadcast.

Definition at line 661 of file ckcausalmlog.C.

References CkCopyMsg(), Converse::CkMyPe(), CkObjID::data, EnvToUsr(), envelope::getGroupNum(), _ObjectID::group, _ObjectID::id, _ObjectID::onPE, sendCommonMsg(), sendNodeGroupMsg(), CkObjID::type, TypeInvalid, TypeNodeGroup, and UsrToEnv().

Referenced by _sendMsgNodeBranch(), and sendNodeGroupMsg().

Here is the call graph for this function:

Here is the caller graph for this function:

void sendCommonMsg	(	CkObjID &	recver,
		envelope *	env,
		int	destPE,
		int	_infoIdx
	)

A method to generate the actual ticket requests for groups, nodegroups or arrays.

Definition at line 736 of file ckcausalmlog.C.

References Converse::CkMyPe(), CmiMemoryCheck(), generalCldEnqueue(), SNToTicket::get(), CkHashtableT< KEY, OBJ >::get(), CkObjID::getObject(), isLocal(), isTeamLocal(), Chare::mlogData, ChareMlogData::nextSN(), sendLocalMsg(), sendMsg(), sendRemoteMsg(), teamSize, ChareMlogData::teamTable, Ticket::TN, CkObjID::toString(), and TypeInvalid.

Referenced by sendArrayMsg(), sendChareMsg(), sendGroupMsg(), and sendNodeGroupMsg().

Here is the call graph for this function:

Here is the caller graph for this function:

void sendRemoteMsg	(	CkObjID &	sender,
		CkObjID &	recver,
		int	destPE,
		MlogEntry *	entry,
		MCount	SN,
		int	resend
	)

Method that does the actual send by creating a ticket request filling it up and sending it.

Definition at line 624 of file ckmessagelogging.C.

References MlogEntry::_infoIdx, ChareMlogData::addLogEntry(), Converse::CkMyPe(), CmiMemoryCheck(), MlogEntry::env, float, generalCldEnqueue(), envelope::getTotalsize(), Chare::mlogData, MLOGFT_totalLogSize, MLOGFT_totalMessages, msgLogSize, numMsgsTarget, sizeMsgsTarget, totalMsgsSize, and totalMsgsTarget.

Referenced by sendCommonMsg().

Here is the call graph for this function:

Here is the caller graph for this function:

void sendLocalMsg	(	envelope *	env,
		int	_infoIdx
	)

Function to send a local message.

It first gets a ticket and then enqueues the message. If we are recovering, then the message is enqueued in a delay queue.

Definition at line 937 of file ckcausalmlog.C.

References _skipCldEnqueue(), CmiMemoryCheck(), and CmiMyPe().

Referenced by sendCommonMsg().

Here is the call graph for this function:

Here is the caller graph for this function:

void _pingHandler ( CkPingMsg * msg )

Definition at line 1343 of file ckcausalmlog.C.

References Converse::CkMyPe(), CmiFree(), and RestartRequest::PE.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _skipCldEnqueue	(	int	pe,
		envelope *	env,
		int	infoFn
	)

Definition at line 1425 of file ck.C.

References _enqObjQueue(), CkFindObjectPtr(), Chare::CkGetObjQueue(), Converse::CkMyPe(), CkPackMessage(), CkRdmaPrepareBcastMsg(), CmiFree(), CmiNodeOf(), CmiPersistentOneSend(), Converse::CmiSyncBroadcast(), Converse::CmiSyncBroadcastAll(), Converse::CmiSyncBroadcastAllAndFree(), Converse::CmiSyncBroadcastAndFree(), Converse::CmiSyncSend(), Converse::CmiSyncSendAndFree(), ConverseDeliver(), CqsEnqueueGeneral(), envelope::getPriobits(), envelope::getPrioPtr(), envelope::getQueueing(), envelope::getTotalsize(), index_objectQHandler, index_skipCldHandler, len, and CkObjectMsgQ::queue().

Referenced by _sendMsgBranch(), CkArrayManagerDeliver(), generalCldEnqueue(), and sendLocalMsg().

Here is the call graph for this function:

Here is the caller graph for this function:

void _noCldNodeEnqueue	(	int	node,
		envelope *	env
	)

Definition at line 1556 of file ck.C.

References CkPackMessage(), CkRdmaPrepareBcastMsg(), CmiFree(), ConverseDeliver(), envelope::getTotalsize(), and len.

Referenced by _sendMsgNodeBranch(), CkSendMsgNodeBranchImmediate(), and generalCldEnqueue().

Here is the call graph for this function:

Here is the caller graph for this function:

void generalCldEnqueue	(	int	destPE,
		envelope *	env,
		int	_infoIdx
	)

Definition at line 1322 of file ckcausalmlog.C.

References _noCldNodeEnqueue(), _skipCldEnqueue(), and TypeNodeGroup.

Referenced by sendCommonMsg(), sendMsg(), and sendRemoteMsg().

Here is the call graph for this function:

Here is the caller graph for this function:

int preProcessReceivedMessage	(	envelope *	env,
		Chare **	objPointer,
		MlogEntry **	localLogEntry
	)

Definition at line 1181 of file ckcausalmlog.C.

References _getTicket(), _recoveryFlag, addBufferedDeterminant(), ChareMlogData::checkAndStoreSsn(), Converse::CkMyPe(), CmiFree(), CmiMemoryCheck(), Converse::CmiSyncSendAndFree(), CqsDequeue(), CqsEmpty(), CqsEnqueue(), CqsEnqueueGeneral(), fault_aware(), flag, CkObjID::getObject(), envelope::getPriobits(), envelope::getPrioPtr(), envelope::getSrcPe(), envelope::getTotalsize(), CkObjID::guessPE(), Chare::mlogData, CkObjID::toString(), and ChareMlogData::tProcessed.

Referenced by _processHandler().

Here is the call graph for this function:

Here is the caller graph for this function:

void postProcessReceivedMessage	(	Chare *	obj,
		CkObjID &	sender,
		MCount	SN,
		MlogEntry *	entry
	)

Updates a few variables once a message has been processed.

Definition at line 1302 of file ckcausalmlog.C.

References Converse::CkMyPe(), CmiMemoryCheck(), MlogEntry::env, CkObjID::guessPE(), Chare::mlogData, and ChareMlogData::tProcessed.

Referenced by _processHandler().

Here is the call graph for this function:

Here is the caller graph for this function:

CpvExtern	(	StoredCheckpoint *	,
		_storedCheckpointData
	)

void CkStartMlogCheckpoint ( CkCallback & cb )

Starts checkpoint phase at PE 0.

Definition at line 832 of file ckmessagelogging.C.

References _startCheckpointIdx, CmiAlloc(), and Converse::CmiSyncBroadcastAllAndFree().

Here is the call graph for this function:

void checkpointAlarm	(	void *	_dummy,
		double	curWallTime
	)

Definition at line 1358 of file ckcausalmlog.C.

References _checkpointRequestHandlerIdx, CcdCallFnAfter(), checkpointAlarm(), chkptPeriod, Converse::CkMyPe(), CmiInitMsgHeader(), Converse::CmiSyncBroadcastAll(), RestartRequest::header, lastCompletedAlarm, RestartRequest::PE, and request.

Referenced by checkpointAlarm(), and startMlogCheckpoint().

Here is the call graph for this function:

Here is the caller graph for this function:

void startMlogCheckpoint	(	void *	_dummy,
		double	curWallTime
	)

Starts the checkpoint phase after migration.

Definition at line 1383 of file ckcausalmlog.C.

References _recoveryFlag, _storeCheckpointHandlerIdx, buf, bufferedDetsSize, buildProcessedTicketLog(), CcdCallFnAfter(), checkpointAlarm(), checkpointCount, chkptPeriod, Converse::CkMyPe(), CkPupGroupData(), CkPupNodeGroupData(), CkPupROData(), CmiAlloc(), CmiMemoryCheck(), CmiMyPe(), Converse::CmiSyncSendAndFree(), CmiTimer(), CmiWallTimer(), CheckPointDataMsg::dataSize, dataSize, forAllCharesDo(), getCheckPointPE(), inCkptFlag, lastCompletedAlarm, msg, msgLogSize, onGoingLoadBalancing, CheckPointDataMsg::PE, processedTicketLog, pupArrayElementsSkip(), PUP::sizer::size(), storedDetsSize, traceUserBracketEvent(), and unAckedCheckpoint.

Referenced by _checkpointRequestHandler(), _receiveMigrationNoticeAckHandler(), _updateHomeRequestHandler(), and startLoadBalancingMlog().

Here is the call graph for this function:

Here is the caller graph for this function:

void pupArrayElementsSkip	(	PUP::er &	p,
		bool	create,
		MigrationRecord *	listToSkip,
		int	listSize = `0`
	)

Pups all the array elements in this processor.

Definition at line 1498 of file ckcausalmlog.C.

References CkCountArrayElements(), Converse::CkMyPe(), CmiMyPe(), flag, CkLocMgr::homePe(), _ckGroupID::idx, idx, MigrationRecord::idx, idx2str(), informLocationHome(), PUP::er::isUnpacking(), CkLocMgr::numLocalElements(), and CkLocMgr::resume().

Referenced by _recvCheckpointHandler(), _recvRestartCheckpointHandler(), _startCheckpointHandler(), and startMlogCheckpoint().

Here is the call graph for this function:

Here is the caller graph for this function:

void _checkpointRequestHandler ( CheckpointRequest * request )

Definition at line 1376 of file ckcausalmlog.C.

References CmiWallTimer(), and startMlogCheckpoint().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _storeCheckpointHandler ( char * msg )

Definition at line 1574 of file ckcausalmlog.C.

References _checkpointAckHandlerIdx, Converse::CkMyPe(), CmiFree(), CmiInitMsgHeader(), CmiMyPe(), Converse::CmiSyncSend(), count, CheckPointDataMsg::dataSize, diskCkptFlag, CheckPointDataMsg::header, migratedNoticeList, CheckPointDataMsg::PE, traceUserBracketEvent(), and writeCheckpointToDisk().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _checkpointAckHandler ( CheckPointAck * ackMsg )

Definition at line 1671 of file ckcausalmlog.C.

References _endCheckpointIdx, Converse::CkMyPe(), CmiAlloc(), CmiFree(), CmiMemoryCheck(), CmiReduce(), CheckPointDataMsg::dataSize, doNothingMsg(), finishedCheckpointLoadBalancing(), inCkptFlag, onGoingLoadBalancing, CheckPointDataMsg::PE, sendRemoveLogRequests(), and unAckedCheckpoint.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _removeProcessedLogHandler ( char * requestMsg )

Removes messages in the log according to the received ticket numbers.

Definition at line 1783 of file ckcausalmlog.C.

References Converse::CkMyPe(), clearUpMigratedRetainedLists(), CmiFree(), CmiMemoryCheck(), forAllCharesDo(), ResendRequest::PE, populateDeterminantTable(), removeProcessedLogs(), request, and traceUserBracketEvent().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void garbageCollectMlog ( )

Garbage collects the message log and other data structures.

In case of synchronized checkpoint, we use an optimization to avoid causal message logging protocol to communicate all determinants to the rest of the processors.

Definition at line 3580 of file ckcausalmlog.C.

References _indexBufferedDets, _numBufferedDets, _phaseBufferedDets, Converse::CkMyPe(), forAllCharesDo(), garbageCollectMlogForChare(), CkHashtableIterator::hasNext(), and CkHashtableIterator::next().

Referenced by _startCheckpointHandler(), and initMlogLBStep().

Here is the call graph for this function:

Here is the caller graph for this function:

void _startCheckpointHandler ( CheckpointBarrierMsg * startMsg )

Starts checkpoint: send its checkpoint to its partner.

This checkpointing strategy is NOT connected to the load balancer, hence onGoingLoadBalancer==0.

Definition at line 847 of file ckmessagelogging.C.

References _storeCheckpointHandlerIdx, buf, bufferedDetsSize, checkpointCount, Converse::CkMyPe(), CkPupGroupData(), CkPupNodeGroupData(), CkPupROData(), CmiAlloc(), CmiFree(), CmiMemoryCheck(), CmiMyPe(), Converse::CmiSyncSendAndFree(), CmiTimer(), CmiWallTimer(), CheckPointDataMsg::dataSize, dataSize, garbageCollectMlog(), getCheckPointPE(), _ckGroupID::idx, inCkptFlag, msgLogSize, CheckPointDataMsg::PE, pupArrayElementsSkip(), PUP::sizer::size(), storedDetsSize, and unAckedCheckpoint.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _endCheckpointHandler ( char * msg )

Finishes checkpoint process by making the callback.

Definition at line 926 of file ckmessagelogging.C.

References CmiFree(), and CkCallback::send().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void CkMlogRestart	(	const char *	dummy,
		CkArgMsg *	dummyMsg
	)

Function for restarting the crashed processor.

It sets the restart flag and contacts the buddy processor to get the latest checkpoint.

Definition at line 1847 of file ckcausalmlog.C.

References _getCheckpointHandlerIdx, _numRestartResponses, _recoveryFlag, _restartFlag, _restartHandlerIdx, Converse::CkMyPe(), Converse::CkNumPes(), CmiInitMsgHeader(), Converse::CmiSyncSend(), CmiWallTimer(), getCheckPointPE(), RestartRequest::header, msg, RestartRequest::PE, and teamSize.

Referenced by _initCharm(), _parseCommandLineOpts(), CkMlogRestartDouble(), and CkMlogRestartLocal().

Here is the call graph for this function:

Here is the caller graph for this function:

void CkMlogRestartDouble	(	void *	,
		double
	)

Definition at line 2063 of file ckcausalmlog.C.

References CkMlogRestart().

Here is the call graph for this function:

void initializeRestart	(	void *	data,
		ChareMlogData *	mlogData
	)

Initializes variables and flags for restarting procedure.

Definition at line 2355 of file ckcausalmlog.C.

References ChareMlogData::receivedTNs, ChareMlogData::resendReplyRecvd, and ChareMlogData::restartFlag.

Referenced by _recvCheckpointHandler(), and _recvRestartCheckpointHandler().

Here is the caller graph for this function:

void distributeRestartedObjects ( )

Distributes objects to accelerate recovery after a failure.

Definition at line 3075 of file ckcausalmlog.C.

References Converse::CkMyPe().

Referenced by _recvGlobalStepHandler(), _sendDetsReplyHandler(), and _updateHomeAckHandler().

Here is the call graph for this function:

Here is the caller graph for this function:

void sendDummyMigration	(	int	restartPE,
		CkGroupID	lbID,
		CkGroupID	locMgrID,
		CkArrayIndexMax &	idx,
		int	locationPE
	)

this method is used to send messages to a restarted processor to tell it that a particular expected object is not going to get to it

Definition at line 3168 of file ckcausalmlog.C.

References _dummyMigrationHandlerIdx, buf, CmiInitMsgHeader(), Converse::CmiSyncSend(), DummyMigrationMsg::flag, DummyMigrationMsg::header, DummyMigrationMsg::idx, DummyMigrationMsg::lbID, DummyMigrationMsg::locationPE, and DummyMigrationMsg::mgrID.

Here is the call graph for this function:

void CkMlogRestartLocal ( )

Definition at line 2068 of file ckcausalmlog.C.

References CkMlogRestart().

Here is the call graph for this function:

void _getCheckpointHandler ( RestartRequest * restartMsg )

Gets the stored checkpoint for its buddy processor.

Definition at line 2075 of file ckcausalmlog.C.

References _recvCheckpointHandlerIdx, _verifyAckRequestHandlerIdx, StoredCheckpoint::buf, buf, StoredCheckpoint::bufSize, RestartProcessorData::checkPointSize, Converse::CkMyPe(), CmiAlloc(), CmiFree(), CmiInitMsgHeader(), CmiMyPe(), CmiPrintf(), Converse::CmiSyncSend(), Converse::CmiSyncSendAndFree(), CmiTimer(), diskCkptFlag, VerifyAckMsg::fromPE, VerifyAckMsg::header, idx, idx2str(), VerifyAckMsg::index, RestartProcessorData::lbGroupID, migratedNoticeList, VerifyAckMsg::migRecord, msg, RestartProcessorData::PE, StoredCheckpoint::PE, RestartRequest::PE, readCheckpointFromDisk(), RestartProcessorData::restartWallTime, sendCheckpointData(), verifyAckCount, and verifyAckTotal.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _recvCheckpointHandler ( char * _restartData )

Receives the checkpoint data from its buddy, restores the state of all the objects and asks everyone else to update its home.

Definition at line 2234 of file ckcausalmlog.C.

References _initDone(), _numRestartResponses, _sendDetsHandlerIdx, adjustChkptPeriod, buf, checkpointCount, RestartProcessorData::checkPointSize, chkptPeriod, Converse::CkMyPe(), CkPupGroupData(), CkPupNodeGroupData(), CkPupROData(), CmiAlloc(), CmiFree(), CmiMyPe(), Converse::CmiSyncBroadcastAndFree(), CmiWallTimer(), createObjIDList(), forAllCharesDo(), getGlobalStep(), initializeRestart(), RestartProcessorData::lbGroupID, ResendRequest::numberObjects, RestartProcessorData::numMigratedAwayElements, ResendRequest::PE, RestartProcessorData::PE, pupArrayElementsSkip(), RestartProcessorData::restartWallTime, and PUP::mem::size().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _resendMessagesHandler ( char * msg )

Resends messages since last checkpoint to the list of objects included in the request.

It also sends stored remote determinants to the particular failed PE.

Definition at line 2663 of file ckcausalmlog.C.

References Converse::CkMyPe(), CmiFree(), CmiMemoryCheck(), CmiResetGlobalReduceSeqID(), CmiWallTimer(), PUP::d, fillTicketForChare(), forAllCharesDo(), isTeamLocal(), lastRestart, ResendData::listObjects, ResendRequest::numberObjects, ResendData::numberObjects, ResendRequest::PE, ResendData::PE, and resendMessageForChare().

Referenced by _messageLoggingInit(), and _recvRestartCheckpointHandler().

Here is the call graph for this function:

Here is the caller graph for this function:

void _sendDetsHandler ( char * msg )

Send all remote determinants to a particular failed PE.

It only sends determinants to those objects on the list.

Definition at line 2557 of file ckcausalmlog.C.

References _sendDetsReplyHandlerIdx, Converse::CkMyPe(), CmiAlloc(), CmiFree(), CmiMemoryCheck(), CmiResetGlobalReduceSeqID(), Converse::CmiSyncSendAndFree(), CmiWallTimer(), PUP::d, int, lastRestart, ResendData::listObjects, ResendRequest::numberObjects, ResendData::numberObjects, ResendRequest::PE, ResendData::PE, printDet(), TProcessedLog::recver, ResendData::ticketVecs, and TProcessedLog::tProcessed.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _sendDetsReplyHandler ( char * msg )

Receives determinants stored on remote nodes.

Definition at line 2719 of file ckcausalmlog.C.

References _numRestartResponses, _receivedDetDataHandlerIdx, _receivedTNDataHandlerIdx, _resendMessagesHandlerIdx, Converse::CkMyPe(), Converse::CkNumPes(), CmiAlloc(), CmiMyPe(), Converse::CmiSyncBroadcastAllAndFree(), Converse::CmiSyncSendAndFree(), CmiWallTimer(), createObjIDList(), distributeRestartedObjects(), fastRecovery, forAllCharesDo(), int, lb, ResendRequest::numberObjects, ReceivedDetData::numDets, ReceivedTNData::numTNs, ResendRequest::PE, processReceivedDet(), processReceivedTN(), CentralLB::ReceiveDummyMigration(), ReceivedDetData::recver, ReceivedTNData::recver, and restartDecisionNumber.

Referenced by _messageLoggingInit(), and _recvGlobalStepHandler().

Here is the call graph for this function:

Here is the caller graph for this function:

void _receivedTNDataHandler ( ReceivedTNData * msg )

Receives a list of TNs coming from the home PE of a migrated object (parallel restart).

Definition at line 2846 of file ckcausalmlog.C.

References CmiFree(), CmiMyPe(), Converse::CmiSyncSendAndFree(), CkObjID::getObject(), CkObjID::guessPE(), Chare::mlogData, ReceivedTNData::numTNs, ChareMlogData::objID, processReceivedTN(), ReceivedTNData::recver, and CkObjID::toString().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _receivedDetDataHandler ( ReceivedDetData * msg )

Receives a list of determinants coming from the home PE of a migrated object (parallel restart).

Definition at line 2828 of file ckcausalmlog.C.

References CmiFree(), CmiMyPe(), Converse::CmiSyncSendAndFree(), CkObjID::getObject(), CkObjID::guessPE(), Chare::mlogData, ReceivedDetData::numDets, ChareMlogData::objID, processReceivedDet(), ReceivedDetData::recver, and CkObjID::toString().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _distributedLocationHandler ( char * receivedMsg )

Handler to update information about an object just received.

Definition at line 3122 of file ckcausalmlog.C.

References _ObjectID::array, buf, Converse::CkMyPe(), CkObjID::data, CkReductionMgr::decGCount(), donotCountMigration, CkLocMgr::elementRec(), globalResumeCount, CkLocMgr::homePe(), _ObjectID::s_array::id, idx, CkReductionMgr::incNumImmigrantRecObjs(), informLocationHome(), CkLocMgr::migratableList(), ChareMlogData::objID, DistributeObjectMsg::PE, CkLocMgr::resume(), and ChareMlogData::toResumeOrNot.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _sendBackLocationHandler ( char * receivedMsg )

Handler to receive back a location.

Definition at line 3085 of file ckcausalmlog.C.

References _ObjectID::array, buf, centralLb, Converse::CkMyPe(), CkObjID::data, CkReductionMgr::decGCount(), CkReductionMgr::decNumEmigrantRecObjs(), donotCountMigration, CkLocMgr::elementRec(), CkLocMgr::homePe(), _ObjectID::s_array::id, idx, informLocationHome(), CkLocMgr::migratableList(), ChareMlogData::objID, DistributeObjectMsg::PE, and CkLocMgr::resume().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _updateHomeRequestHandler ( RestartRequest * updateRequest )

Updates the homePe for all chares in this processor.

Definition at line 2388 of file ckcausalmlog.C.

References _receiveMigrationNoticeHandlerIdx, _updateHomeAckHandlerIdx, checkpointCount, CmiInitMsgHeader(), CmiMyPe(), CmiPrintf(), Converse::CmiSyncSend(), Converse::CmiSyncSendAndFree(), forAllCharesDo(), getCheckPointPE(), MigrationNotice::header, MigrationNotice::migRecord, RestartRequest::PE, MigrationNotice::record, retainedObjectList, startMlogCheckpoint(), unAckedCheckpoint, and updateHomePE().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _updateHomeAckHandler ( RestartRequest * updateHomeAck )

Receives the updateHome ACKs from all other processors.

Once everybody has replied, it sends a request to resend the logged messages.

Definition at line 2307 of file ckcausalmlog.C.

References _resendMessagesHandlerIdx, Converse::CkMyPe(), CmiAlloc(), CmiFree(), Converse::CmiSyncBroadcastAllAndFree(), CmiWallTimer(), countUpdateHomeAcks, createObjIDList(), distributeRestartedObjects(), fastRecovery, forAllCharesDo(), lb, ResendRequest::numberObjects, ResendRequest::PE, CentralLB::ReceiveDummyMigration(), and restartDecisionNumber.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _verifyAckRequestHandler ( VerifyAckMsg * verifyRequest )

Definition at line 2113 of file ckcausalmlog.C.

References _verifyAckHandlerIdx, CmiMyPe(), CmiPrintf(), Converse::CmiSyncSendAndFree(), CkLocMgr::elementNrec(), VerifyAckMsg::fromPE, CkLocRec::getLBDB(), CkLocRec::getLdHandle(), MigrationRecord::gID, _ckGroupID::idx, MigrationRecord::idx, idx2str(), VerifyAckMsg::migRecord, CkLocMgr::reclaim(), CkLocMgr::setDuringMigration(), LBDatabase::UnregisterObj(), and verifyAckedRequests.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _verifyAckHandler ( VerifyAckMsg * verifyReply )

Definition at line 2139 of file ckcausalmlog.C.

References CmiMyPe(), CmiPrintf(), idx, idx2str(), VerifyAckMsg::index, index, migratedNoticeList, VerifyAckMsg::migRecord, sendCheckpointData(), verifyAckCount, and verifyAckTotal.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _dummyMigrationHandler ( DummyMigrationMsg * msg )

this handler is used to process a dummy migration msg.

it looks up the load balancer and calls migrated for it

Definition at line 3203 of file ckcausalmlog.C.

References CmiFree(), CmiMyPe(), CmiPrintf(), DummyMigrationMsg::count, DummyMigrationMsg::flag, h, DummyMigrationMsg::idx, _ckGroupID::idx, idx2str(), lb, DummyMigrationMsg::lbID, DummyMigrationMsg::locationPE, DummyMigrationMsg::mgrID, CentralLB::Migrated(), and verifyAckedRequests.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _restartHandler ( RestartRequest * restartMsg )

Function to restart this processor.

The handler is invoked by a member of its same team in message logging.

Definition at line 1879 of file ckcausalmlog.C.

References _getRestartCheckpointHandlerIdx, _numRestartResponses, _restartFlag, Converse::CkMyPe(), CmiInitMsgHeader(), Converse::CmiSyncSend(), CmiWallTimer(), getCheckPointPE(), RestartRequest::header, msg, and RestartRequest::PE.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _getRestartCheckpointHandler ( RestartRequest * restartMsg )

Gets the stored checkpoint but calls another function in the sender.

Definition at line 1911 of file ckcausalmlog.C.

References _verifyAckRequestHandlerIdx, CmiInitMsgHeader(), CmiMyPe(), CmiPrintf(), Converse::CmiSyncSend(), VerifyAckMsg::fromPE, VerifyAckMsg::header, idx, idx2str(), VerifyAckMsg::index, migratedNoticeList, VerifyAckMsg::migRecord, msg, StoredCheckpoint::PE, RestartRequest::PE, sendCheckpointData(), verifyAckCount, and verifyAckTotal.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _recvRestartCheckpointHandler ( char * _restartData )

Receives the checkpoint coming from its buddy.

This is the case of restart for one team member that did not crash.

Definition at line 1951 of file ckcausalmlog.C.

References _resendMessagesHandler(), _resendMessagesHandlerIdx, adjustChkptPeriod, buf, checkpointCount, RestartProcessorData::checkPointSize, chkptPeriod, Converse::CkMyPe(), Converse::CkNumPes(), CkPupGroupData(), CkPupNodeGroupData(), CkPupROData(), CmiAlloc(), CmiFree(), CmiMyPe(), Converse::CmiSyncSend(), CmiWallTimer(), createObjIDList(), forAllCharesDo(), initializeRestart(), lb, RestartProcessorData::lbGroupID, ResendRequest::numberObjects, RestartProcessorData::numMigratedAwayElements, ResendRequest::PE, RestartProcessorData::PE, pupArrayElementsSkip(), CentralLB::ReceiveDummyMigration(), restartDecisionNumber, RestartProcessorData::restartWallTime, setTeamRecovery(), PUP::mem::size(), sleep(), teamSize, and unsetTeamRecovery().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void startLoadBalancingMlog	(	void()(void )	fnPtr,
		void *	_centralLb
	)

Load Balancing.

Definition at line 3394 of file ckcausalmlog.C.

References centralLb, CmiMyPe(), CmiWallTimer(), countLBMigratedAway, countLBToMigrate, migrationDoneCalled, resumeLbFnPtr, and startMlogCheckpoint().

Referenced by CentralLB::MigrationDoneImpl().

Here is the call graph for this function:

Here is the caller graph for this function:

void finishedCheckpointLoadBalancing ( )

Definition at line 3407 of file ckcausalmlog.C.

References _checkpointBarrierHandlerIdx, CmiAlloc(), CmiMyPe(), CmiReduce(), and doNothingMsg().

Referenced by _checkpointAckHandler().

Here is the call graph for this function:

Here is the caller graph for this function:

void sendMlogLocation	(	int	targetPE,
		envelope *	env
	)

void resumeFromSyncRestart	(	void *	data,
		ChareMlogData *	mlogData
	)

Definition at line 3513 of file ckcausalmlog.C.

void restoreParallelRecovery	(	void()(void )	fnPtr,
		void *	_centralLb
	)

Restores objects after parallel recovery, either by sending back the immigrant objects or by waiting for all emigrant objects to be back.

Definition at line 3377 of file ckcausalmlog.C.

References centralLb, resumeLbFnPtr, and sendBackImmigrantRecObjs().

Referenced by CentralLB::ReceiveMigration().

Here is the call graph for this function:

Here is the caller graph for this function:

void _receiveMlogLocationHandler ( void * buf )

Definition at line 3499 of file ckcausalmlog.C.

References Converse::CkMyPe(), CkUnpackMessage(), EnvToUsr(), envelope::getTotalsize(), CkArrayElementMigrateMessage::gid, _ckGroupID::idx, and CkLocMgr::immigrate().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _receiveMigrationNoticeHandler ( MigrationNotice * msg )

Definition at line 3468 of file ckcausalmlog.C.

References _receiveMigrationNoticeAckHandlerIdx, MigrationRecord::ackFrom, MigrationRecord::ackTo, buf, CmiInitMsgHeader(), Converse::CmiSyncSend(), getCheckPointPE(), MigrationNoticeAck::header, migratedNoticeList, MigrationNotice::migRecord, MigrationNotice::record, and MigrationNoticeAck::record.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _receiveMigrationNoticeAckHandler ( MigrationNoticeAck * msg )

Definition at line 3479 of file ckcausalmlog.C.

References _receiveMlogLocationHandlerIdx, RetainedMigratedObject::acked, CmiMyPe(), Converse::CmiSyncSend(), CmiWallTimer(), countLBMigratedAway, countLBToMigrate, MigrationRecord::gID, CkLocMgr::homePe(), MigrationRecord::idx, informLocationHome(), migrationDoneCalled, RetainedMigratedObject::migRecord, RetainedMigratedObject::msg, MigrationNoticeAck::record, RetainedMigratedObject::size, startMlogCheckpoint(), and MigrationRecord::toPE.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _getGlobalStepHandler ( LBStepMsg * msg )

Definition at line 3668 of file ckcausalmlog.C.

References _recvGlobalStepHandlerIdx, CmiMyPe(), CmiPrintf(), Converse::CmiSyncSend(), LBStepMsg::fromPE, _ckGroupID::idx, lb, LBStepMsg::lbID, and LBStepMsg::step.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _recvGlobalStepHandler ( LBStepMsg * msg )

Receives the global step handler from PE 0.

Definition at line 3680 of file ckcausalmlog.C.

References _resendMessagesHandlerIdx, _sendDetsReplyHandler(), Converse::CkMyPe(), CmiAlloc(), CmiFree(), CmiMyPe(), CmiPrintf(), Converse::CmiSyncBroadcastAllAndFree(), CmiWallTimer(), createObjIDList(), distributeRestartedObjects(), fastRecovery, forAllCharesDo(), lb, ResendRequest::numberObjects, ResendRequest::PE, CentralLB::ReceiveDummyMigration(), restartDecisionNumber, and LBStepMsg::step.

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _checkpointBarrierHandler ( CheckpointBarrierMsg * msg )

Processor 0 receives a contribution from every other processor after checkpoint.

Definition at line 3527 of file ckcausalmlog.C.

References _checkpointBarrierAckHandlerIdx, CmiAlloc(), CmiFree(), and Converse::CmiSyncBroadcastAllAndFree().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void _checkpointBarrierAckHandler ( CheckpointBarrierMsg * msg )

Definition at line 3538 of file ckcausalmlog.C.

References centralLb, Converse::CkMyPe(), CmiFree(), CmiMyPe(), CmiPrintf(), inCkptFlag, and sendRemoveLogRequests().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

int getCheckPointPE ( )

Getting the pe number of the current processor's buddy.

In the team-based approach each processor might checkpoint in the next team, but currently teams are only meant to reduce memory overhead. Note: function getReverseCheckPointPE performs the reverse map. It must be changed accordingly.

Definition at line 4097 of file ckcausalmlog.C.

References CmiMyPe().

Referenced by _receiveMigrationNoticeHandler(), _restartHandler(), _startCheckpointHandler(), _updateHomeRequestHandler(), CkMlogRestart(), heartBeatPartner(), sendMlogLocation(), and startMlogCheckpoint().

Here is the call graph for this function:

Here is the caller graph for this function:

void forAllCharesDo	(	MlogFn	fnPointer,
		void *	data
	)

Map function pointed by fnPointer over all the chares living in this processor.

Definition at line 3254 of file ckcausalmlog.C.

References caller, and Chare::mlogData.

Referenced by _recvCheckpointHandler(), _recvGlobalStepHandler(), _recvRestartCheckpointHandler(), _removeProcessedLogHandler(), _resendMessagesHandler(), _sendDetsReplyHandler(), _updateHomeAckHandler(), _updateHomeRequestHandler(), garbageCollectMlog(), and startMlogCheckpoint().

Here is the caller graph for this function:

envelope* copyEnvelope ( envelope * env )

Definition at line 4109 of file ckcausalmlog.C.

References CmiAlloc(), and envelope::getTotalsize().

Referenced by resendMessageForChare().

Here is the call graph for this function:

Here is the caller graph for this function:

void _resetNodeBocInitVec ( void )

void informLocationHome	(	CkGroupID	mgrID,
		CkArrayIndexMax	idx,
		int	homePE,
		int	currentPE
	)

method that informs an array elements home processor of its current location It is a converse method to bypass the charm++ message logging framework

Definition at line 3610 of file ckcausalmlog.C.

References _receiveLocationHandlerIdx, Converse::CkMyPe(), CmiInitMsgHeader(), CmiMyPe(), CmiPrintf(), Converse::CmiSyncSend(), CmiWallTimer(), CurrentLocationMsg::fromPE, CurrentLocationMsg::header, _ckGroupID::idx, CurrentLocationMsg::idx, idx2str(), CurrentLocationMsg::locationPE, CurrentLocationMsg::mgrID, and traceUserBracketEvent().

Referenced by _distributedLocationHandler(), _receiveMigrationNoticeAckHandler(), _sendBackLocationHandler(), pupArrayElementsSkip(), and updateHomePE().

Here is the call graph for this function:

Here is the caller graph for this function:

void _receiveLocationHandler ( CurrentLocationMsg * data )

Definition at line 3626 of file ckcausalmlog.C.

References Converse::CkMyPe(), CmiAbort(), CmiFree(), CmiMyPe(), CmiPrintf(), Converse::CmiSyncSend(), CmiWallTimer(), CkLocMgr::elementNrec(), CurrentLocationMsg::fromPE, CurrentLocationMsg::idx, idx2str(), CkLocMgr::inform(), CkLocMgr::lastKnown(), CurrentLocationMsg::locationPE, CurrentLocationMsg::mgrID, and traceUserBracketEvent().

Referenced by _messageLoggingInit().

Here is the call graph for this function:

Here is the caller graph for this function:

void CmiDeliverRemoteMsgHandlerRange	(	int	lowerHandler,
		int	higherHandler
	)

Variable Documentation

char objString[100]

Definition at line 177 of file ckcausalmlog.C.

Referenced by buildProcessedTicketLog(), createObjIDList(), and processReceivedTN().

int _checkpointRequestHandlerIdx

Definition at line 178 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and checkpointAlarm().

int _storeCheckpointHandlerIdx

Definition at line 179 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), _startCheckpointHandler(), and startMlogCheckpoint().

int _checkpointAckHandlerIdx

Definition at line 180 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and _storeCheckpointHandler().

int _removeProcessedLogHandlerIdx

Definition at line 183 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and sendRemoveLogRequests().

int _getCheckpointHandlerIdx

Definition at line 181 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and CkMlogRestart().

int _recvCheckpointHandlerIdx

Definition at line 182 of file ckcausalmlog.C.

Referenced by _getCheckpointHandler(), _messageLoggingInit(), and sendCheckpointData().

int _resendMessagesHandlerIdx

Definition at line 195 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), _recvGlobalStepHandler(), _recvRestartCheckpointHandler(), _sendDetsReplyHandler(), and _updateHomeAckHandler().

int _sendDetsHandlerIdx

Definition at line 196 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and _recvCheckpointHandler().

int _sendDetsReplyHandlerIdx

Definition at line 197 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and _sendDetsHandler().

int _receivedTNDataHandlerIdx

Definition at line 198 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and _sendDetsReplyHandler().

int _receivedDetDataHandlerIdx

Definition at line 199 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and _sendDetsReplyHandler().

int _distributedLocationHandlerIdx

Definition at line 200 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and ElementDistributor::addLocation().

int _updateHomeRequestHandlerIdx

Definition at line 193 of file ckcausalmlog.C.

Referenced by _messageLoggingInit().

int _updateHomeAckHandlerIdx

Definition at line 194 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and _updateHomeRequestHandler().

int _verifyAckRequestHandlerIdx

Definition at line 185 of file ckcausalmlog.C.

Referenced by _getCheckpointHandler(), _getRestartCheckpointHandler(), and _messageLoggingInit().

int _verifyAckHandlerIdx

Definition at line 186 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and _verifyAckRequestHandler().

int _dummyMigrationHandlerIdx

Definition at line 187 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), sendDummyMigration(), and sendDummyMigrationCounts().

int onGoingLoadBalancing

For testing on clusters we might carry out restarts on a porcessor without actually starting it 1 -> false restart 0 -> restart after an actual crash.

Definition at line 227 of file ckcausalmlog.C.

Referenced by _checkpointAckHandler(), initMlogLBStep(), and startMlogCheckpoint().

void* centralLb

Definition at line 228 of file ckcausalmlog.C.

Referenced by _checkpointBarrierAckHandler(), _sendBackLocationHandler(), restoreParallelRecovery(), and startLoadBalancingMlog().

void(* resumeLbFnPtr)(void *)

Referenced by restoreParallelRecovery(), and startLoadBalancingMlog().

int _receiveMlogLocationHandlerIdx

Definition at line 230 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and _receiveMigrationNoticeAckHandler().

int _receiveMigrationNoticeHandlerIdx

Definition at line 231 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), _updateHomeRequestHandler(), and sendMlogLocation().

int _receiveMigrationNoticeAckHandlerIdx

Definition at line 232 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and _receiveMigrationNoticeHandler().

int _getGlobalStepHandlerIdx

Definition at line 190 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and getGlobalStep().

int _recvGlobalStepHandlerIdx

Definition at line 191 of file ckcausalmlog.C.

Referenced by _getGlobalStepHandler(), and _messageLoggingInit().

int _checkpointBarrierHandlerIdx

Definition at line 233 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and finishedCheckpointLoadBalancing().

int _checkpointBarrierAckHandlerIdx

Definition at line 234 of file ckcausalmlog.C.

Referenced by _checkpointBarrierHandler(), and _messageLoggingInit().

std::vector<MigrationRecord> migratedNoticeList

Definition at line 236 of file ckcausalmlog.C.

Referenced by _getCheckpointHandler(), _getRestartCheckpointHandler(), _receiveMigrationNoticeHandler(), _storeCheckpointHandler(), _verifyAckHandler(), clearUpMigratedRetainedLists(), and sendCheckpointData().

std::vector<RetainedMigratedObject *> retainedObjectList

Definition at line 237 of file ckcausalmlog.C.

Referenced by _updateHomeRequestHandler(), clearUpMigratedRetainedLists(), and sendMlogLocation().

int _receiveLocationHandlerIdx

Definition at line 251 of file ckcausalmlog.C.

Referenced by _messageLoggingInit(), and informLocationHome().


Data Structures
class	RSSN
	Class that stores all received-sender-sequence-numbers (rssn) from another object. More...
class	ChareMlogData
	This file includes the definition of the class for storing the meta data associdated with the message logging protocol. More...
class	MlogEntry
	Entry in a message log. More...
class	StoredCheckpoint
struct	CheckPointDataMsg
struct	DistributeObjectMsg
struct	RestartRequest
	Struct to request a particular action during restart. More...
struct	RestartProcessorData
struct	ResendRequest
struct	ReceivedTNData
struct	ReceivedDetData
struct	ResendData
struct	MigrationRecord
struct	MigrationNotice
struct	MigrationNoticeAck
struct	RetainedMigratedObject
struct	VerifyAckMsg
struct	CheckpointBarrierMsg
struct	CurrentLocationMsg
struct	LBStepMsg
struct	DummyMigrationMsg
Typedefs
typedef CheckPointDataMsg	CheckPointAck
typedef RestartRequest	CkPingMsg
typedef RestartRequest	CheckpointRequest
typedef ResendRequest	RemoveLogRequest
typedef void(*	MlogFn )(void , ChareMlogData )
Functions
	CpvExtern (Chare *, _currentObj)
	CpvExtern (int, _numImmigrantRecObjs)
void	_messageLoggingInit ()
	Initialize message logging data structures and register handlers.
void	sendGroupMsg (envelope *env, int destPE, int _infoIdx)
	Sends a group message that might be a broadcast.
void	sendArrayMsg (envelope *env, int destPE, int _infoIdx)
	Sends a message to an array element.
void	sendChareMsg (envelope env, int destPE, int _infoIdx, const CkChareID pCid)
	Sends a message to a singleton chare.
void	sendNodeGroupMsg (envelope *env, int destNode, int _infoIdx)
	Sends a nodegroup message that might be a broadcast.
void	sendCommonMsg (CkObjID &recver, envelope *env, int destPE, int _infoIdx)
	A method to generate the actual ticket requests for groups, nodegroups or arrays.
void	sendRemoteMsg (CkObjID &sender, CkObjID &recver, int destPE, MlogEntry *entry, MCount SN, int resend)
	Method that does the actual send by creating a ticket request filling it up and sending it.
void	sendLocalMsg (envelope *env, int _infoIdx)
	Function to send a local message.
void	_pingHandler (CkPingMsg *msg)
void	_skipCldEnqueue (int pe, envelope *env, int infoFn)
void	_noCldNodeEnqueue (int node, envelope *env)
void	generalCldEnqueue (int destPE, envelope *env, int _infoIdx)
int	preProcessReceivedMessage (envelope env, Chare objPointer, MlogEntry *localLogEntry)
void	postProcessReceivedMessage (Chare obj, CkObjID &sender, MCount SN, MlogEntry entry)
	Updates a few variables once a message has been processed.
	CpvExtern (StoredCheckpoint *, _storedCheckpointData)
void	CkStartMlogCheckpoint (CkCallback &cb)
	Starts checkpoint phase at PE 0.
void	checkpointAlarm (void *_dummy, double curWallTime)
void	startMlogCheckpoint (void *_dummy, double curWallTime)
	Starts the checkpoint phase after migration.
void	pupArrayElementsSkip (PUP::er &p, bool create, MigrationRecord *listToSkip, int listSize=0)
	Pups all the array elements in this processor.
void	_checkpointRequestHandler (CheckpointRequest *request)
void	_storeCheckpointHandler (char *msg)
void	_checkpointAckHandler (CheckPointAck *ackMsg)
void	_removeProcessedLogHandler (char *requestMsg)
	Removes messages in the log according to the received ticket numbers.
void	garbageCollectMlog ()
	Garbage collects the message log and other data structures.
void	_startCheckpointHandler (CheckpointBarrierMsg *msg)
	Starts checkpoint: send its checkpoint to its partner.
void	_endCheckpointHandler (char *msg)
	Finishes checkpoint process by making the callback.
void	CkMlogRestart (const char dummy, CkArgMsg dummyMsg)
	Function for restarting the crashed processor.
void	CkMlogRestartDouble (void *, double)
void	initializeRestart (void data, ChareMlogData mlogData)
	Initializes variables and flags for restarting procedure.
void	distributeRestartedObjects ()
	Distributes objects to accelerate recovery after a failure.
void	sendDummyMigration (int restartPE, CkGroupID lbID, CkGroupID locMgrID, CkArrayIndexMax &idx, int locationPE)
	this method is used to send messages to a restarted processor to tell it that a particular expected object is not going to get to it
void	CkMlogRestartLocal ()
void	_getCheckpointHandler (RestartRequest *restartMsg)
	Gets the stored checkpoint for its buddy processor.
void	_recvCheckpointHandler (char *_restartData)
	Receives the checkpoint data from its buddy, restores the state of all the objects and asks everyone else to update its home.
void	_resendMessagesHandler (char *msg)
	Resends messages since last checkpoint to the list of objects included in the request.
void	_sendDetsHandler (char *msg)
	Send all remote determinants to a particular failed PE.
void	_sendDetsReplyHandler (char *msg)
	Receives determinants stored on remote nodes.
void	_receivedTNDataHandler (ReceivedTNData *msg)
	Receives a list of TNs coming from the home PE of a migrated object (parallel restart).
void	_receivedDetDataHandler (ReceivedDetData *msg)
	Receives a list of determinants coming from the home PE of a migrated object (parallel restart).
void	_distributedLocationHandler (char *receivedMsg)
	Handler to update information about an object just received.
void	_sendBackLocationHandler (char *receivedMsg)
	Handler to receive back a location.
void	_updateHomeRequestHandler (RestartRequest *updateRequest)
	Updates the homePe for all chares in this processor.
void	_updateHomeAckHandler (RestartRequest *updateHomeAck)
	Receives the updateHome ACKs from all other processors.
void	_verifyAckRequestHandler (VerifyAckMsg *verifyRequest)
void	_verifyAckHandler (VerifyAckMsg *verifyReply)
void	_dummyMigrationHandler (DummyMigrationMsg *msg)
	this handler is used to process a dummy migration msg.
void	_restartHandler (RestartRequest *restartMsg)
	Function to restart this processor.
void	_getRestartCheckpointHandler (RestartRequest *restartMsg)
	Gets the stored checkpoint but calls another function in the sender.
void	_recvRestartCheckpointHandler (char *_restartData)
	Receives the checkpoint coming from its buddy.
void	startLoadBalancingMlog (void(fnPtr)(void ), void *_centralLb)
	Load Balancing.
void	finishedCheckpointLoadBalancing ()
void	sendMlogLocation (int targetPE, envelope *env)
void	resumeFromSyncRestart (void data, ChareMlogData mlogData)
void	restoreParallelRecovery (void(fnPtr)(void ), void *_centralLb)
	Restores objects after parallel recovery, either by sending back the immigrant objects or by waiting for all emigrant objects to be back.
void	_receiveMlogLocationHandler (void *buf)
void	_receiveMigrationNoticeHandler (MigrationNotice *msg)
void	_receiveMigrationNoticeAckHandler (MigrationNoticeAck *msg)
void	_getGlobalStepHandler (LBStepMsg *msg)
void	_recvGlobalStepHandler (LBStepMsg *msg)
	Receives the global step handler from PE 0.
void	_checkpointBarrierHandler (CheckpointBarrierMsg *msg)
	Processor 0 receives a contribution from every other processor after checkpoint.
void	_checkpointBarrierAckHandler (CheckpointBarrierMsg *msg)
int	getCheckPointPE ()
	Getting the pe number of the current processor's buddy.
void	forAllCharesDo (MlogFn fnPointer, void *data)
	Map function pointed by fnPointer over all the chares living in this processor.
envelope *	copyEnvelope (envelope *env)
void	_initDone (void)
	This function (not a handler) is called once and only once per processor.
void	_resetNodeBocInitVec (void)
void	informLocationHome (CkGroupID mgrID, CkArrayIndexMax idx, int homePE, int currentPE)
	method that informs an array elements home processor of its current location It is a converse method to bypass the charm++ message logging framework
void	_receiveLocationHandler (CurrentLocationMsg *data)
void	CmiDeliverRemoteMsgHandlerRange (int lowerHandler, int higherHandler)
Variables
char	objString [100]
int	_checkpointRequestHandlerIdx
int	_storeCheckpointHandlerIdx
int	_checkpointAckHandlerIdx
int	_removeProcessedLogHandlerIdx
int	_getCheckpointHandlerIdx
int	_recvCheckpointHandlerIdx
int	_resendMessagesHandlerIdx
int	_sendDetsHandlerIdx
int	_sendDetsReplyHandlerIdx
int	_receivedTNDataHandlerIdx
int	_receivedDetDataHandlerIdx
int	_distributedLocationHandlerIdx
int	_updateHomeRequestHandlerIdx
int	_updateHomeAckHandlerIdx
int	_verifyAckRequestHandlerIdx
int	_verifyAckHandlerIdx
int	_dummyMigrationHandlerIdx
int	onGoingLoadBalancing
	For testing on clusters we might carry out restarts on a porcessor without actually starting it 1 -> false restart 0 -> restart after an actual crash.
void *	centralLb
void(*	resumeLbFnPtr )(void *)
int	_receiveMlogLocationHandlerIdx
int	_receiveMigrationNoticeHandlerIdx
int	_receiveMigrationNoticeAckHandlerIdx
int	_getGlobalStepHandlerIdx
int	_recvGlobalStepHandlerIdx
int	_checkpointBarrierHandlerIdx
int	_checkpointBarrierAckHandlerIdx
std::vector< MigrationRecord >	migratedNoticeList
std::vector < RetainedMigratedObject * >	retainedObjectList
int	_receiveLocationHandlerIdx