Test case that shows problem with Nalu-Tioga link #69
Conversation
| linear_solvers: | ||
|
|
||
| - name: solve_scalar | ||
| type: tpetra |
There was a problem hiding this comment.
Can we switch these to hypre? These tpetra solver settings are super old and we've not been keeping up with them. I would rather not see another copy of milestone.xml put under version control if we can avoid it.
|
Thanks for the input, @psakievich. I wanted a quick way to share it with Ilker, so I put it all here at first. I'll make the changes you requested. |
|
Sounds good. My requests are only if you want to merge it for a long term test. Otherwise, fine as is. |
|
It probably would be useful to be added as a long-term part of the repository, but it doesn't fit well as a true regression test because it's just a one time step thing without a real physical meaning. I'll think about it some more. |
|
I don't think a regression test has to always be physically meaningful. One timestep to verify hole-cutting in parallel hasn't regressed seems like a perfect regression test candidate IMO. |
|
Solved with nalu-wind fix related to "set_resolutions" Exawind/nalu-wind#1267 |
In Nalu-Wind, the velocity is set to 0.0 and the density is set to 1.0. In AMR-Wind, the velocity is set to [0.1, -0.1, 0.0] and the density is set to 1.5. This shows that when multiple processors are used for Nalu-Wind, there are errors in communicating the fields properly into Nalu-Wind. The iblank field (its boundary shown in black), does not have issues, though, thanks to the recent changes to Nalu-Wind.