Opm::Accelerator::rocsparseBILU0< Scalar, block_size > Class Template Reference

This class implements a Blocked ILU0 preconditioner The decomposition is done on GPU, using exact decomposition, or ChowPatel decomposition The preconditioner is applied via two exact triangular solves. More...

#include <rocsparseBILU0.hpp>

Inheritance diagram for Opm::Accelerator::rocsparseBILU0< Scalar, block_size >:

Public Member Functions
	rocsparseBILU0 (int verbosity_)
bool	initialize (std::shared_ptr< BlockedMatrix< Scalar > > matrix, std::shared_ptr< BlockedMatrix< Scalar > > jacMatrix, rocsparse_int *d_Arows, rocsparse_int *d_Acols) override
	Initialize GPU and allocate memory.
bool	analyze_matrix (BlockedMatrix< Scalar > *mat) override
	Analysis, extract parallelism if specified.
bool	analyze_matrix (BlockedMatrix< Scalar > *mat, BlockedMatrix< Scalar > *jacMat) override
	Analysis, extract parallelism if specified.
bool	create_preconditioner (BlockedMatrix< Scalar > *mat) override
	ILU decomposition.
bool	create_preconditioner (BlockedMatrix< Scalar > *mat, BlockedMatrix< Scalar > *jacMat) override
	ILU decomposition.
void	apply (const Scalar &y, Scalar &x) override
	Apply preconditioner, x = prec(y) via Lz = y and Ux = z.
void	copy_system_to_gpu (Scalar *mVals) override
	Copy matrix A values to GPU.
void	update_system_on_gpu (Scalar *b) override
	Reassign pointers, in case the addresses of the Dune variables have changed --> TODO: check when/if we need this method.
Public Member Functions inherited from Opm::Accelerator::rocsparsePreconditioner< Scalar, block_size >
void	set_matrix_analysis (rocsparse_mat_descr descr_L, rocsparse_mat_descr descr_U)
void	set_context (rocsparse_handle handle, rocsparse_direction dir, rocsparse_operation operation, hipStream_t stream)
void	setJacMat (const BlockedMatrix< Scalar > &jacMat)
Public Member Functions inherited from Opm::Accelerator::Preconditioner< Scalar, block_size, ApplyScalar >
virtual void	apply (const ApplyScalar &y, ApplyScalar &x)=0

Additional Inherited Members
Static Public Member Functions inherited from Opm::Accelerator::rocsparsePreconditioner< Scalar, block_size >
static std::unique_ptr< rocsparsePreconditioner< Scalar, block_size > >	create (PreconditionerType type, int verbosity)
Static Public Member Functions inherited from Opm::Accelerator::Preconditioner< Scalar, block_size, ApplyScalar >
static std::unique_ptr< Preconditioner >	create (PreconditionerType type, bool opencl_ilu_parallel, int verbosity)
Public Attributes inherited from Opm::Accelerator::rocsparsePreconditioner< Scalar, block_size >
int	nnzbs_prec = 0
bool	useJacMatrix = false
std::shared_ptr< BlockedMatrix< Scalar > >	jacMat {}
Protected Member Functions inherited from Opm::Accelerator::rocsparsePreconditioner< Scalar, block_size >
	rocsparsePreconditioner (int verbosity_)
Protected Member Functions inherited from Opm::Accelerator::Preconditioner< Scalar, block_size, ApplyScalar >
	Preconditioner (int verbosity_)
Protected Attributes inherited from Opm::Accelerator::rocsparsePreconditioner< Scalar, block_size >
rocsparse_handle	handle
rocsparse_direction	dir = rocsparse_direction_row
rocsparse_operation	operation = rocsparse_operation_none
rocsparse_mat_descr	descr_L
rocsparse_mat_descr	descr_U
hipStream_t	stream
Protected Attributes inherited from Opm::Accelerator::Preconditioner< Scalar, block_size, ApplyScalar >
int	N = 0
int	Nb = 0
int	nnz = 0
int	nnzb = 0
int	verbosity = 0

Detailed Description

template<class Scalar, unsigned int block_size>
class Opm::Accelerator::rocsparseBILU0< Scalar, block_size >

This class implements a Blocked ILU0 preconditioner The decomposition is done on GPU, using exact decomposition, or ChowPatel decomposition The preconditioner is applied via two exact triangular solves.

Member Function Documentation

analyze_matrix() [1/2]

template<class Scalar , unsigned int block_size>

bool Opm::Accelerator::rocsparseBILU0< Scalar, block_size >::analyze_matrix ( BlockedMatrix< Scalar > *  mat )

overridevirtual

Analysis, extract parallelism if specified.

Parameters

[in]

mat

matrix A

Implements Opm::Accelerator::Preconditioner< Scalar, block_size, ApplyScalar >.

analyze_matrix() [2/2]

template<class Scalar , unsigned int block_size>

bool Opm::Accelerator::rocsparseBILU0< Scalar, block_size >::analyze_matrix ( BlockedMatrix< Scalar > *  mat, BlockedMatrix< Scalar > *  jacMat  )

overridevirtual

Analysis, extract parallelism if specified.

Parameters

[in]	mat	matrix A
[in]	jacMat	matrix for preconditioner, analyze this as well

Implements Opm::Accelerator::Preconditioner< Scalar, block_size, ApplyScalar >.

apply()

template<class Scalar , unsigned int block_size>

void Opm::Accelerator::rocsparseBILU0< Scalar, block_size >::apply ( const Scalar &  y, Scalar &  x  )

override

Apply preconditioner, x = prec(y) via Lz = y and Ux = z.

Parameters

[in]	y	Input y vector
[out]	x	Output x vector

copy_system_to_gpu()

template<class Scalar , unsigned int block_size>

void Opm::Accelerator::rocsparseBILU0< Scalar, block_size >::copy_system_to_gpu ( Scalar *  mVals )

overridevirtual

Copy matrix A values to GPU.

Parameters

[in]

mVals

Input values

Implements Opm::Accelerator::rocsparsePreconditioner< Scalar, block_size >.

create_preconditioner() [1/2]

template<class Scalar , unsigned int block_size>

bool Opm::Accelerator::rocsparseBILU0< Scalar, block_size >::create_preconditioner ( BlockedMatrix< Scalar > *  mat )

overridevirtual

ILU decomposition.

Parameters

[in]

mat

matrix A to decompose

Implements Opm::Accelerator::Preconditioner< Scalar, block_size, ApplyScalar >.

create_preconditioner() [2/2]

template<class Scalar , unsigned int block_size>

bool Opm::Accelerator::rocsparseBILU0< Scalar, block_size >::create_preconditioner ( BlockedMatrix< Scalar > *  mat, BlockedMatrix< Scalar > *  jacMat  )

overridevirtual

ILU decomposition.

Parameters

[in]	mat	matrix A
[in]	jacMat	matrix for preconditioner, decompose this one if used

Implements Opm::Accelerator::Preconditioner< Scalar, block_size, ApplyScalar >.

initialize()

template<class Scalar , unsigned int block_size>

bool Opm::Accelerator::rocsparseBILU0< Scalar, block_size >::initialize ( std::shared_ptr< BlockedMatrix< Scalar > >  matrix, std::shared_ptr< BlockedMatrix< Scalar > >  jacMatrix, rocsparse_int *  d_Arows, rocsparse_int *  d_Acols  )

overridevirtual

Initialize GPU and allocate memory.

Parameters

[in]	matrix	matrix A
[in]	jacMatrix	matrix for preconditioner

Implements Opm::Accelerator::rocsparsePreconditioner< Scalar, block_size >.

update_system_on_gpu()

template<class Scalar , unsigned int block_size>

void Opm::Accelerator::rocsparseBILU0< Scalar, block_size >::update_system_on_gpu ( Scalar *  b )

overridevirtual

Reassign pointers, in case the addresses of the Dune variables have changed --> TODO: check when/if we need this method.

Update GPU values after a new assembly is done

Parameters

[in]

b

New b vector

Implements Opm::Accelerator::rocsparsePreconditioner< Scalar, block_size >.

---

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

• opm/simulators/linalg/gpubridge/rocm/rocsparseBILU0.hpp
• opm/simulators/linalg/gpubridge/rocm/rocsparseBILU0.cpp