MAGMA  1.5.0
Matrix Algebra for GPU and Multicore Architectures
 All Functions Groups
double precision

Functions

magma_int_t magma_dbaiter (magma_d_sparse_matrix A, magma_d_vector b, magma_d_vector *x, magma_d_solver_par *solver_par)
 Solves a system of linear equations A * x = b via the block asynchronous iteration method on GPU. More...
 
magma_int_t magma_dbcsrlu (magma_d_sparse_matrix A, magma_d_vector b, magma_d_vector *x, magma_d_solver_par *solver_par)
 LU decomposition and solution via triangular solves for a BCSR matrix A. More...
 
magma_int_t magma_dbcsrlutrf (magma_d_sparse_matrix A, magma_d_sparse_matrix *M, magma_int_t *ipiv, magma_int_t version)
 LU decomposition for a BCSR matrix A. More...
 
magma_int_t magma_dbcsrlusv (magma_d_sparse_matrix A, magma_d_vector b, magma_d_vector *x, magma_d_solver_par *solver_par, magma_int_t *ipiv)
 LU solve for a BCSR matrix A. More...
 
magma_int_t magma_dbicgstab (magma_d_sparse_matrix A, magma_d_vector b, magma_d_vector *x, magma_d_solver_par *solver_par)
 Solves a system of linear equations A * X = B where A is a real symmetric N-by-N positive definite matrix A. More...
 
magma_int_t magma_dbicgstab_merge (magma_d_sparse_matrix A, magma_d_vector b, magma_d_vector *x, magma_d_solver_par *solver_par)
 Solves a system of linear equations A * X = B where A is a real symmetric N-by-N positive definite matrix A. More...
 
magma_int_t magma_dbicgstab_merge2 (magma_d_sparse_matrix A, magma_d_vector b, magma_d_vector *x, magma_d_solver_par *solver_par)
 Solves a system of linear equations A * X = B where A is a real symmetric N-by-N positive definite matrix A. More...
 
magma_int_t magma_dgmres (magma_d_sparse_matrix A, magma_d_vector b, magma_d_vector *x, magma_d_solver_par *solver_par)
 Solves a system of linear equations A * X = B where A is a real sparse matrix stored in the GPU memory. More...
 
magma_int_t magma_diterref (magma_d_sparse_matrix A, magma_d_vector b, magma_d_vector *x, magma_d_solver_par *solver_par, magma_d_preconditioner *precond_par)
 Solves a system of linear equations A * X = B where A is a real symmetric N-by-N positive definite matrix A. More...
 
magma_int_t magma_djacobi (magma_d_sparse_matrix A, magma_d_vector b, magma_d_vector *x, magma_d_solver_par *solver_par)
 Solves a system of linear equations A * X = B where A is a real symmetric N-by-N positive definite matrix A. More...
 
magma_int_t magma_dpastixsetup (magma_d_sparse_matrix A, magma_d_vector b, magma_d_preconditioner *precond)
 Prepares the PASTIX solver. More...
 
magma_int_t magma_dpgmres (magma_d_sparse_matrix A, magma_d_vector b, magma_d_vector *x, magma_d_solver_par *solver_par, magma_d_preconditioner *precond_par)
 Solves a system of linear equations A * X = B where A is a real sparse matrix stored in the GPU memory. More...
 

Detailed Description

Function Documentation

magma_int_t magma_dbaiter ( magma_d_sparse_matrix  A,
magma_d_vector  b,
magma_d_vector *  x,
magma_d_solver_par *  solver_par 
)

Solves a system of linear equations A * x = b via the block asynchronous iteration method on GPU.

Parameters
Amagma_d_sparse_matrix input matrix A
bmagma_d_vector RHS b
xmagma_d_vector* solution approximation
solver_parmagma_d_solver_par* solver parameters
magma_int_t magma_dbcsrlu ( magma_d_sparse_matrix  A,
magma_d_vector  b,
magma_d_vector *  x,
magma_d_solver_par *  solver_par 
)

LU decomposition and solution via triangular solves for a BCSR matrix A.

We assume all diagonal blocks to be nonzero.

Parameters
Amagma_d_sparse_matrix descriptor for matrix A
bmagma_d_vector RHS b vector
xmagma_d_vector* solution approximation
solver_parmagma_d_solver_par* solver parameters
magma_int_t magma_dbcsrlusv ( magma_d_sparse_matrix  A,
magma_d_vector  b,
magma_d_vector *  x,
magma_d_solver_par *  solver_par,
magma_int_t *  ipiv 
)

LU solve for a BCSR matrix A.

Parameters
Amagma_d_sparse_matrix input matrix A (on DEV)
bmagma_d_vector RHS (on DEV)
xmagma_d_vector solution (on DEV)
solver_parmagma_d_solver_par* solver parameters
ipivmagma_int_t* pivot vector
magma_int_t magma_dbcsrlutrf ( magma_d_sparse_matrix  A,
magma_d_sparse_matrix *  M,
magma_int_t *  ipiv,
magma_int_t  version 
)

LU decomposition for a BCSR matrix A.

We assume all diagonal blocks to be nonzero.

Parameters
Amagma_d_sparse_matrix input matrix A (on DEV)
mmagma_d_sparse_matrix* output matrix containing LU decomposition
ipivmagma_int_t* pivot vector
magma_int_t magma_dbicgstab ( magma_d_sparse_matrix  A,
magma_d_vector  b,
magma_d_vector *  x,
magma_d_solver_par *  solver_par 
)

Solves a system of linear equations A * X = B where A is a real symmetric N-by-N positive definite matrix A.

This is a GPU implementation of the Biconjugate Gradient Stabelized method.

Parameters
Amagma_d_sparse_matrix input matrix A
bmagma_d_vector RHS b
xmagma_d_vector* solution approximation
solver_parmagma_d_solver_par* solver parameters
magma_int_t magma_dbicgstab_merge ( magma_d_sparse_matrix  A,
magma_d_vector  b,
magma_d_vector *  x,
magma_d_solver_par *  solver_par 
)

Solves a system of linear equations A * X = B where A is a real symmetric N-by-N positive definite matrix A.

This is a GPU implementation of the Biconjugate Gradient Stabelized method. The difference to magma_dbicgstab is that we use specifically designed kernels merging multiple operations into one kernel.

Parameters
Amagma_d_sparse_matrix input matrix A
bmagma_d_vector RHS b
xmagma_d_vector* solution approximation
solver_parmagma_d_solver_par* solver parameters
magma_int_t magma_dbicgstab_merge2 ( magma_d_sparse_matrix  A,
magma_d_vector  b,
magma_d_vector *  x,
magma_d_solver_par *  solver_par 
)

Solves a system of linear equations A * X = B where A is a real symmetric N-by-N positive definite matrix A.

This is a GPU implementation of the Biconjugate Gradient Stabelized method. The difference to magma_dbicgstab is that we use specifically designed kernels merging multiple operations into one kernel.

Parameters
Amagma_d_sparse_matrix input matrix A
bmagma_d_vector RHS b
xmagma_d_vector* solution approximation
solver_parmagma_d_solver_par* solver parameters
magma_int_t magma_dgmres ( magma_d_sparse_matrix  A,
magma_d_vector  b,
magma_d_vector *  x,
magma_d_solver_par *  solver_par 
)

Solves a system of linear equations A * X = B where A is a real sparse matrix stored in the GPU memory.

X and B are real vectors stored on the GPU memory. This is a GPU implementation of the right-preconditioned GMRES method.

Parameters
Amagma_d_sparse_matrix descriptor for matrix A
bmagma_d_vector RHS b vector
xmagma_d_vector* solution approximation
solver_parmagma_d_solver_par* solver parameters
magma_int_t magma_diterref ( magma_d_sparse_matrix  A,
magma_d_vector  b,
magma_d_vector *  x,
magma_d_solver_par *  solver_par,
magma_d_preconditioner *  precond_par 
)

Solves a system of linear equations A * X = B where A is a real symmetric N-by-N positive definite matrix A.

This is a GPU implementation of the Iterative Refinement method. The inner solver is passed via the preconditioner argument.

Parameters
Amagma_d_sparse_matrix input matrix A
bmagma_d_vector RHS b
xmagma_d_vector* solution approximation
solver_parmagma_d_solver_par* solver parameters
precond_parmagma_d_preconditioner* inner solver
magma_int_t magma_djacobi ( magma_d_sparse_matrix  A,
magma_d_vector  b,
magma_d_vector *  x,
magma_d_solver_par *  solver_par 
)

Solves a system of linear equations A * X = B where A is a real symmetric N-by-N positive definite matrix A.

This is a GPU implementation of the Jacobi method.

Parameters
Amagma_d_sparse_matrix input matrix A
bmagma_d_vector RHS b
xmagma_d_vector* solution approximation
solver_parmagma_d_solver_par* solver parameters
magma_int_t magma_dpastixsetup ( magma_d_sparse_matrix  A,
magma_d_vector  b,
magma_d_preconditioner *  precond 
)

Prepares the PASTIX solver.

Parameters
Amagma_d_sparse_matrix input matrix A
bmagma_d_vector RHS b
precondmagma_d_preconditioner* preconditioner parameter
magma_int_t magma_dpgmres ( magma_d_sparse_matrix  A,
magma_d_vector  b,
magma_d_vector *  x,
magma_d_solver_par *  solver_par,
magma_d_preconditioner *  precond_par 
)

Solves a system of linear equations A * X = B where A is a real sparse matrix stored in the GPU memory.

X and B are real vectors stored on the GPU memory. This is a GPU implementation of the right-preconditioned GMRES method.

Parameters
Amagma_d_sparse_matrix descriptor for matrix A
bmagma_d_vector RHS b vector
xmagma_d_vector* solution approximation
solver_parmagma_d_solver_par* solver parameters
precond_parmagma_d_preconditioner* preconditioner