e3115846f5
To correct the previous commit, the BlockSolverPerformance has been enhanced
by a replace(...) and max(...) memeber function.
The matrix solve functions of fvMatrix, faMatrix and tetFemMatrix now store
solver performance data for all components in the solverPerformance dictionary
of solution. However, they will still return a scalar-valued
lduSolverPerformance based on the component with the max residual.
This is necessary as otherwise the type of a non-scalar field, which may be
specified in fvSolution::residualControl does not match the type which is
written to and read from the solverPerformance dictionary in the
maxTypeResidual(...) from solutionControl. If e.g. the field U is given
in fvSolution::residualControl of solutionControl, the constructor of
BlockSolverPerformance in the maxTypeResidual(...) from solutionControl
expects to read a vector ( u1 u2 u3 ) from the dictionary and not just one
component.
The changes are inspired and directly related to the following vanilla
commit:
|
||
|---|---|---|
| .. | ||
| blockLduSystem | ||
| cfdTools | ||
| fields | ||
| finiteVolume | ||
| fvMatrices | ||
| fvMesh | ||
| interpolation | ||
| Make | ||
| surfaceMesh | ||
| volMesh | ||