|
schwz
Generated automatically from develop
|
The struct that contains the solver settings and the parameters to be set by the user. More...
#include <settings.hpp>
Classes | |
| struct | comm_settings |
| The settings for the various available communication paradigms. More... | |
| struct | convergence_settings |
| The various convergence settings available. More... | |
Public Types | |
| enum | partition_settings { partition_regular = 0x0, partition_regular2d = 0x4, partition_metis = 0x1, partition_zoltan = 0x2, partition_custom = 0x3 } |
| The partition algorithm to be used for partitioning the matrix. | |
| enum | local_solver_settings { direct_solver_cholmod = 0x0, direct_solver_umfpack = 0x5, direct_solver_ginkgo = 0x1, iterative_solver_ginkgo = 0x2, iterative_solver_dealii = 0x3, solver_custom = 0x4 } |
| The local solver algorithm for the local subdomain solves. | |
Public Member Functions | |
| Settings (std::string executor_string="reference") | |
Public Attributes | |
| std::string | executor_string |
| The string that contains the ginkgo executor paradigm. | |
| std::shared_ptr< gko::Executor > | executor = gko::ReferenceExecutor::create() |
| The ginkgo executor the code is to be executed on. | |
| std::shared_ptr< device_guard > | cuda_device_guard |
| The ginkgo executor the code is to be executed on. | |
| partition_settings | partition = partition_settings::partition_regular |
| gko::int32 | overlap = 2 |
| The overlap between the subdomains. | |
| std::string | matrix_filename = "null" |
| The string that contains the matrix file name to read from . | |
| bool | explicit_laplacian = true |
| Flag if the laplacian matrix should be generated within the library. More... | |
| bool | use_mixed_precision = false |
| Flag if mixed precision should be used. | |
| bool | enable_random_rhs = false |
| Flag to enable a random rhs. | |
| bool | print_matrices = false |
| Flag to enable printing of matrices. | |
| bool | debug_print = false |
| Flag to enable some debug printing. | |
| local_solver_settings | local_solver |
| bool | non_symmetric_matrix = false |
| Is the matrix non-symmetric ? , Use GMRES for local solves. | |
| unsigned int | restart_iter = 1u |
| The restart iter for the GMRES solver. | |
| int | reset_local_crit_iter = -1 |
| The global iter at which to reset the local solver criterion. | |
| bool | naturally_ordered_factor = false |
| Disables the re-ordering of the matrix before computing the triangular factors during the CHOLMOD factorization. More... | |
| std::string | metis_objtype |
| This setting defines the objective type for the metis partitioning. | |
| bool | use_precond = false |
| Enable the block jacobi local preconditioner for the local solver. | |
| bool | write_debug_out = false |
| Enable the writing of debug out to file. | |
| bool | write_iters_and_residuals = false |
| Enable writing the iters and residuals to a file. | |
| bool | enable_logging = false |
| Flag to enable logging for local iterative solvers. More... | |
| bool | write_perm_data = false |
| Enable the local permutations from CHOLMOD to a file. | |
| int | shifted_iter = 1 |
| Iteration shift for node local communication. | |
| comm_settings | comm_settings |
| convergence_settings | convergence_settings |
| std::string | factorization = "cholmod" |
| The factorization for the local direct solver. | |
| std::string | reorder |
| The reordering for the local solve. | |
The struct that contains the solver settings and the parameters to be set by the user.
settings
| bool schwz::Settings::enable_logging = false |
Flag to enable logging for local iterative solvers.
Note: Probably will have a significant performance hit.
| bool schwz::Settings::explicit_laplacian = true |
Flag if the laplacian matrix should be generated within the library.
If false, an external matrix and rhs needs to be provided
Referenced by schwz::SchwarzBase< ValueType, IndexType, MixedValueType >::initialize().
| local_solver_settings schwz::Settings::local_solver |
| bool schwz::Settings::naturally_ordered_factor = false |
Disables the re-ordering of the matrix before computing the triangular factors during the CHOLMOD factorization.
1.8.13