nonsymmetric dynamics
ML 5.0 Smoothed Aggregation User's Guide - Trilinos - Sandia ...
ML 5.0 Smoothed Aggregation User's Guide - Trilinos - Sandia ...
- No tags were found...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
aggregation: aux: threshold [double]. The matrix entry A i,j is ignored if<br />
L i,j < threshold ∗ max j,i≠j L i,j . Note: the diagonal<br />
of A is also modified so that constants are<br />
still in the null space (assuming that this was<br />
true for the unmodified A). Default: 0.0.<br />
6.5 Default Parameter Settings for Common Problem Types<br />
The MultiLevelPreconditioner class provides default values for five different preconditioner<br />
types:<br />
1. Classical smoothed aggregation for symmetric positive definite or nearly symmetric<br />
positive definite systems.<br />
2. Classical smoothed aggregation-based 2-level domain decomposition<br />
3. 3-level algebraic domain decomposition<br />
4. Eddy current formulation of Maxwell’s equations<br />
5. Energy-based minimizing smoothed aggregation suitable for highly convective <strong>nonsymmetric</strong><br />
fluid flow problems.<br />
Default values are listed in Table 6. In the table, SA refers to “classical” smoothed aggregation<br />
(with small aggregates and relative large number of levels), DD and DD-ML to<br />
domain decomposition methods (whose coarse matrix is defined using aggressive coarsening<br />
and limited number of levels). Maxwell refers to the solution of Maxwell’s equations.<br />
NSSA is a <strong>nonsymmetric</strong> smoothed aggregation variant which may be appropriate for highly<br />
<strong>nonsymmetric</strong> operators.<br />
Default values for the parameter list can be set by ML Epetra::SetDefaults(). The<br />
user can easily put the desired default values in a given parameter list as follows:<br />
Teuchos::ParameterList MLList;<br />
ML_Epetra::SetDefaults(ProblemType, MLList);<br />
Teuchos::ParameterList MLList;<br />
ML_Epetra::SetDefaults(ProblemType, MLList, options, params);<br />
Teuchos::ParameterList MLList;<br />
ML_Epetra::SetDefaults(ProblemType, MLList, options, params, true);<br />
In the second usage, options and params are vectors that control the Aztec smoother. In<br />
the third usage, the boolean argument true permits SetDefaults to override any previously<br />
set parameters.<br />
For DD and DD-ML, the default smoother is Aztec, with an incomplete factorization ILUT,<br />
and minimal overlap. Memory for the two Aztec vectors, options and params, is allocated<br />
using Teuchos reference-counted pointers, and so will be freed automatically.<br />
34