dftefe::linearAlgebra::elpaScalaOpInternal Namespace Reference

Functions
void	setupELPAHandleParameters (const utils::mpi::MPIComm &mpi_communicator, utils::mpi::MPIComm &processGridCommunicatorActive, const std::shared_ptr< const ProcessGrid > &processGrid, const size_type na, const size_type nev, const size_type blockSize, elpa_t &elpaHandle, const bool useELPADeviceKernel=false)
	Contains internal functions used in linearAlgebra. More...
void	createProcessGridSquareMatrix (const utils::mpi::MPIComm &mpi_communicator, const size_type size, std::shared_ptr< const ProcessGrid > &processGrid, const size_type scalapackParalProcs, const bool useELPA, const bool useOnlyThumbRule=false)
	Wrapper function to create a two dimensional processor grid for a square matrix in ScaLAPACKMatrix storage format. More...
void	createProcessGridRectangularMatrix (const utils::mpi::MPIComm &mpi_communicator, const size_type sizeRows, const size_type sizeColumns, std::shared_ptr< const ProcessGrid > &processGrid)
	Wrapper function to create a two dimensional processor grid for a rectangular matrix in ScaLAPACKMatrix storage format. More...
template<typename T >
void	createGlobalToLocalIdMapsScaLAPACKMat (const std::shared_ptr< const ProcessGrid > &processGrid, const ScaLAPACKMatrix< T > &mat, std::unordered_map< size_type, size_type > &globalToLocalRowIdMap, std::unordered_map< size_type, size_type > &globalToLocalColumnIdMap)
	Creates global row/column id to local row/column ids for ScaLAPACKMatrix. More...
template<typename T >
void	scaleScaLAPACKMat (const std::shared_ptr< const ProcessGrid > &processGrid, ScaLAPACKMatrix< T > &mat, const T scalar)
	scale a ScaLAPACKMat with a scalar More...
template<typename T >
void	fillParallelOverlapMatrix (const T *X, const size_type XLocalSize, const size_type numberVectors, const std::shared_ptr< const ProcessGrid > &processGrid, const utils::mpi::MPIComm &mpiComm, const size_type wfcBlockSize, ScaLAPACKMatrix< T > &overlapMatPar)
	Computes Sc=X^{T}*Xc and stores in a parallel ScaLAPACK matrix. X^{T} is the subspaceVectorsArray stored in the column major format (N x M). Sc is the overlapMatPar. More...
template<typename ValueType , typename utils::MemorySpace memorySpace>
void	subspaceRotation (ValueType *X, const size_type M, const size_type N, const std::shared_ptr< const ProcessGrid > &processGrid, const utils::mpi::MPIComm &mpiCommDomain, LinAlgOpContext< memorySpace > &linAlgOpContext, const ScaLAPACKMatrix< ValueType > &rotationMatPar, const size_type subspaceRotDofsBlockSize, const size_type wfcBlockSize, const bool rotationMatTranspose=false, const bool isRotationMatLowerTria=false)
	Computes X^{T}=Q*X^{T} inplace. X^{T} is the subspaceVectorsArray stored in the column major format (N x M). Q is rotationMatPar (N x N). More...

Function Documentation

createGlobalToLocalIdMapsScaLAPACKMat()

template<typename T >

void dftefe::linearAlgebra::elpaScalaOpInternal::createGlobalToLocalIdMapsScaLAPACKMat	(	const std::shared_ptr< const ProcessGrid > &	processGrid,
		const ScaLAPACKMatrix< T > &	mat,
		std::unordered_map< size_type, size_type > &	globalToLocalRowIdMap,
		std::unordered_map< size_type, size_type > &	globalToLocalColumnIdMap
	)

Creates global row/column id to local row/column ids for ScaLAPACKMatrix.

Here is the call graph for this function:

Here is the caller graph for this function:

createProcessGridRectangularMatrix()

void dftefe::linearAlgebra::elpaScalaOpInternal::createProcessGridRectangularMatrix	(	const utils::mpi::MPIComm &	mpi_communicator,
		const size_type	sizeRows,
		const size_type	sizeColumns,
		std::shared_ptr< const ProcessGrid > &	processGrid
	)

Wrapper function to create a two dimensional processor grid for a rectangular matrix in ScaLAPACKMatrix storage format.

Here is the call graph for this function:

createProcessGridSquareMatrix()

void dftefe::linearAlgebra::elpaScalaOpInternal::createProcessGridSquareMatrix	(	const utils::mpi::MPIComm &	mpi_communicator,
		const size_type	size,
		std::shared_ptr< const ProcessGrid > &	processGrid,
		const size_type	scalapackParalProcs,
		const bool	useELPA,
		const bool	useOnlyThumbRule = false
	)

Wrapper function to create a two dimensional processor grid for a square matrix in ScaLAPACKMatrix storage format.

Here is the call graph for this function:

Here is the caller graph for this function:

fillParallelOverlapMatrix()

template<typename T >

void dftefe::linearAlgebra::elpaScalaOpInternal::fillParallelOverlapMatrix	(	const T *	X,
		const size_type	XLocalSize,
		const size_type	numberVectors,
		const std::shared_ptr< const ProcessGrid > &	processGrid,
		const utils::mpi::MPIComm &	mpiComm,
		const size_type	wfcBlockSize,
		ScaLAPACKMatrix< T > &	overlapMatPar
	)

Computes Sc=X^{T}*Xc and stores in a parallel ScaLAPACK matrix. X^{T} is the subspaceVectorsArray stored in the column major format (N x M). Sc is the overlapMatPar.

The overlap matrix computation and filling is done in a blocked approach which avoids creation of full serial overlap matrix memory, and also avoids creation of another full X memory.

Here is the call graph for this function:

scaleScaLAPACKMat()

template<typename T >

void dftefe::linearAlgebra::elpaScalaOpInternal::scaleScaLAPACKMat	(	const std::shared_ptr< const ProcessGrid > &	processGrid,
		ScaLAPACKMatrix< T > &	mat,
		const T	scalar
	)

scale a ScaLAPACKMat with a scalar

setupELPAHandleParameters()

void dftefe::linearAlgebra::elpaScalaOpInternal::setupELPAHandleParameters	(	const utils::mpi::MPIComm &	mpi_communicator,
		utils::mpi::MPIComm &	processGridCommunicatorActive,
		const std::shared_ptr< const ProcessGrid > &	processGrid,
		const size_type	na,
		const size_type	nev,
		const size_type	blockSize,
		elpa_t &	elpaHandle,
		const bool	useELPADeviceKernel = false
	)

Contains internal functions used in linearAlgebra.

Author

Sambit Das, Aviup Sircar

setup ELPA parameters.

Here is the call graph for this function:

Here is the caller graph for this function:

subspaceRotation()

template<typename ValueType , typename utils::MemorySpace memorySpace>

void dftefe::linearAlgebra::elpaScalaOpInternal::subspaceRotation	(	ValueType *	X,
		const size_type	M,
		const size_type	N,
		const std::shared_ptr< const ProcessGrid > &	processGrid,
		const utils::mpi::MPIComm &	mpiCommDomain,
		LinAlgOpContext< memorySpace > &	linAlgOpContext,
		const ScaLAPACKMatrix< ValueType > &	rotationMatPar,
		const size_type	subspaceRotDofsBlockSize,
		const size_type	wfcBlockSize,
		const bool	rotationMatTranspose = false,
		const bool	isRotationMatLowerTria = false
	)

Computes X^{T}=Q*X^{T} inplace. X^{T} is the subspaceVectorsArray stored in the column major format (N x M). Q is rotationMatPar (N x N).

The subspace rotation inside this function is done in a blocked approach which avoids creation of full serial rotation matrix memory, and also avoids creation of another full subspaceVectorsArray memory. subspaceVectorsArrayLocalSize=N*M

Here is the call graph for this function: