Inheritance diagram for Opm::WellContributionsCuda:

Public Member Functions
void	setCudaStream (cudaStream_t stream)
	Set a cudaStream to be used. More...

void	apply (double d_x, double d_y)
	Apply all Wells in this object performs y -= (C^T * (D^-1 * (B*x))) for all Wells. More...

Public Member Functions inherited from Opm::WellContributions
unsigned int	getNumWells ()

void	addNumBlocks (unsigned int numBlocks)
	Indicate how large the next StandardWell is, this function cannot be called after alloc() is called. More...

void	alloc ()
	Allocate memory for the StandardWells.

virtual	~WellContributions ()=default
	Empty destructor.

void	setBlockSize (unsigned int dim, unsigned int dim_wells)
	Indicate how large the blocks of the StandardWell (C and B) are. More...

void	setVectorSize (unsigned N)
	Set size of vector that the wells are applied to. More...

void	addMatrix (MatrixType type, int colIndices, double values, unsigned int val_size)
	Store a matrix in this object, in blocked csr format, can only be called after alloc() is called. More...

void	addMultisegmentWellContribution (unsigned int dim, unsigned int dim_wells, unsigned int Mb, std::vector< double > &Bvalues, std::vector< unsigned int > &BcolIndices, std::vector< unsigned int > &BrowPointers, unsigned int DnumBlocks, double Dvalues, UMFPackIndex DcolPointers, UMFPackIndex *DrowIndices, std::vector< double > &Cvalues)
	Add a MultisegmentWellContribution, actually creates an object on heap that is destroyed in the destructor Matrices C and B are passed in Blocked CSR, matrix D in CSC. More...

Protected Member Functions
void	APIalloc () override
	Allocate memory for the StandardWells. More...

void	APIaddMatrix (MatrixType type, int colIndices, double values, unsigned int val_size) override
	Store a matrix in this object, in blocked csr format, can only be called after alloc() is called. More...

virtual void	APIalloc ()
	API specific allocation. More...

virtual void	APIaddMatrix (MatrixType, int , double , unsigned int)
	Api specific upload of matrix. More...

Protected Attributes
cudaStream_t	stream

double *	d_Cnnzs = nullptr

double *	d_Dnnzs = nullptr

double *	d_Bnnzs = nullptr

int *	d_Ccols = nullptr

int *	d_Bcols = nullptr

double *	d_z1 = nullptr

double *	d_z2 = nullptr

unsigned int *	d_val_pointers = nullptr

double *	h_x = nullptr

double *	h_y = nullptr

Protected Attributes inherited from Opm::WellContributions
bool	allocated = false

unsigned int	N

unsigned int	dim

unsigned int	dim_wells

unsigned int	num_blocks = 0

unsigned int	num_std_wells = 0

unsigned int	num_ms_wells = 0

unsigned int	num_blocks_so_far = 0

unsigned int	num_std_wells_so_far = 0

std::vector< unsigned int >	val_pointers

std::vector< std::unique_ptr< MultisegmentWellContribution > >	multisegments

Additional Inherited Members
Public Types inherited from Opm::WellContributions
enum class	MatrixType { C , D , B }
	StandardWell has C, D and B matrices that need to be copied.

using	UMFPackIndex = int

Static Public Member Functions inherited from Opm::WellContributions
static std::unique_ptr< WellContributions >	create (const std::string &accelerator_mode, bool useWellConn)

Member Function Documentation

◆ APIaddMatrix()

void Opm::WellContributionsCuda::APIaddMatrix	(	MatrixType	type,
		int *	colIndices,
		double *	values,
		unsigned int	val_size
	)

overrideprotectedvirtual

Store a matrix in this object, in blocked csr format, can only be called after alloc() is called.

Parameters

[in]	type	indicate if C, D or B is sent
[in]	colIndices	columnindices of blocks in C or B, ignored for D
[in]	values	array of nonzeroes
[in]	val_size	number of blocks in C or B, ignored for D

Reimplemented from Opm::WellContributions.

◆ APIalloc()

void Opm::WellContributionsCuda::APIalloc ( )

overrideprotectedvirtual

Allocate memory for the StandardWells.

Reimplemented from Opm::WellContributions.

◆ apply()

void Opm::WellContributionsCuda::apply	(	double *	d_x,
		double *	d_y
	)

Apply all Wells in this object performs y -= (C^T * (D^-1 * (B*x))) for all Wells.

Parameters

[in]	d_x	vector x, must be on GPU
[in,out]	d_y	vector y, must be on GPU

◆ setCudaStream()

void Opm::WellContributionsCuda::setCudaStream ( cudaStream_t stream )

Set a cudaStream to be used.

Parameters

[in] stream the cudaStream that is used to launch the kernel in

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

opm/simulators/linalg/bda/cuda/cuWellContributions.hpp

Public Member Functions

Protected Member Functions

Protected Attributes

Additional Inherited Members

Member Function Documentation

◆ APIaddMatrix()

◆ APIalloc()

◆ apply()

◆ setCudaStream()