this feature is necessary to keep readable output if we use funny value types, like Vc::double_v

This patch removes the assumption that the field_type and / or real_type
can always be implicitly converted to double and can be implicitly
constructed from an int (in constructs like field_type d = 0;).

This patch is required for compatibility with the FLOP counter in PDELab.

[Bugfix] fixed order of the scalar product when applying the Arnoldi algorithm in GMRes solver, which is important in the complex case

minor cleanups in the MinRes solver

[Bugfix] implemented complex Givens rotations and fixed MinRes and GMRes for complex valued problems

Function attribute have to be placed in front of the function name,
not behind it.

Marian Piatkowski authored 10 years ago and Steffen Müthing committed 10 years ago

made the distinction between field_type and real_type which is the same for real numbers but differs for std::complex

Marian Piatkowski authored 10 years ago and Steffen Müthing committed 10 years ago

added the old constructors called with the variable recalc_defect to RestartedGMResSolver and set them as deprecated, since they should not be used anymore

Marian Piatkowski authored 11 years ago and Steffen Müthing committed 10 years ago

replaced the usual arrays for the Givens-rotation coefficients and the search directions etc. with the arrays from Dune

implemented the BLAS-routine drotg for greater robustness which was a TODO note to compute the Givens rotation

Marian Piatkowski authored 11 years ago and Steffen Müthing committed 10 years ago

removed bool recalc_defect
reimplement the update function with respect to optimality
fixed strange behaviour when maxiter is reached and GMRes wants to restart infinitely many times
fixed strange behaviour when linear reduction is reached but GMRes wants to restart and continue calculation
fixed the printing of the results

Fixes flyspray https://dune-project.org/flyspray/index.php?do=details&task_id=1393
(not stopping if there is no convergence).
In addition each residual is only printed once now.

The full krylov basis is copied in the call of the method update(),
working with a reference to the krylov basis reduced the peak amount of allocated memory
by a factor of about 1.7 and improves the runtime by 4%.

Steffen: Also constified all applicable function arguments as suggested
         by Christian.

Closes FS#1378

Jurgis Pods authored 11 years ago and Markus Blatt committed 11 years ago

This is essentially equivalent to Uli Sack's patch from Flyspray task FS#1202, ported to the current master.

Currently, GMRes is inconsistent when comparing current ('norm') and initial defect ('norm_0'). While the current residual is taken from the preconditioned residual, the initial defect - at least in the case the flag '_recalc_defect' is not set, which is the default - is taken from the unpreconditioned residual.

This causes the following problems:
1) The calculated reduction is wrong.
2) The solver might even exit in iteration 0 (without entering the iteration loop) because 'norm' and 'norm_0' differ by a factor greater than the requested reduction, returning a zero update 'x'.

By using the preconditioned residual in any case, the defect calculation is made consistent again.

The last commits caused compile errors for e.g. clang as I used
incomplete (only forward declared) types in template classes.
This patch moves the definition of InverseOperator and Preconditioners
from the headers solvers.hh and preconditioners into new separate
headers solver.hh and preconditioner.hh. This allows to include these
definitions in the previous headers without cause circular include directives.

Previously if the solver did not converge the number of iterations
that was printed and stored was one higher than the actual value.

Four warnings are left for two variables - I don't know what they do.

[[Imported from SVN: r1908]]

Include some other header that are used, to make headercheck
pass again.

[[Imported from SVN: r1905]]

[[Imported from SVN: r1776]]

[[Imported from SVN: r1749]]

[[Imported from SVN: r1748]]

Shouldn't subversion prevent such a thing? I even could
not use svn merge to restall the changes of r1739. 
Really weired.

[[Imported from SVN: r1747]]

[[Imported from SVN: r1746]]

[[Imported from SVN: r1743]]

This solver allows the preconditioner
to between iterations or even be nonlinear.

[[Imported from SVN: r1740]]

[[Imported from SVN: r1739]]

Removed superfluous Preconditioner.apply in GMRes, see FS#1201.

(thanks to Uli Sack for the patch)

[[Imported from SVN: r1729]]

- fix some real_type vs field_type issues
- make more solvers work with complex
(thanks to Matthias Wohlmuth)

[[Imported from SVN: r1650]]

[[Imported from SVN: r1628]]