Class CudaSolveNativeMethods.Sparse

The cuSolverSP library was mainly designed to a solve sparse linear system AxB and the least-squares problem x = argmin||A*z-b||

Inheritance

System.Object

CudaSolveNativeMethods.Sparse

Inherited Members

System.Object.Equals(System.Object)

System.Object.Equals(System.Object, System.Object)

System.Object.GetHashCode()

System.Object.GetType()

System.Object.MemberwiseClone()

System.Object.ReferenceEquals(System.Object, System.Object)

System.Object.ToString()

Namespace: ManagedCuda.CudaSolve

Assembly: CudaSolve.dll

Syntax

public static class Sparse

Methods

cusolverSpCcsreigsHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, cuFloatComplex[], Int32[], Int32[], cuFloatComplex, cuFloatComplex, ref Int32)

Declaration

public static cusolverStatus cusolverSpCcsreigsHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, cuFloatComplex[] csrValA, int[] csrRowPtrA, int[] csrColIndA, cuFloatComplex left_bottom_corner, cuFloatComplex right_upper_corner, ref int num_eigs)

Parameters

Type	Name	Description
cusolverSpHandle	handle
System.Int32	m
System.Int32	nnz
cusparseMatDescr	descrA
cuFloatComplex[]	csrValA
System.Int32[]	csrRowPtrA
System.Int32[]	csrColIndA
cuFloatComplex	left_bottom_corner
cuFloatComplex	right_upper_corner
System.Int32	num_eigs

Returns

Type	Description
cusolverStatus

cusolverSpCcsreigvsi(cusolverSpHandle, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, cuFloatComplex, CUdeviceptr, Int32, Single, ref cuFloatComplex, CUdeviceptr)

This function solves the simple eigenvalue problem Ax=lambdax by shift-inverse method.

Declaration

public static cusolverStatus cusolverSpCcsreigvsi(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, CUdeviceptr csrValA, CUdeviceptr csrRowPtrA, CUdeviceptr csrColIndA, cuFloatComplex mu0, CUdeviceptr x0, int maxite, float tol, ref cuFloatComplex mu, CUdeviceptr x)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
CUdeviceptr	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
cuFloatComplex	mu0	initial guess of eigenvalue.
CUdeviceptr	x0	initial guess of eigenvector, a vecotr of size m.
System.Int32	maxite	maximum iterations in shift-inverse method.
System.Single	tol	tolerance for convergence.
cuFloatComplex	mu	approximated eigenvalue nearest mu0 under tolerance.
CUdeviceptr	x	approximated eigenvector of size m.

Returns

Type	Description
cusolverStatus

cusolverSpCcsreigvsiHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, cuFloatComplex[], Int32[], Int32[], cuFloatComplex, cuFloatComplex[], Int32, Single, ref cuFloatComplex, cuFloatComplex[])

This function solves the simple eigenvalue problem Ax=lambdax by shift-inverse method.

Declaration

public static cusolverStatus cusolverSpCcsreigvsiHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, cuFloatComplex[] csrValA, int[] csrRowPtrA, int[] csrColIndA, cuFloatComplex mu0, cuFloatComplex[] x0, int maxite, float tol, ref cuFloatComplex mu, cuFloatComplex[] x)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
cuFloatComplex[]	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
cuFloatComplex	mu0	initial guess of eigenvalue.
cuFloatComplex[]	x0	initial guess of eigenvector, a vecotr of size m.
System.Int32	maxite	maximum iterations in shift-inverse method.
System.Single	tol	tolerance for convergence.
cuFloatComplex	mu	approximated eigenvalue nearest mu0 under tolerance.
cuFloatComplex[]	x	approximated eigenvector of size m.

Returns

Type	Description
cusolverStatus

cusolverSpCcsrlsqvqrHost(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, cuFloatComplex[], Int32[], Int32[], cuFloatComplex[], Single, ref Int32, cuFloatComplex[], Int32[], ref Single)

This function solves the following least-square problem x = argmin||A*z-b||

Declaration

public static cusolverStatus cusolverSpCcsrlsqvqrHost(cusolverSpHandle handle, int m, int n, int nnz, cusparseMatDescr descrA, cuFloatComplex[] csrValA, int[] csrRowPtrA, int[] csrColIndA, cuFloatComplex[] b, float tol, ref int rankA, cuFloatComplex[] x, int[] p, ref float min_norm)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of matrix A.
System.Int32	n	number of columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
cuFloatComplex[]	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
cuFloatComplex[]	b	right hand side vector of size m.
System.Single	tol	tolerance to decide rank of A.
System.Int32	rankA	numerical rank of A.
cuFloatComplex[]	x	solution vector of size n, x=pinv(A)*b.
System.Int32[]	p	a vector of size n, which represents the permuation matrix P satisfying AP^T=QR.
System.Single	min_norm	||A*x-b||, x=pinv(A)*b.

Returns

Type	Description
cusolverStatus

cusolverSpCcsrlsvchol(cusolverSpHandle, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Single, Int32, CUdeviceptr, ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpCcsrlsvchol(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, CUdeviceptr csrVal, CUdeviceptr csrRowPtr, CUdeviceptr csrColInd, CUdeviceptr b, float tol, int reorder, CUdeviceptr x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrVal	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
CUdeviceptr	csrRowPtr	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColInd	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
CUdeviceptr	b	right hand side vector of size m.
System.Single	tol	tolerance to decide singularity.
System.Int32	reorder	no effect.
CUdeviceptr	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is symmetric postive definite.

Returns

Type	Description
cusolverStatus

cusolverSpCcsrlsvcholHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, cuFloatComplex[], Int32[], Int32[], cuFloatComplex[], Single, Int32, cuFloatComplex[], ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpCcsrlsvcholHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, cuFloatComplex[] csrVal, int[] csrRowPtr, int[] csrColInd, cuFloatComplex[] b, float tol, int reorder, cuFloatComplex[] x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
cuFloatComplex[]	csrVal	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtr	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColInd	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
cuFloatComplex[]	b	right hand side vector of size m.
System.Single	tol	tolerance to decide singularity.
System.Int32	reorder	no effect.
cuFloatComplex[]	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is symmetric postive definite.

Returns

Type	Description
cusolverStatus

cusolverSpCcsrlsvluHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, cuFloatComplex[], Int32[], Int32[], cuFloatComplex[], Single, Int32, cuFloatComplex[], ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpCcsrlsvluHost(cusolverSpHandle handle, int n, int nnzA, cusparseMatDescr descrA, cuFloatComplex[] csrValA, int[] csrRowPtrA, int[] csrColIndA, cuFloatComplex[] b, float tol, int reorder, cuFloatComplex[] x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	n	number of rows and columns of matrix A.
System.Int32	nnzA	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
cuFloatComplex[]	csrValA	array of nnzA (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnzA (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
cuFloatComplex[]	b	right hand side vector of size n.
System.Single	tol	tolerance to decide if singular or not.
System.Int32	reorder	no ordering if reorder=0. Otherwise, symrcm is used to reduce zero fill-in.
cuFloatComplex[]	x	solution vector of size n, x = inv(A)*b.
System.Int32	singularity	-1 if A is invertible. Otherwise, first index j such that U(j,j)≈0

Returns

Type	Description
cusolverStatus

cusolverSpCcsrlsvqr(cusolverSpHandle, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Single, Int32, CUdeviceptr, ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpCcsrlsvqr(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, CUdeviceptr csrValA, CUdeviceptr csrRowPtrA, CUdeviceptr csrColIndA, CUdeviceptr b, float tol, int reorder, CUdeviceptr x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
CUdeviceptr	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
CUdeviceptr	b	right hand side vector of size m.
System.Single	tol	tolerance to decide if singular or not.
System.Int32	reorder	no effect.
CUdeviceptr	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is invertible. Otherwise, first index j such that R(j,j)≈0

Returns

Type	Description
cusolverStatus

cusolverSpCcsrlsvqrHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, cuFloatComplex[], Int32[], Int32[], cuFloatComplex[], Single, Int32, cuFloatComplex[], ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpCcsrlsvqrHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, cuFloatComplex[] csrValA, int[] csrRowPtrA, int[] csrColIndA, cuFloatComplex[] b, float tol, int reorder, cuFloatComplex[] x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
cuFloatComplex[]	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
cuFloatComplex[]	b	right hand side vector of size m.
System.Single	tol	tolerance to decide if singular or not.
System.Int32	reorder	no effect.
cuFloatComplex[]	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is invertible. Otherwise, first index j such that R(j,j)≈0

Returns

Type	Description
cusolverStatus

cusolverSpCcsrqrBufferInfoBatched(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Int32, csrqrInfo, ref SizeT, ref SizeT)

The batched sparse QR factorization is used to solve either a set of least-squares problems or a set of linear systems

Declaration

public static cusolverStatus cusolverSpCcsrqrBufferInfoBatched(cusolverSpHandle handle, int m, int n, int nnz, cusparseMatDescr descrA, CUdeviceptr csrVal, CUdeviceptr csrRowPtr, CUdeviceptr csrColInd, int batchSize, csrqrInfo info, ref SizeT internalDataInBytes, ref SizeT workspaceInBytes)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of each matrix Aj.
System.Int32	n	number of columns of each matrix Aj.
System.Int32	nnz	number of nonzeros of each matrix Aj. It is the size csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrVal	array of nnzA*batchSize nonzero elements of matrices A0, A1, .... All matrices are aggregated one after another.
CUdeviceptr	csrRowPtr	integer array of m+1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColInd	integer array of nnzAcolumn indices of the nonzero elements of each matrix Aj.
System.Int32	batchSize	number of systems to be solved.
csrqrInfo	info	opaque structure for QR factorization.
SizeT	internalDataInBytes	number of bytes of the internal data.
SizeT	workspaceInBytes	number of bytes of the buffer in numerical factorization.

Returns

Type	Description
cusolverStatus

cusolverSpCcsrqrsvBatched(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Int32, csrqrInfo, CUdeviceptr)

The batched sparse QR factorization is used to solve either a set of least-squares problems or a set of linear systems

Declaration

public static cusolverStatus cusolverSpCcsrqrsvBatched(cusolverSpHandle handle, int m, int n, int nnz, cusparseMatDescr descrA, CUdeviceptr csrValA, CUdeviceptr csrRowPtrA, CUdeviceptr csrColIndA, CUdeviceptr b, CUdeviceptr x, int batchSize, csrqrInfo info, CUdeviceptr pBuffer)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of each matrix Aj.
System.Int32	n	number of columns of each matrix Aj.
System.Int32	nnz	number of nonzeros of each matrix Aj. It is the size csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrValA	array of nnzA*batchSize nonzero elements of matrices A0, A1, .... All matrices are aggregated one after another.
CUdeviceptr	csrRowPtrA	integer array of m+1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColIndA	integer array of nnzAcolumn indices of the nonzero elements of each matrix Aj.
CUdeviceptr	b	array of m*batchSize of right-hand-side vectors b0, b1, .... All vectors are aggregated one after another.
CUdeviceptr	x	array of m*batchSize of solution vectors x0, x1, .... All vectors are aggregated one after another.
System.Int32	batchSize	number of systems to be solved.
csrqrInfo	info	opaque structure for QR factorization.
CUdeviceptr	pBuffer	buffer allocated by the user, the size is returned by cusolverSpXcsrqrBufferInfoBatched().

Returns

Type	Description
cusolverStatus

cusolverSpCreate(ref cusolverSpHandle)

This function initializes the cuSolverSP library and creates a handle on the cuSolver context. It must be called before any other cuSolverSP API function is invoked. It allocates hardware resources necessary for accessing the GPU.

Declaration

public static cusolverStatus cusolverSpCreate(ref cusolverSpHandle handle)

Parameters

Type	Name	Description
cusolverSpHandle	handle	the pointer to the handle to the cuSolverSP context.

Returns

Type	Description
cusolverStatus

cusolverSpCreateCsrqrInfo(ref csrqrInfo)

The batched sparse QR factorization is used to solve either a set of least-squares problems or a set of linear systems

Declaration

public static cusolverStatus cusolverSpCreateCsrqrInfo(ref csrqrInfo info)

Parameters

Type	Name	Description
csrqrInfo	info	opaque structure for QR factorization.

Returns

Type	Description
cusolverStatus

cusolverSpDcsreigsHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Double[], Int32[], Int32[], cuDoubleComplex, cuDoubleComplex, ref Int32)

Declaration

public static cusolverStatus cusolverSpDcsreigsHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, double[] csrValA, int[] csrRowPtrA, int[] csrColIndA, cuDoubleComplex left_bottom_corner, cuDoubleComplex right_upper_corner, ref int num_eigs)

Parameters

Type	Name	Description
cusolverSpHandle	handle
System.Int32	m
System.Int32	nnz
cusparseMatDescr	descrA
System.Double[]	csrValA
System.Int32[]	csrRowPtrA
System.Int32[]	csrColIndA
cuDoubleComplex	left_bottom_corner
cuDoubleComplex	right_upper_corner
System.Int32	num_eigs

Returns

Type	Description
cusolverStatus

cusolverSpDcsreigvsi(cusolverSpHandle, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Double, CUdeviceptr, Int32, Double, ref Double, CUdeviceptr)

This function solves the simple eigenvalue problem Ax=lambdax by shift-inverse method.

Declaration

public static cusolverStatus cusolverSpDcsreigvsi(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, CUdeviceptr csrValA, CUdeviceptr csrRowPtrA, CUdeviceptr csrColIndA, double mu0, CUdeviceptr x0, int maxite, double tol, ref double mu, CUdeviceptr x)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
CUdeviceptr	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Double	mu0	initial guess of eigenvalue.
CUdeviceptr	x0	initial guess of eigenvector, a vecotr of size m.
System.Int32	maxite	maximum iterations in shift-inverse method.
System.Double	tol	tolerance for convergence.
System.Double	mu	approximated eigenvalue nearest mu0 under tolerance.
CUdeviceptr	x	approximated eigenvector of size m.

Returns

Type	Description
cusolverStatus

cusolverSpDcsreigvsiHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Double[], Int32[], Int32[], Double, Double[], Int32, Double, ref Double, Double[])

This function solves the simple eigenvalue problem Ax=lambdax by shift-inverse method.

Declaration

public static cusolverStatus cusolverSpDcsreigvsiHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, double[] csrValA, int[] csrRowPtrA, int[] csrColIndA, double mu0, double[] x0, int maxite, double tol, ref double mu, double[] x)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Double[]	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Double	mu0	initial guess of eigenvalue.
System.Double[]	x0	initial guess of eigenvector, a vecotr of size m.
System.Int32	maxite	maximum iterations in shift-inverse method.
System.Double	tol	tolerance for convergence.
System.Double	mu	approximated eigenvalue nearest mu0 under tolerance.
System.Double[]	x	approximated eigenvector of size m.

Returns

Type	Description
cusolverStatus

cusolverSpDcsrlsqvqrHost(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, Double[], Int32[], Int32[], Double[], Double, ref Int32, Double[], Int32[], ref Double)

This function solves the following least-square problem x = argmin||A*z-b||

Declaration

public static cusolverStatus cusolverSpDcsrlsqvqrHost(cusolverSpHandle handle, int m, int n, int nnz, cusparseMatDescr descrA, double[] csrValA, int[] csrRowPtrA, int[] csrColIndA, double[] b, double tol, ref int rankA, double[] x, int[] p, ref double min_norm)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of matrix A.
System.Int32	n	number of columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Double[]	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Double[]	b	right hand side vector of size m.
System.Double	tol	tolerance to decide rank of A.
System.Int32	rankA	numerical rank of A.
System.Double[]	x	solution vector of size n, x=pinv(A)*b.
System.Int32[]	p	a vector of size n, which represents the permuation matrix P satisfying AP^T=QR.
System.Double	min_norm	||A*x-b||, x=pinv(A)*b.

Returns

Type	Description
cusolverStatus

cusolverSpDcsrlsvchol(cusolverSpHandle, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Double, Int32, CUdeviceptr, ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpDcsrlsvchol(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, CUdeviceptr csrVal, CUdeviceptr csrRowPtr, CUdeviceptr csrColInd, CUdeviceptr b, double tol, int reorder, CUdeviceptr x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrVal	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
CUdeviceptr	csrRowPtr	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColInd	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
CUdeviceptr	b	right hand side vector of size m.
System.Double	tol	tolerance to decide singularity.
System.Int32	reorder	no effect.
CUdeviceptr	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is symmetric postive definite.

Returns

Type	Description
cusolverStatus

cusolverSpDcsrlsvcholHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Double[], Int32[], Int32[], Double[], Double, Int32, Double[], ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpDcsrlsvcholHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, double[] csrVal, int[] csrRowPtr, int[] csrColInd, double[] b, double tol, int reorder, double[] x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Double[]	csrVal	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtr	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColInd	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Double[]	b	right hand side vector of size m.
System.Double	tol	tolerance to decide singularity.
System.Int32	reorder	no effect.
System.Double[]	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is symmetric postive definite.

Returns

Type	Description
cusolverStatus

cusolverSpDcsrlsvluHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Double[], Int32[], Int32[], Double[], Double, Int32, Double[], ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpDcsrlsvluHost(cusolverSpHandle handle, int n, int nnzA, cusparseMatDescr descrA, double[] csrValA, int[] csrRowPtrA, int[] csrColIndA, double[] b, double tol, int reorder, double[] x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	n	number of rows and columns of matrix A.
System.Int32	nnzA	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Double[]	csrValA	array of nnzA (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnzA (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Double[]	b	right hand side vector of size n.
System.Double	tol	tolerance to decide if singular or not.
System.Int32	reorder	no ordering if reorder=0. Otherwise, symrcm is used to reduce zero fill-in.
System.Double[]	x	solution vector of size n, x = inv(A)*b.
System.Int32	singularity	-1 if A is invertible. Otherwise, first index j such that U(j,j)≈0

Returns

Type	Description
cusolverStatus

cusolverSpDcsrlsvqr(cusolverSpHandle, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Double, Int32, CUdeviceptr, ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpDcsrlsvqr(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, CUdeviceptr csrValA, CUdeviceptr csrRowPtrA, CUdeviceptr csrColIndA, CUdeviceptr b, double tol, int reorder, CUdeviceptr x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
CUdeviceptr	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
CUdeviceptr	b	right hand side vector of size m.
System.Double	tol	tolerance to decide if singular or not.
System.Int32	reorder	no effect.
CUdeviceptr	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is invertible. Otherwise, first index j such that R(j,j)≈0

Returns

Type	Description
cusolverStatus

cusolverSpDcsrlsvqrHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Double[], Int32[], Int32[], Double[], Double, Int32, Double[], ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpDcsrlsvqrHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, double[] csrValA, int[] csrRowPtrA, int[] csrColIndA, double[] b, double tol, int reorder, double[] x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Double[]	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Double[]	b	right hand side vector of size m.
System.Double	tol	tolerance to decide if singular or not.
System.Int32	reorder	no effect.
System.Double[]	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is invertible. Otherwise, first index j such that R(j,j)≈0

Returns

Type	Description
cusolverStatus

cusolverSpDcsrqrBufferInfoBatched(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Int32, csrqrInfo, ref SizeT, ref SizeT)

The batched sparse QR factorization is used to solve either a set of least-squares problems or a set of linear systems

Declaration

public static cusolverStatus cusolverSpDcsrqrBufferInfoBatched(cusolverSpHandle handle, int m, int n, int nnz, cusparseMatDescr descrA, CUdeviceptr csrVal, CUdeviceptr csrRowPtr, CUdeviceptr csrColInd, int batchSize, csrqrInfo info, ref SizeT internalDataInBytes, ref SizeT workspaceInBytes)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of each matrix Aj.
System.Int32	n	number of columns of each matrix Aj.
System.Int32	nnz	number of nonzeros of each matrix Aj. It is the size csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrVal	array of nnzA*batchSize nonzero elements of matrices A0, A1, .... All matrices are aggregated one after another.
CUdeviceptr	csrRowPtr	integer array of m+1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColInd	integer array of nnzAcolumn indices of the nonzero elements of each matrix Aj.
System.Int32	batchSize	number of systems to be solved.
csrqrInfo	info	opaque structure for QR factorization.
SizeT	internalDataInBytes	number of bytes of the internal data.
SizeT	workspaceInBytes	number of bytes of the buffer in numerical factorization.

Returns

Type	Description
cusolverStatus

cusolverSpDcsrqrsvBatched(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Int32, csrqrInfo, CUdeviceptr)

The batched sparse QR factorization is used to solve either a set of least-squares problems or a set of linear systems

Declaration

public static cusolverStatus cusolverSpDcsrqrsvBatched(cusolverSpHandle handle, int m, int n, int nnz, cusparseMatDescr descrA, CUdeviceptr csrValA, CUdeviceptr csrRowPtrA, CUdeviceptr csrColIndA, CUdeviceptr b, CUdeviceptr x, int batchSize, csrqrInfo info, CUdeviceptr pBuffer)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of each matrix Aj.
System.Int32	n	number of columns of each matrix Aj.
System.Int32	nnz	number of nonzeros of each matrix Aj. It is the size csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrValA	array of nnzA*batchSize nonzero elements of matrices A0, A1, .... All matrices are aggregated one after another.
CUdeviceptr	csrRowPtrA	integer array of m+1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColIndA	integer array of nnzAcolumn indices of the nonzero elements of each matrix Aj.
CUdeviceptr	b	array of m*batchSize of right-hand-side vectors b0, b1, .... All vectors are aggregated one after another.
CUdeviceptr	x	array of m*batchSize of solution vectors x0, x1, .... All vectors are aggregated one after another.
System.Int32	batchSize	number of systems to be solved.
csrqrInfo	info	opaque structure for QR factorization.
CUdeviceptr	pBuffer	buffer allocated by the user, the size is returned by cusolverSpXcsrqrBufferInfoBatched().

Returns

Type	Description
cusolverStatus

cusolverSpDestroy(cusolverSpHandle)

This function releases CPU-side resources used by the cuSolverSP library.

Declaration

public static cusolverStatus cusolverSpDestroy(cusolverSpHandle handle)

Parameters

Type	Name	Description
cusolverSpHandle	handle	the handle to the cuSolverSP context.

Returns

Type	Description
cusolverStatus

cusolverSpDestroyCsrqrInfo(csrqrInfo)

The batched sparse QR factorization is used to solve either a set of least-squares problems or a set of linear systems

Declaration

public static cusolverStatus cusolverSpDestroyCsrqrInfo(csrqrInfo info)

Parameters

Type	Name	Description
csrqrInfo	info	opaque structure for QR factorization.

Returns

Type	Description
cusolverStatus

cusolverSpGetStream(cusolverSpHandle, ref CUstream)

This function gets the stream to be used by the cuSolverSP library to execute its routines.

Declaration

public static cusolverStatus cusolverSpGetStream(cusolverSpHandle handle, ref CUstream streamId)

Parameters

Type	Name	Description
cusolverSpHandle	handle	the handle to the cuSolverSP context.
CUstream	streamId	the stream to be used by the library.

Returns

Type	Description
cusolverStatus

cusolverSpScsreigsHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Single[], Int32[], Int32[], cuFloatComplex, cuFloatComplex, ref Int32)

Declaration

public static cusolverStatus cusolverSpScsreigsHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, float[] csrValA, int[] csrRowPtrA, int[] csrColIndA, cuFloatComplex left_bottom_corner, cuFloatComplex right_upper_corner, ref int num_eigs)

Parameters

Type	Name	Description
cusolverSpHandle	handle
System.Int32	m
System.Int32	nnz
cusparseMatDescr	descrA
System.Single[]	csrValA
System.Int32[]	csrRowPtrA
System.Int32[]	csrColIndA
cuFloatComplex	left_bottom_corner
cuFloatComplex	right_upper_corner
System.Int32	num_eigs

Returns

Type	Description
cusolverStatus

cusolverSpScsreigvsi(cusolverSpHandle, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Single, CUdeviceptr, Int32, Single, ref Single, CUdeviceptr)

This function solves the simple eigenvalue problem Ax=lambdax by shift-inverse method.

Declaration

public static cusolverStatus cusolverSpScsreigvsi(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, CUdeviceptr csrValA, CUdeviceptr csrRowPtrA, CUdeviceptr csrColIndA, float mu0, CUdeviceptr x0, int maxite, float tol, ref float mu, CUdeviceptr x)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
CUdeviceptr	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Single	mu0	initial guess of eigenvalue.
CUdeviceptr	x0	initial guess of eigenvector, a vecotr of size m.
System.Int32	maxite	maximum iterations in shift-inverse method.
System.Single	tol	tolerance for convergence.
System.Single	mu	approximated eigenvalue nearest mu0 under tolerance.
CUdeviceptr	x	approximated eigenvector of size m.

Returns

Type	Description
cusolverStatus

cusolverSpScsreigvsiHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Single[], Int32[], Int32[], Single, Single[], Int32, Single, ref Single, Single[])

This function solves the simple eigenvalue problem Ax=lambdax by shift-inverse method.

Declaration

public static cusolverStatus cusolverSpScsreigvsiHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, float[] csrValA, int[] csrRowPtrA, int[] csrColIndA, float mu0, float[] x0, int maxite, float tol, ref float mu, float[] x)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Single[]	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Single	mu0	initial guess of eigenvalue.
System.Single[]	x0	initial guess of eigenvector, a vecotr of size m.
System.Int32	maxite	maximum iterations in shift-inverse method.
System.Single	tol	tolerance for convergence.
System.Single	mu	approximated eigenvalue nearest mu0 under tolerance.
System.Single[]	x	approximated eigenvector of size m.

Returns

Type	Description
cusolverStatus

cusolverSpScsrlsqvqrHost(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, Single[], Int32[], Int32[], Single[], Single, ref Int32, Single[], Int32[], ref Single)

This function solves the following least-square problem x = argmin||A*z-b||

Declaration

public static cusolverStatus cusolverSpScsrlsqvqrHost(cusolverSpHandle handle, int m, int n, int nnz, cusparseMatDescr descrA, float[] csrValA, int[] csrRowPtrA, int[] csrColIndA, float[] b, float tol, ref int rankA, float[] x, int[] p, ref float min_norm)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of matrix A.
System.Int32	n	number of columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Single[]	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Single[]	b	right hand side vector of size m.
System.Single	tol	tolerance to decide rank of A.
System.Int32	rankA	numerical rank of A.
System.Single[]	x	solution vector of size n, x=pinv(A)*b.
System.Int32[]	p	a vector of size n, which represents the permuation matrix P satisfying AP^T=QR.
System.Single	min_norm	||A*x-b||, x=pinv(A)*b.

Returns

Type	Description
cusolverStatus

cusolverSpScsrlsvchol(cusolverSpHandle, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Single, Int32, CUdeviceptr, ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpScsrlsvchol(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, CUdeviceptr csrVal, CUdeviceptr csrRowPtr, CUdeviceptr csrColInd, CUdeviceptr b, float tol, int reorder, CUdeviceptr x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrVal	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
CUdeviceptr	csrRowPtr	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColInd	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
CUdeviceptr	b	right hand side vector of size m.
System.Single	tol	tolerance to decide singularity.
System.Int32	reorder	no effect.
CUdeviceptr	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is symmetric postive definite.

Returns

Type	Description
cusolverStatus

cusolverSpScsrlsvcholHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Single[], Int32[], Int32[], Single[], Single, Int32, Single[], ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpScsrlsvcholHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, float[] csrVal, int[] csrRowPtr, int[] csrColInd, float[] b, float tol, int reorder, float[] x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Single[]	csrVal	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtr	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColInd	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Single[]	b	right hand side vector of size m.
System.Single	tol	tolerance to decide singularity.
System.Int32	reorder	no effect.
System.Single[]	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is symmetric postive definite.

Returns

Type	Description
cusolverStatus

cusolverSpScsrlsvluHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Single[], Int32[], Int32[], Single[], Single, Int32, Single[], ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpScsrlsvluHost(cusolverSpHandle handle, int n, int nnzA, cusparseMatDescr descrA, float[] csrValA, int[] csrRowPtrA, int[] csrColIndA, float[] b, float tol, int reorder, float[] x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	n	number of rows and columns of matrix A.
System.Int32	nnzA	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Single[]	csrValA	array of nnzA (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnzA (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Single[]	b	right hand side vector of size n.
System.Single	tol	tolerance to decide if singular or not.
System.Int32	reorder	no ordering if reorder=0. Otherwise, symrcm is used to reduce zero fill-in.
System.Single[]	x	solution vector of size n, x = inv(A)*b.
System.Int32	singularity	-1 if A is invertible. Otherwise, first index j such that U(j,j)≈0

Returns

Type	Description
cusolverStatus

cusolverSpScsrlsvqr(cusolverSpHandle, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Single, Int32, CUdeviceptr, ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpScsrlsvqr(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, CUdeviceptr csrValA, CUdeviceptr csrRowPtrA, CUdeviceptr csrColIndA, CUdeviceptr b, float tol, int reorder, CUdeviceptr x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
CUdeviceptr	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
CUdeviceptr	b	right hand side vector of size m.
System.Single	tol	tolerance to decide if singular or not.
System.Int32	reorder	no effect.
CUdeviceptr	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is invertible. Otherwise, first index j such that R(j,j)≈0

Returns

Type	Description
cusolverStatus

cusolverSpScsrlsvqrHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Single[], Int32[], Int32[], Single[], Single, Int32, Single[], ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpScsrlsvqrHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, float[] csrValA, int[] csrRowPtrA, int[] csrColIndA, float[] b, float tol, int reorder, float[] x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Single[]	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Single[]	b	right hand side vector of size m.
System.Single	tol	tolerance to decide if singular or not.
System.Int32	reorder	no effect.
System.Single[]	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is invertible. Otherwise, first index j such that R(j,j)≈0

Returns

Type	Description
cusolverStatus

cusolverSpScsrqrBufferInfoBatched(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Int32, csrqrInfo, ref SizeT, ref SizeT)

The batched sparse QR factorization is used to solve either a set of least-squares problems or a set of linear systems

Declaration

public static cusolverStatus cusolverSpScsrqrBufferInfoBatched(cusolverSpHandle handle, int m, int n, int nnz, cusparseMatDescr descrA, CUdeviceptr csrVal, CUdeviceptr csrRowPtr, CUdeviceptr csrColInd, int batchSize, csrqrInfo info, ref SizeT internalDataInBytes, ref SizeT workspaceInBytes)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of each matrix Aj.
System.Int32	n	number of columns of each matrix Aj.
System.Int32	nnz	number of nonzeros of each matrix Aj. It is the size csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrVal	array of nnzA*batchSize nonzero elements of matrices A0, A1, .... All matrices are aggregated one after another.
CUdeviceptr	csrRowPtr	integer array of m+1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColInd	integer array of nnzAcolumn indices of the nonzero elements of each matrix Aj.
System.Int32	batchSize	number of systems to be solved.
csrqrInfo	info	opaque structure for QR factorization.
SizeT	internalDataInBytes	number of bytes of the internal data.
SizeT	workspaceInBytes	number of bytes of the buffer in numerical factorization.

Returns

Type	Description
cusolverStatus

cusolverSpScsrqrsvBatched(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Int32, csrqrInfo, CUdeviceptr)

The batched sparse QR factorization is used to solve either a set of least-squares problems or a set of linear systems

Declaration

public static cusolverStatus cusolverSpScsrqrsvBatched(cusolverSpHandle handle, int m, int n, int nnz, cusparseMatDescr descrA, CUdeviceptr csrValA, CUdeviceptr csrRowPtrA, CUdeviceptr csrColIndA, CUdeviceptr b, CUdeviceptr x, int batchSize, csrqrInfo info, CUdeviceptr pBuffer)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of each matrix Aj.
System.Int32	n	number of columns of each matrix Aj.
System.Int32	nnz	number of nonzeros of each matrix Aj. It is the size csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrValA	array of nnzA*batchSize nonzero elements of matrices A0, A1, .... All matrices are aggregated one after another.
CUdeviceptr	csrRowPtrA	integer array of m+1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColIndA	integer array of nnzAcolumn indices of the nonzero elements of each matrix Aj.
CUdeviceptr	b	array of m*batchSize of right-hand-side vectors b0, b1, .... All vectors are aggregated one after another.
CUdeviceptr	x	array of m*batchSize of solution vectors x0, x1, .... All vectors are aggregated one after another.
System.Int32	batchSize	number of systems to be solved.
csrqrInfo	info	opaque structure for QR factorization.
CUdeviceptr	pBuffer	buffer allocated by the user, the size is returned by cusolverSpXcsrqrBufferInfoBatched().

Returns

Type	Description
cusolverStatus

cusolverSpSetStream(cusolverSpHandle, CUstream)

This function sets the stream to be used by the cuSolverSP library to execute its routines.

Declaration

public static cusolverStatus cusolverSpSetStream(cusolverSpHandle handle, CUstream streamId)

Parameters

Type	Name	Description
cusolverSpHandle	handle	the handle to the cuSolverSP context.
CUstream	streamId	the stream to be used by the library.

Returns

Type	Description
cusolverStatus

cusolverSpXcsrissymHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Int32[], Int32[], Int32[], ref Int32)

This function checks if A has symmetric pattern or not. The output parameter issym reports 1 if A is symmetric; otherwise, it reports 0.

Declaration

public static cusolverStatus cusolverSpXcsrissymHost(cusolverSpHandle handle, int m, int nnzA, cusparseMatDescr descrA, int[] csrRowPtrA, int[] csrEndPtrA, int[] csrColIndA, ref int issym)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnzA	number of nonzeros of matrix A. It is the size of csrValA and csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Int32[]	csrRowPtrA	integer array of m elements that contains the start of every row.
System.Int32[]	csrEndPtrA	integer array of m elements that contains the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnzAcolumn indices of the nonzero elements of matrix A.
System.Int32	issym	1 if A is symmetric; 0 otherwise.

Returns

Type	Description
cusolverStatus

cusolverSpXcsrperm_bufferSizeHost(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, Int32[], Int32[], Int32[], Int32[], ref SizeT)

Given a left permutation vector p which corresponds to permutation matrix P and a right permutation vector q which corresponds to permutation matrix Q, this function computes permutation of matrix A by B = PAQ^T

Declaration

public static cusolverStatus cusolverSpXcsrperm_bufferSizeHost(cusolverSpHandle handle, int m, int n, int nnzA, cusparseMatDescr descrA, int[] csrRowPtrA, int[] csrColIndA, int[] p, int[] q, ref SizeT bufferSizeInBytes)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of matrix A.
System.Int32	n	number of columns of matrix A.
System.Int32	nnzA	number of nonzeros of matrix A. It is the size of csrValA and csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Int32[]	p	left permutation vector of size m.
System.Int32[]	q	right permutation vector of size n.
SizeT	bufferSizeInBytes	number of bytes of the buffer.

Returns

Type	Description
cusolverStatus

cusolverSpXcsrpermHost(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, Int32[], Int32[], Int32[], Int32[], Int32[], Byte[])

Given a left permutation vector p which corresponds to permutation matrix P and a right permutation vector q which corresponds to permutation matrix Q, this function computes permutation of matrix A by B = PAQ^T

Declaration

public static cusolverStatus cusolverSpXcsrpermHost(cusolverSpHandle handle, int m, int n, int nnzA, cusparseMatDescr descrA, int[] csrRowPtrA, int[] csrColIndA, int[] p, int[] q, int[] map, byte[] pBuffer)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of matrix A.
System.Int32	n	number of columns of matrix A.
System.Int32	nnzA	number of nonzeros of matrix A. It is the size of csrValA and csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Int32[]	p	left permutation vector of size m.
System.Int32[]	q	right permutation vector of size n.
System.Int32[]	map	integer array of nnzA indices. If the user wants to get relationship between A and B, map must be set 0:1:(nnzA-1).
System.Byte[]	pBuffer	buffer allocated by the user, the size is returned by csrperm_bufferSize().

Returns

Type	Description
cusolverStatus

cusolverSpXcsrqrAnalysisBatched(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, csrqrInfo)

The batched sparse QR factorization is used to solve either a set of least-squares problems or a set of linear systems

Declaration

public static cusolverStatus cusolverSpXcsrqrAnalysisBatched(cusolverSpHandle handle, int m, int n, int nnzA, cusparseMatDescr descrA, CUdeviceptr csrRowPtrA, CUdeviceptr csrColIndA, csrqrInfo info)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of each matrix Aj.
System.Int32	n	number of columns of each matrix Aj.
System.Int32	nnzA	number of nonzeros of each matrix Aj. It is the size csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrRowPtrA	integer array of m+1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColIndA	integer array of nnzAcolumn indices of the nonzero elements of each matrix Aj.
csrqrInfo	info	opaque structure for QR factorization.

Returns

Type	Description
cusolverStatus

cusolverSpXcsrsymamdHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Int32[], Int32[], Int32[])

Symmetric Approximate minimum degree algorithm based on quotient graph.

This function implements Symmetric Approximate Minimum Degree Algorithm based on Quotient Graph. It returns a permutation vector p such that A(p,p) would have less zero fill-in during Cholesky factorization.

Declaration

public static cusolverStatus cusolverSpXcsrsymamdHost(cusolverSpHandle handle, int n, int nnzA, cusparseMatDescr descrA, int[] csrRowPtrA, int[] csrColIndA, int[] p)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	n	number of rows and columns of matrix A.
System.Int32	nnzA	number of nonzeros of matrix A. It is the size of csrValA and csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Int32[]	p	permutation vector of size n.

Returns

Type	Description
cusolverStatus

cusolverSpXcsrsymmdqHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Int32[], Int32[], Int32[])

Symmetric minimum degree algorithm based on quotient graph.

This function implements Symmetric Minimum Degree Algorithm based on Quotient Graph. It returns a permutation vector p such that A(p,p) would have less zero fill-in during Cholesky factorization.

Declaration

public static cusolverStatus cusolverSpXcsrsymmdqHost(cusolverSpHandle handle, int n, int nnzA, cusparseMatDescr descrA, int[] csrRowPtrA, int[] csrColIndA, int[] p)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	n	number of rows and columns of matrix A.
System.Int32	nnzA	number of nonzeros of matrix A. It is the size of csrValA and csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Int32[]	p	permutation vector of size n.

Returns

Type	Description
cusolverStatus

cusolverSpXcsrsymrcmHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, Int32[], Int32[], Int32[])

This function implements Symmetric Reverse Cuthill-McKee permutation. It returns a permutation vector p such that A(p,p) would concentrate nonzeros to diagonal. This is equivalent to symrcm in MATLAB, however the result may not be the same because of different heuristics in the pseudoperipheral finder.

Declaration

public static cusolverStatus cusolverSpXcsrsymrcmHost(cusolverSpHandle handle, int n, int nnzA, cusparseMatDescr descrA, int[] csrRowPtrA, int[] csrColIndA, int[] p)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	n	number of rows and columns of matrix A.
System.Int32	nnzA	number of nonzeros of matrix A. It is the size of csrValA and csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
System.Int32[]	p	permutation vector of size n.

Returns

Type	Description
cusolverStatus

cusolverSpZcsreigsHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, cuDoubleComplex[], Int32[], Int32[], cuDoubleComplex, cuDoubleComplex, ref Int32)

Declaration

public static cusolverStatus cusolverSpZcsreigsHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, cuDoubleComplex[] csrValA, int[] csrRowPtrA, int[] csrColIndA, cuDoubleComplex left_bottom_corner, cuDoubleComplex right_upper_corner, ref int num_eigs)

Parameters

Type	Name	Description
cusolverSpHandle	handle
System.Int32	m
System.Int32	nnz
cusparseMatDescr	descrA
cuDoubleComplex[]	csrValA
System.Int32[]	csrRowPtrA
System.Int32[]	csrColIndA
cuDoubleComplex	left_bottom_corner
cuDoubleComplex	right_upper_corner
System.Int32	num_eigs

Returns

Type	Description
cusolverStatus

cusolverSpZcsreigvsi(cusolverSpHandle, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, cuDoubleComplex, CUdeviceptr, Int32, Double, ref cuDoubleComplex, CUdeviceptr)

This function solves the simple eigenvalue problem Ax=lambdax by shift-inverse method.

Declaration

public static cusolverStatus cusolverSpZcsreigvsi(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, CUdeviceptr csrValA, CUdeviceptr csrRowPtrA, CUdeviceptr csrColIndA, cuDoubleComplex mu0, CUdeviceptr x0, int maxite, double tol, ref cuDoubleComplex mu, CUdeviceptr x)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
CUdeviceptr	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
cuDoubleComplex	mu0	initial guess of eigenvalue.
CUdeviceptr	x0	initial guess of eigenvector, a vecotr of size m.
System.Int32	maxite	maximum iterations in shift-inverse method.
System.Double	tol	tolerance for convergence.
cuDoubleComplex	mu	approximated eigenvalue nearest mu0 under tolerance.
CUdeviceptr	x	approximated eigenvector of size m.

Returns

Type	Description
cusolverStatus

cusolverSpZcsreigvsiHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, cuDoubleComplex[], Int32[], Int32[], cuDoubleComplex, cuDoubleComplex[], Int32, Double, ref cuDoubleComplex, cuDoubleComplex[])

This function solves the simple eigenvalue problem Ax=lambdax by shift-inverse method.

Declaration

public static cusolverStatus cusolverSpZcsreigvsiHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, cuDoubleComplex[] csrValA, int[] csrRowPtrA, int[] csrColIndA, cuDoubleComplex mu0, cuDoubleComplex[] x0, int maxite, double tol, ref cuDoubleComplex mu, cuDoubleComplex[] x)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
cuDoubleComplex[]	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
cuDoubleComplex	mu0	initial guess of eigenvalue.
cuDoubleComplex[]	x0	initial guess of eigenvector, a vecotr of size m.
System.Int32	maxite	maximum iterations in shift-inverse method.
System.Double	tol	tolerance for convergence.
cuDoubleComplex	mu	approximated eigenvalue nearest mu0 under tolerance.
cuDoubleComplex[]	x	approximated eigenvector of size m.

Returns

Type	Description
cusolverStatus

cusolverSpZcsrlsqvqrHost(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, cuDoubleComplex[], Int32[], Int32[], cuDoubleComplex[], Double, ref Int32, cuDoubleComplex[], Int32[], ref Double)

This function solves the following least-square problem x = argmin||A*z-b||

Declaration

public static cusolverStatus cusolverSpZcsrlsqvqrHost(cusolverSpHandle handle, int m, int n, int nnz, cusparseMatDescr descrA, cuDoubleComplex[] csrValA, int[] csrRowPtrA, int[] csrColIndA, cuDoubleComplex[] b, double tol, ref int rankA, cuDoubleComplex[] x, int[] p, ref double min_norm)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of matrix A.
System.Int32	n	number of columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
cuDoubleComplex[]	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
cuDoubleComplex[]	b	right hand side vector of size m.
System.Double	tol	tolerance to decide rank of A.
System.Int32	rankA	numerical rank of A.
cuDoubleComplex[]	x	solution vector of size n, x=pinv(A)*b.
System.Int32[]	p	a vector of size n, which represents the permuation matrix P satisfying AP^T=QR.
System.Double	min_norm	||A*x-b||, x=pinv(A)*b.

Returns

Type	Description
cusolverStatus

cusolverSpZcsrlsvchol(cusolverSpHandle, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Double, Int32, CUdeviceptr, ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpZcsrlsvchol(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, CUdeviceptr csrVal, CUdeviceptr csrRowPtr, CUdeviceptr csrColInd, CUdeviceptr b, double tol, int reorder, CUdeviceptr x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrVal	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
CUdeviceptr	csrRowPtr	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColInd	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
CUdeviceptr	b	right hand side vector of size m.
System.Double	tol	tolerance to decide singularity.
System.Int32	reorder	no effect.
CUdeviceptr	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is symmetric postive definite.

Returns

Type	Description
cusolverStatus

cusolverSpZcsrlsvcholHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, cuDoubleComplex[], Int32[], Int32[], cuDoubleComplex[], Double, Int32, cuDoubleComplex[], ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpZcsrlsvcholHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, cuDoubleComplex[] csrVal, int[] csrRowPtr, int[] csrColInd, cuDoubleComplex[] b, double tol, int reorder, cuDoubleComplex[] x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
cuDoubleComplex[]	csrVal	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtr	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColInd	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
cuDoubleComplex[]	b	right hand side vector of size m.
System.Double	tol	tolerance to decide singularity.
System.Int32	reorder	no effect.
cuDoubleComplex[]	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is symmetric postive definite.

Returns

Type	Description
cusolverStatus

cusolverSpZcsrlsvluHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, cuDoubleComplex[], Int32[], Int32[], cuDoubleComplex[], Double, Int32, cuDoubleComplex[], ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpZcsrlsvluHost(cusolverSpHandle handle, int n, int nnzA, cusparseMatDescr descrA, cuDoubleComplex[] csrValA, int[] csrRowPtrA, int[] csrColIndA, cuDoubleComplex[] b, double tol, int reorder, cuDoubleComplex[] x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	n	number of rows and columns of matrix A.
System.Int32	nnzA	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
cuDoubleComplex[]	csrValA	array of nnzA (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnzA (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
cuDoubleComplex[]	b	right hand side vector of size n.
System.Double	tol	tolerance to decide if singular or not.
System.Int32	reorder	no ordering if reorder=0. Otherwise, symrcm is used to reduce zero fill-in.
cuDoubleComplex[]	x	solution vector of size n, x = inv(A)*b.
System.Int32	singularity	-1 if A is invertible. Otherwise, first index j such that U(j,j)≈0

Returns

Type	Description
cusolverStatus

cusolverSpZcsrlsvqr(cusolverSpHandle, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Double, Int32, CUdeviceptr, ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpZcsrlsvqr(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, CUdeviceptr csrValA, CUdeviceptr csrRowPtrA, CUdeviceptr csrColIndA, CUdeviceptr b, double tol, int reorder, CUdeviceptr x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
CUdeviceptr	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
CUdeviceptr	b	right hand side vector of size m.
System.Double	tol	tolerance to decide if singular or not.
System.Int32	reorder	no effect.
CUdeviceptr	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is invertible. Otherwise, first index j such that R(j,j)≈0

Returns

Type	Description
cusolverStatus

cusolverSpZcsrlsvqrHost(cusolverSpHandle, Int32, Int32, cusparseMatDescr, cuDoubleComplex[], Int32[], Int32[], cuDoubleComplex[], Double, Int32, cuDoubleComplex[], ref Int32)

This function solves the linear system A*x=b

Declaration

public static cusolverStatus cusolverSpZcsrlsvqrHost(cusolverSpHandle handle, int m, int nnz, cusparseMatDescr descrA, cuDoubleComplex[] csrValA, int[] csrRowPtrA, int[] csrColIndA, cuDoubleComplex[] b, double tol, int reorder, cuDoubleComplex[] x, ref int singularity)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows and columns of matrix A.
System.Int32	nnz	number of nonzeros of matrix A.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
cuDoubleComplex[]	csrValA	array of nnz (= csrRowPtrA(n) * csrRowPtrA(0)) nonzero elements of matrix A.
System.Int32[]	csrRowPtrA	integer array of n + 1 elements that contains the start of every row and the end of the last row plus one.
System.Int32[]	csrColIndA	integer array of nnz (=csrRowPtrA(n) * csrRowPtrA(0)) column indices of the nonzero elements of matrix A.
cuDoubleComplex[]	b	right hand side vector of size m.
System.Double	tol	tolerance to decide if singular or not.
System.Int32	reorder	no effect.
cuDoubleComplex[]	x	solution vector of size m, x = inv(A)*b.
System.Int32	singularity	-1 if A is invertible. Otherwise, first index j such that R(j,j)≈0

Returns

Type	Description
cusolverStatus

cusolverSpZcsrqrBufferInfoBatched(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Int32, csrqrInfo, ref SizeT, ref SizeT)

The batched sparse QR factorization is used to solve either a set of least-squares problems or a set of linear systems

Declaration

public static cusolverStatus cusolverSpZcsrqrBufferInfoBatched(cusolverSpHandle handle, int m, int n, int nnz, cusparseMatDescr descrA, CUdeviceptr csrVal, CUdeviceptr csrRowPtr, CUdeviceptr csrColInd, int batchSize, csrqrInfo info, ref SizeT internalDataInBytes, ref SizeT workspaceInBytes)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of each matrix Aj.
System.Int32	n	number of columns of each matrix Aj.
System.Int32	nnz	number of nonzeros of each matrix Aj. It is the size csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrVal	array of nnzA*batchSize nonzero elements of matrices A0, A1, .... All matrices are aggregated one after another.
CUdeviceptr	csrRowPtr	integer array of m+1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColInd	integer array of nnzAcolumn indices of the nonzero elements of each matrix Aj.
System.Int32	batchSize	number of systems to be solved.
csrqrInfo	info	opaque structure for QR factorization.
SizeT	internalDataInBytes	number of bytes of the internal data.
SizeT	workspaceInBytes	number of bytes of the buffer in numerical factorization.

Returns

Type	Description
cusolverStatus

cusolverSpZcsrqrsvBatched(cusolverSpHandle, Int32, Int32, Int32, cusparseMatDescr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, CUdeviceptr, Int32, csrqrInfo, CUdeviceptr)

The batched sparse QR factorization is used to solve either a set of least-squares problems or a set of linear systems

Declaration

public static cusolverStatus cusolverSpZcsrqrsvBatched(cusolverSpHandle handle, int m, int n, int nnz, cusparseMatDescr descrA, CUdeviceptr csrValA, CUdeviceptr csrRowPtrA, CUdeviceptr csrColIndA, CUdeviceptr b, CUdeviceptr x, int batchSize, csrqrInfo info, CUdeviceptr pBuffer)

Parameters

Type	Name	Description
cusolverSpHandle	handle	handle to the cuSolverSP library context.
System.Int32	m	number of rows of each matrix Aj.
System.Int32	n	number of columns of each matrix Aj.
System.Int32	nnz	number of nonzeros of each matrix Aj. It is the size csrColIndA.
cusparseMatDescr	descrA	the descriptor of matrix A. The supported matrix type is CUSPARSE_MATRIX_TYPE_GENERAL. Also, the supported index bases are CUSPARSE_INDEX_BASE_ZERO and CUSPARSE_INDEX_BASE_ONE.
CUdeviceptr	csrValA	array of nnzA*batchSize nonzero elements of matrices A0, A1, .... All matrices are aggregated one after another.
CUdeviceptr	csrRowPtrA	integer array of m+1 elements that contains the start of every row and the end of the last row plus one.
CUdeviceptr	csrColIndA	integer array of nnzAcolumn indices of the nonzero elements of each matrix Aj.
CUdeviceptr	b	array of m*batchSize of right-hand-side vectors b0, b1, .... All vectors are aggregated one after another.
CUdeviceptr	x	array of m*batchSize of solution vectors x0, x1, .... All vectors are aggregated one after another.
System.Int32	batchSize	number of systems to be solved.
csrqrInfo	info	opaque structure for QR factorization.
CUdeviceptr	pBuffer	buffer allocated by the user, the size is returned by cusolverSpXcsrqrBufferInfoBatched().

Returns

Type	Description
cusolverStatus