Hello Winnifried, > I noticed that apparently a change in the SolutionTransfer class is coming with the next version. Particularly, the function 'prepare_for_pure_refinement' seems to be removed (was present in version 9.6.2 but is gone in git)?
We had separate implementations of `SolutionTransfer` each for a different Triangulation class. With the coming release, we decided to merge them into one, which helps us to better maintain the functionality, but also means that a common interface had to be found. Unfortunately `prepare_for_pure_refinement` did not make it. See also https://github.com/dealii/dealii/pull/15706 If you need the old implementation, you can use the `Legacy::SolutionTransfer` class. However, it is deprecated and will be removed in the future. https://dealii.org/developer/doxygen/deal.II/classLegacy_1_1SolutionTransfer.html > Now with 'prepare_for_coarsening_and_refinement' it seems to me that I have to load all timesteps to memory as the vectors need to be passed to the function? Is there a way to avoid this? You were using a SolutionTransfer object and set it up once with `prepare_for_pure_refinement`, and loaded one vector after the other from disk into memory. The mesh doesn't change, so you are not using mesh refinement. The serial SolutionTransfer class just had the responsibility to transfer vectors between differently refined meshes. How did you use it to load vectors from disk? I don't remember that it had such a functionality. If the mesh doesn't change, and you only store and load the vectors during the runtime of the program, then I would suggest to serialize them with boost. We have a new tutorial on serialization. Please have a look at step-83 for more information. https://dealii.org/developer/doxygen/deal.II/step_83.html Let me know if that helps. Best, Marc On Thursday, January 23, 2025 at 12:16:51 PM UTC+1 winnifried wrote: > Dear all, > I noticed that apparently a change in the SolutionTransfer class is > coming with the next version. > Particularly, the function 'prepare_for_pure_refinement' seems to be > removed (was present in version 9.6.2 but is gone in git)? > My code presently relies on the possibility to utilize this mesh > transfer as it can be prepared without prior knowledge of the vectors to > be interpolated. Is there a fundamental reason why this function is > removed - if so do you have a suggestion for a resolution of the > following problem: > > In a time dependent problem (on a single mesh for all timesteps), I have > the possibilities to store finite element vectors associated to certain > timesteps on disk - to avoid keeping all vectors in RAM. > Now with 'prepare_for_coarsening_and_refinement' it seems to me that I > have to load all timesteps to memory as the vectors need to be passed to > the function? Is there a way to avoid this? > > Thanks > Winni > > > -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion visit https://groups.google.com/d/msgid/dealii/99068493-f3c9-46f1-a29c-4f5d1cfbcc87n%40googlegroups.com.
