dftefe::utils::mpi::MPIRequestersNBX Class Reference

#include <MPIRequestersNBX.h>

Inheritance diagram for dftefe::utils::mpi::MPIRequestersNBX:

Collaboration diagram for dftefe::utils::mpi::MPIRequestersNBX:

Public Member Functions
	MPIRequestersNBX (const std::vector< size_type > &targetIDs, const MPIComm &comm)
	MPIRequestersNBX ()=default
std::vector< size_type >	getRequestingRankIds () override
Public Member Functions inherited from dftefe::utils::mpi::MPIRequestersBase
virtual	~MPIRequestersBase ()=default
virtual std::vector< size_type >	getRequestingRankIds ()=0

Private Member Functions
bool	haveAllLocalSendReceived ()
void	signalLocalSendCompletion ()
bool	haveAllIncomingMsgsReceived ()
void	probeAndReceiveIncomingMsg ()
void	startLocalSend ()
void	finish ()

Private Attributes
std::vector< size_type >	d_targetIDs
std::vector< int >	d_sendBuffers
std::vector< MPIRequest >	d_sendRequests
std::vector< std::unique_ptr< int > >	d_recvBuffers
std::vector< std::unique_ptr< MPIRequest > >	d_recvRequests
MPIRequest	d_barrierRequest
const MPIComm &	d_comm
std::set< size_type >	d_requestingProcesses
int	d_numProcessors
int	d_myRank

Constructor & Destructor Documentation

MPIRequestersNBX() [1/2]

dftefe::utils::mpi::MPIRequestersNBX::MPIRequestersNBX	(	const std::vector< size_type > &	targetIDs,
		const MPIComm &	comm
	)

Here is the call graph for this function:

MPIRequestersNBX() [2/2]

dftefe::utils::mpi::MPIRequestersNBX::MPIRequestersNBX ( )

default

Member Function Documentation

finish()

void dftefe::utils::mpi::MPIRequestersNBX::finish ( )

private

After all processors have received all the incoming messages, the MPI data structures can be freed and the received messages can be processed.

Here is the call graph for this function:

Here is the caller graph for this function:

getRequestingRankIds()

std::vector< size_type > dftefe::utils::mpi::MPIRequestersNBX::getRequestingRankIds ( )

overridevirtual

Implements dftefe::utils::mpi::MPIRequestersBase.

Here is the call graph for this function:

haveAllIncomingMsgsReceived()

bool dftefe::utils::mpi::MPIRequestersNBX::haveAllIncomingMsgsReceived ( )

private

Check whether all of the incoming messages from other processors to the current processor have been received.

Here is the call graph for this function:

Here is the caller graph for this function:

haveAllLocalSendReceived()

bool dftefe::utils::mpi::MPIRequestersNBX::haveAllLocalSendReceived ( )

private

Check whether all of message sent from the current processor to other processors have been received or not

Here is the call graph for this function:

Here is the caller graph for this function:

probeAndReceiveIncomingMsg()

void dftefe::utils::mpi::MPIRequestersNBX::probeAndReceiveIncomingMsg ( )

private

Probe for an incoming message and if there is one receive it

Here is the call graph for this function:

Here is the caller graph for this function:

signalLocalSendCompletion()

void dftefe::utils::mpi::MPIRequestersNBX::signalLocalSendCompletion ( )

private

Signal to all other processors that for this processor all its message sent to other processors have been received. This is done nonblocking barrier (i.e., MPI_IBarrier).

Here is the call graph for this function:

Here is the caller graph for this function:

startLocalSend()

void dftefe::utils::mpi::MPIRequestersNBX::startLocalSend ( )

private

Start to sending message to all the target processors

Here is the call graph for this function:

Here is the caller graph for this function:

Member Data Documentation

d_barrierRequest

MPIRequest dftefe::utils::mpi::MPIRequestersNBX::d_barrierRequest

private

d_comm

const MPIComm& dftefe::utils::mpi::MPIRequestersNBX::d_comm

private

d_myRank

int dftefe::utils::mpi::MPIRequestersNBX::d_myRank

private

d_numProcessors

int dftefe::utils::mpi::MPIRequestersNBX::d_numProcessors

private

d_recvBuffers

std::vector<std::unique_ptr<int> > dftefe::utils::mpi::MPIRequestersNBX::d_recvBuffers

private

Buffers for receiving requests. We use a vector of unique pointers because that guarantees that the buffers themselves are never moved around in memory, even if the vector is resized and consequently its elements (the pointers) are moved around.

d_recvRequests

std::vector<std::unique_ptr<MPIRequest> > dftefe::utils::mpi::MPIRequestersNBX::d_recvRequests

private

Requests for receiving requests.

d_requestingProcesses

std::set<size_type> dftefe::utils::mpi::MPIRequestersNBX::d_requestingProcesses

private

List of processes who have made a request to this process.

d_sendBuffers

std::vector<int> dftefe::utils::mpi::MPIRequestersNBX::d_sendBuffers

private

Buffers for sending requests.

d_sendRequests

std::vector<MPIRequest> dftefe::utils::mpi::MPIRequestersNBX::d_sendRequests

private

Requests for sending requests.

d_targetIDs

std::vector<size_type> dftefe::utils::mpi::MPIRequestersNBX::d_targetIDs

private

List of processes this processor wants to send requests to.

---

The documentation for this class was generated from the following files:

• src/utils/MPIRequestersNBX.h
• src/utils/MPIRequestersNBX.cpp