|
Belos Version of the Day
|
Default parameters common to most Belos solvers. More...
#include <BelosTypes.hpp>
Static Public Attributes | |
| static const double | convTol = 1.0e-8 |
| Default convergence tolerance. | |
| static const double | polyTol = 1.0e-12 |
| Relative residual tolerance for matrix polynomial construction. | |
| static const double | orthoKappa = -1.0 |
| DGKS orthogonalization constant. | |
| static const double | resScaleFactor = 1.0 |
| User-defined residual scaling factor. | |
| static const double | impTolScale = 10.0 |
| "Implicit Tolerance Scale Factor" | |
Default parameters common to most Belos solvers.
Both Belos solvers and users may override these defaults. Real floating-point values are deliberately double, in order to avoid issues with constexpr construction of certain MagnitudeTypes.
Definition at line 251 of file BelosTypes.hpp.
Default convergence tolerance.
This assumes that implicit conversion from double to Teuchos::ScalarTraits<Scalar>::magnitudeType always works, but Belos already assumed that. See discussion starting here:
https://github.com/trilinos/Trilinos/pull/2677#issuecomment-395453521
Initialization of values moved to BelosTypes.cpp to fix ODR-used issues
Definition at line 261 of file BelosTypes.hpp.
Relative residual tolerance for matrix polynomial construction.
Definition at line 264 of file BelosTypes.hpp.
DGKS orthogonalization constant.
Definition at line 267 of file BelosTypes.hpp.
User-defined residual scaling factor.
Definition at line 270 of file BelosTypes.hpp.
"Implicit Tolerance Scale Factor"
Definition at line 273 of file BelosTypes.hpp.
1.9.8