#include <communicator.h>

Public Member Functions
	Communicator ()

	~Communicator ()

Static Public Member Functions
static void	init (int numberOfArguments, char **cmdLineArguments)
	Parallel::Communicator::init initializes the communicater and sets up the lattice geometry. More...

static void	initializeSubLattice ()
	Parallel::Communicator::initializeSubLattice. More...

static SU3	getPositiveLink (Lattice< SU3 > lattice, std::vector< int > n, int mu, int muIndex, int SU3Dir)
	Parallel::Communicator::getPositiveLink fetches a link in the positive direction. More...

static SU3	getNegativeLink (Lattice< SU3 > lattice, std::vector< int > n, int mu, int muIndex, int SU3Dir)
	Parallel::Communicator::getNegativeLink fetches a link in the negative direction. More...

static SU3	getNeighboursNeighbourLink (Lattice< SU3 > lattice, std::vector< int > n, int mu, int muIndex, int nu, int *nuIndex, int SU3Dir)
	Parallel::Communicator::getNeighboursNeighbourLink fetches a neighbours neighbor link. More...

static SU3	getNeighboursNeighbourNegativeLink (Lattice< SU3 > lattice, std::vector< int > n, int mu, int muIndex, int nu, int *nuIndex, int SU3Dir)
	Parallel::Communicator::getNeighboursNeighbourNegativeLink. More...

static int	getProcessRank ()

static int	getNumProc ()

static void	setN (std::vector< unsigned int > N)
	Parallel::Communicator::setN sets the lattice dimensions in the Parallel::Communicator class. More...

static void	MPIExit (std::string message)
	Parallel::Communicator::MPIExit exits the program. Frees MPI groups before it exits. More...

static void	MPIPrint (std::string message)
	Parallel::Communicator::MPIPrint prints a message from rank 0. Includes barriers. More...

static void	setBarrier ()
	Parallel::Communicator::setBarrier. More...

static void	setBarrierActive ()
	Parallel::Communicator::setBarrierActive. More...

static void	gatherDoubleResults (double *data, unsigned int N)
	Parallel::Communicator::gatherDoubleResults. More...

static void	freeMPIGroups ()
	Parallel::Communicator::freeMPIGroups. More...

static void	reduceToTemporalDimension (std::vector< double > &obsResults, std::vector< double > obs)
	Parallel::Communicator::reduceToTemporalDimension reduces the results to the temporal dimensions, i.e. Euclidean time. More...

static void	checkProcessorValidity ()
	Parallel::Communicator::checkProcessorValidity checks that we do not have an odd number of processors. More...

static void	checkSubLatticeDimensionsValidity ()
	Parallel::Communicator::checkSubLatticeDimensionsValidity. More...

static void	checkSubLatticeValidity ()
	Parallel::Communicator::checkSubLatticeValidity runs a series of tests to ensure that the sub lattices has been correctly set up. More...

static void	checkLattice (Lattice< SU3 > *lattice, std::string message)
	Parallel::Communicator::checkLattice checks if the lattice containts invalid numbers/nan. More...

Constructor & Destructor Documentation

◆ Communicator()

Communicator::Communicator ( )

◆ ~Communicator()

Communicator::~Communicator ( )

Member Function Documentation

◆ checkLattice()

void Communicator::checkLattice	(	Lattice< SU3 > *	lattice,
		std::string	message
	)

static

Parallel::Communicator::checkLattice checks if the lattice containts invalid numbers/nan.

Parameters

lattice	lattice to check.
message	to print if it contains nan numbers.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ checkProcessorValidity()

void Communicator::checkProcessorValidity ( )

static

Parallel::Communicator::checkProcessorValidity checks that we do not have an odd number of processors.

Todo:: : this could probably be changed such that it just leaves out one processor that is left over.

◆ checkSubLatticeDimensionsValidity()

void Communicator::checkSubLatticeDimensionsValidity ( )

static

Parallel::Communicator::checkSubLatticeDimensionsValidity.

Ensures that the lattice size is valid. If it is of size 2 or less, we exit.

◆ checkSubLatticeValidity()

void Communicator::checkSubLatticeValidity ( )

static

Parallel::Communicator::checkSubLatticeValidity runs a series of tests to ensure that the sub lattices has been correctly set up.

Here is the call graph for this function:

◆ freeMPIGroups()

void Communicator::freeMPIGroups ( )

static

Parallel::Communicator::freeMPIGroups.

Frees MPI groups and communicators.

Here is the caller graph for this function:

◆ gatherDoubleResults()

void Communicator::gatherDoubleResults	(	double *	data,
		unsigned int	N
	)

static

Parallel::Communicator::gatherDoubleResults.

Parameters

data	to reduce.
N	number of points in data to reduce.

Todo:: Remove the unsigned long int, and use instead just unsigned long? Change globally to only use long?

Here is the caller graph for this function:

◆ getNegativeLink()

SU3 Communicator::getNegativeLink	(	Lattice< SU3 > *	lattice,
		std::vector< int >	n,
		int	mu,
		int *	muIndex,
		int	SU3Dir
	)

static

Parallel::Communicator::getNegativeLink fetches a link in the negative direction.

Parameters

lattice	a lattice pointer for all four dimensions.
n	position in lattice to fetch link from.
mu	direction to shift in. Always negative in x, y, z and t directions. Index is the same as the step direction of muIndex.
muIndex	is a unit vector containing a step in the direction of sharing. It is \(\hat{\mu}\) in \(U_\nu (n + \hat{mu}) \).
SU3Dir	is the index of the tensor \(\mu\) in link \(U_{\mu}\).

Returns: fetched SU3 matrix.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ getNeighboursNeighbourLink()

SU3 Communicator::getNeighboursNeighbourLink	(	Lattice< SU3 > *	lattice,
		std::vector< int >	n,
		int	mu,
		int *	muIndex,
		int	nu,
		int *	nuIndex,
		int	SU3Dir
	)

static

Parallel::Communicator::getNeighboursNeighbourLink fetches a neighbours neighbor link.

Fetches the link when it is given as \(U_\mu(n + \hat{\mu} - \hat{\nu})\).

Parameters

lattice	a lattice pointer for all four dimensions.
n	position in lattice to fetch link from.
mu	index of the muIndex vector we are sharing.
muIndex	is a unit vector containing a step in the direction of sharing.
nu	index of the nuIndex vector we are sharing.
nuIndex	is a unit vector containing a step in the direction of sharing.
SU3Dir	is the index of the tensor \(\mu\) in link \(U_{\mu}\).

Returns: the fetched SU3 matrix.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ getNeighboursNeighbourNegativeLink()

SU3 Communicator::getNeighboursNeighbourNegativeLink	(	Lattice< SU3 > *	lattice,
		std::vector< int >	n,
		int	mu,
		int *	muIndex,
		int	nu,
		int *	nuIndex,
		int	SU3Dir
	)

static

Parallel::Communicator::getNeighboursNeighbourNegativeLink.

Fetches the link when it is given as \(U_\mu(n - \hat{\mu} - \hat{\nu})\).

Parameters

lattice	a lattice pointer for all four dimensions.
n	position in lattice to fetch link from.
mu	index of the muIndex vector we are sharing.
muIndex	is a unit vector containing a step in the direction of sharing.
nu	index of the nuIndex vector we are sharing.
nuIndex	is a unit vector containing a step in the direction of sharing.
SU3Dir	is the index of the tensor \(\mu\) in link \(U_{\mu}\).

Returns: the fetched SU3 matrix.

Todo:: Should probably pass n by reference, and set all elements as const.

Here is the call graph for this function:

◆ getNumProc()

static int Parallel::Communicator::getNumProc ( )

inlinestatic

Here is the caller graph for this function:

◆ getPositiveLink()

SU3 Communicator::getPositiveLink	(	Lattice< SU3 > *	lattice,
		std::vector< int >	n,
		int	mu,
		int *	muIndex,
		int	SU3Dir
	)

static

Parallel::Communicator::getPositiveLink fetches a link in the positive direction.

Parameters

lattice	a lattice pointer for all four dimensions.
n	position in lattice to fetch link from.
mu	direction to shift in. Always negative in x, y, z and t directions. Index is the same as the step direction of muIndex.
muIndex	is a unit vector containing a step in the direction of sharing. It is \(\hat{\mu}\) in \(U_\nu (n + \hat{mu}) \).
SU3Dir	is the index of the tensor \(\mu\) in link \(U_{\mu}\).

Returns: fetched SU3 matrix.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ getProcessRank()

static int Parallel::Communicator::getProcessRank ( )

inlinestatic

Here is the caller graph for this function:

◆ init()

void Communicator::init	(	int *	numberOfArguments,
		char ***	cmdLineArguments
	)

static

Parallel::Communicator::init initializes the communicater and sets up the lattice geometry.

Parameters

numberOfArguments	number of command line arguments.
cmdLineArguments	command line arguments.

Here is the caller graph for this function:

◆ initializeSubLattice()

void Communicator::initializeSubLattice ( )

static

Parallel::Communicator::initializeSubLattice.

Sets up sublattices. Either by retrieving it from the Parameters class, or by setting it up manually.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ MPIExit()

void Communicator::MPIExit ( std::string message )

static

Parallel::Communicator::MPIExit exits the program. Frees MPI groups before it exits.

Parameters

message message to print before exiting.

Here is the caller graph for this function:

◆ MPIPrint()

void Communicator::MPIPrint ( std::string message )

static

Parallel::Communicator::MPIPrint prints a message from rank 0. Includes barriers.

Parameters

message to print.

◆ reduceToTemporalDimension()

void Communicator::reduceToTemporalDimension	(	std::vector< double > &	obsResults,
		std::vector< double >	obs
	)

static

Parallel::Communicator::reduceToTemporalDimension reduces the results to the temporal dimensions, i.e. Euclidean time.

Parameters

obsResults	contigious vector that results will be placed in.
obs	vector we are gathering results in.

Todo:

Should probably pass by the obs by reference.

Should probably set obs as const.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ setBarrier()

void Communicator::setBarrier ( )

static

Parallel::Communicator::setBarrier.

A MPI_Barrier for all processors.

Here is the caller graph for this function:

◆ setBarrierActive()

void Communicator::setBarrierActive ( )

static

Parallel::Communicator::setBarrierActive.

A MPI_Barrier for only the active processors, i.e. those used in flow or cfg. generation.

Here is the caller graph for this function:

◆ setN()

void Communicator::setN ( std::vector< unsigned int > N )

static

Parallel::Communicator::setN sets the lattice dimensions in the Parallel::Communicator class.

Parameters

N	vector of global lattice dimensions

Here is the caller graph for this function:

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

GLAC/parallelization/communicator.h
GLAC/parallelization/communicator.cpp

Public Member Functions

Static Public Member Functions

Constructor & Destructor Documentation

◆ Communicator()

◆ ~Communicator()

Member Function Documentation

◆ checkLattice()

◆ checkProcessorValidity()

◆ checkSubLatticeDimensionsValidity()

◆ checkSubLatticeValidity()

◆ freeMPIGroups()

◆ gatherDoubleResults()

◆ getNegativeLink()

◆ getNeighboursNeighbourLink()

◆ getNeighboursNeighbourNegativeLink()

◆ getNumProc()

◆ getPositiveLink()

◆ getProcessRank()

◆ init()

◆ initializeSubLattice()

◆ MPIExit()

◆ MPIPrint()

◆ reduceToTemporalDimension()

◆ setBarrier()

◆ setBarrierActive()

◆ setN()