[Dune] Initialization of BlockVector with C array or Eigen array

Tobias Ritschel tobiasritschel at gmail.com
Wed May 27 10:29:58 CEST 2015


Alright, it sounds like it might be the best choice to use the ISTL
matrix/vector functionality for creating the Jacobian and right-hand-sides,
and as far as I can see there is plenty of functionality. It would be nice
to be able to use AMG.

@Christian Engwer: I am currently using the opm-autodiff
<http://www.opm-project.org/documentation/master/opm-autodiff/html/class_opm_1_1_auto_diff_block.php>
library
to obtain the Jacobian and residual equations so everything is currently in
the Eigen SparseMatrix and array format.

@Oliver Sander: I am writing my master thesis on high-permance computing in
optimization of oil production on the Technical University of Denmark. I
have a fully implicit two-phase solver which also uses the adjoint method
for computation of gradients.

Then I use TAO <http://www.mcs.anl.gov/research/projects/tao/> which is
distributed together with PETSc <http://www.mcs.anl.gov/petsc/> to carry
out single shooting optimization. I would like to use ISTL to be able to
solve the equations in parallel since we have an HPC cluster
<http://www.cc.dtu.dk> here on the university. Currently, I am using
BiCGStab from Eigen but it doesn't run in parallel and the other linear
solvers in Eigen are not supported.

At some point I would also like to use the Constrained Pressure Residual
preconditioner for the Newton iterations and adjoint equations (there is
some adjusted CPR preconditioner for this).

My deadline is in one month, however, so I may have to implement the ISTL
solvers at a later point.

Thanks for the help
- Tobias
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.dune-project.org/pipermail/dune/attachments/20150527/b1b189c2/attachment.htm>


More information about the Dune mailing list