[WIP] Multizone discrete adjoints for turbomachinery #2317
Conversation
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
|
|
||
| su2double gammaIn_BC = 0; | ||
| if (nDim == 3){ | ||
| gammaIn_BC = FlowDir[2]; //atan(FlowDir[2]/FlowDir[0]); |
Check notice
Code scanning / CodeQL
Commented-out code
|
@joshkellyjak @pcarruscag Should I make some first commits for a "tag type / debug" recording to this branch? (This problem is the main motivation but technically it's quite unrelated.) We need it here, but it might be better to merge it from a dedicated branch. |
…l write restart files during runtime (there is a bug, potentially when clearing the tape).
Personally I would like this to selfishly enhance my understanding, but I don't think it should be merged when we complete the PR. Could be useful documentation for the website. |
Well I think the functionality should be merged at some point. It might become too much for this PR(?). Let me just go ahead and create a dedicated branch for the tag tape. We could just use that branch as a helper then and prepare it for develop after this one is merged. |
|
Ah now I got the comment. I would keep adding the debug tape separate from fixing the turbo issues |
…zation.hpp (to cover RealReverseTag).
|
Reverting an accidental merge |
…his is a temporary change to avoid false positives when a tag tape is used. They might be unnecessary anyway though.
Proposed Changes
This is, for now, a brief note on @joshkellyjak 's and my WIP on a discrete adjoint functionality for multizone (multi-stage, multi-row) turbomachinery cases. We intend to realize everything through the native multizone discrete adjoint solver, though some turbo-specific problems need to be resolved.
This will, probably, involve further additions/corrections/cleanups around the mixing plane interface so that it is recorded properly. The current recording seems to still be not fully correct.
Related Work
None, so far.