- 23 Jun, 2022 3 commits
-
-
oliver.sander_at_tu-dresden.de authored
CholmodSolver: Allow to reuse a factorization See merge request !67
-
Computing Cholesky factorizations is expensive, and they absolutely have to be reused if the same linear system is solved with several different right hand sides. The previous implementation of CholmodSolver wouldn't allow this: The 'solve' method factorized and solved together. This patch makes the interface of the CholmodSolver class richer to allow for factorization reuse. It is now possible to set the matrix, rhs and solution storage separately. The matrix can be factorized by calling a new 'factorize' method, and if 'solve' is called afterwards, the matrix won't be factorized again. On the other hand, when calling 'solve' by itself the 'solve' method will do the factorization. In other words, the patch is fully backward-compatible.
-
-
- 27 Jan, 2022 1 commit
-
-
Patrick Jaap authored
-
- 25 Jan, 2022 2 commits
-
-
oliver.sander_at_tu-dresden.de authored
Replace DUNE_DEPRECATED by [[deprecated]] See merge request !66
-
lisa_julia.nebel_at_tu-dresden.de authored
-
- 06 Jan, 2022 1 commit
-
-
Patrick Jaap authored
TupleVector: inherit all constructors from std::tuple See merge request !65
-
- 04 Jan, 2022 1 commit
-
-
Patrick Jaap authored
The previous implementation caused unwanted passes of std::is_assignable with bool. Furthermore, the previous implementation provided a variadic constructor without any restrictions causing hard to track compiler errors in the std::forward part.
-
- 20 Dec, 2021 2 commits
-
-
oliver.sander_at_tu-dresden.de authored
Replace HAVE_UG by HAVE_DUNE_UGGRID See merge request !64
-
oliver.sander_at_tu-dresden.de authored
Because HAVE_UG has been removed.
-
- 16 Nov, 2021 1 commit
-
-
Patrick Jaap authored
EnergyNorm: add getMatrix() method See merge request !62
-
- 15 Nov, 2021 1 commit
-
-
Patrick Jaap authored
-
- 09 Nov, 2021 2 commits
-
-
oliver.sander_at_tu-dresden.de authored
cmake: use keyword signature of target_link_libraries See merge request !61
-
lisa_julia.nebel_at_tu-dresden.de authored
-
- 26 Oct, 2021 3 commits
-
-
Patrick Jaap authored
Fix: LoopSolver has to execute the criteria regardless of the verbosity mode See merge request !59
-
Patrick Jaap authored
Fix return type in IterativeSolver See merge request !60
-
Patrick Jaap authored
-
- 21 Oct, 2021 1 commit
-
-
Patrick Jaap authored
In the current implementation this caused different solution behavior depending on the chosen verbosity mode. Plain stop critera were ignored in the SOLVER::QUIET case. This patch fixed the problem.
-
- 25 Aug, 2021 1 commit
-
-
Patrick Jaap authored
LoopSolver: Add getter for the current convergence rate See merge request !56
-
- 20 Aug, 2021 1 commit
-
-
Patrick Jaap authored
-
- 06 Jul, 2021 1 commit
-
-
oliver.sander_at_tu-dresden.de authored
Make TwoNorm more generic See merge request !55
-
- 05 Jul, 2021 2 commits
-
-
oliver.sander_at_tu-dresden.de authored
Adapt CholmodSolver to (almost) arbitrary matrix types See merge request !53
-
Patrick Jaap authored
This is a follow-up change from the dune-istl Cholmod changes in https://gitlab.dune-project.org/core/dune-istl/-/merge_requests/424
-
- 02 Jul, 2021 1 commit
-
-
Patrick Jaap authored
Now it also works for multi type vectors
-
- 30 Jun, 2021 2 commits
-
-
oliver.sander_at_tu-dresden.de authored
Replace DUNE_UNUSED by [[maybe_unused]] See merge request !54
-
Patrick Jaap authored
-
- 26 Nov, 2020 2 commits
-
-
oliver.sander_at_tu-dresden.de authored
Add missing namespace fixes for addToDiagonal See merge request !52
-
oliver.sander_at_tu-dresden.de authored
Apparently the calls are not used anywhere in dune-solvers itself, that's why the CI system didn't catch the problem.
-
- 23 Nov, 2020 1 commit
-
-
oliver.sander_at_tu-dresden.de authored
UMFPackSolver with scalar matrix entries See merge request !51
-
- 22 Nov, 2020 3 commits
-
-
oliver.sander_at_tu-dresden.de authored
-
oliver.sander_at_tu-dresden.de authored
-
oliver.sander_at_tu-dresden.de authored
Previously, the class SymmetricSampleProblem would always produce matrices and vectors with FieldMatrix/FieldVector entries of a given user-controllable blocksize. With this patch, the blocksize can also be set to 0. In that case, the resulting matrices and vectors have 'double' entries.
-
- 19 Nov, 2020 1 commit
-
-
oliver.sander_at_tu-dresden.de authored
In that case the BitVector will be the input vector type, but with entry type 'char'.
-
- 21 Oct, 2020 1 commit
-
-
Patrick Jaap authored
Interval: handle negative multiplication See merge request !50
-
- 09 Oct, 2020 1 commit
-
-
Patrick Jaap authored
-
- 30 Sep, 2020 2 commits
-
-
oliver.sander_at_tu-dresden.de authored
Interval: scalar multiplication See merge request !49
-
Patrick Jaap authored
-
- 05 Sep, 2020 1 commit
-
-
oliver.sander_at_tu-dresden.de authored
Fix bug in MultigridStep constructor See merge request !45
-
- 04 Sep, 2020 2 commits
-
-
oliver.sander_at_tu-dresden.de authored
Remove deprecated stuff from tests See merge request !46
-
oliver.sander_at_tu-dresden.de authored
-