-
Notifications
You must be signed in to change notification settings - Fork 71
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Adding Cardinal model of a 7-Pin LFR demonstration case #360
Conversation
Job VTB Documentation on 6d85ed0 wanted to post the following: View the site here This comment will be updated on new commits. |
Hi @GiudGiud @travismui, are we having significant degradation of SAM performance after recent update? The CI tests failed for 3 SAM models in the repository, and they are
The first SAM model timed out at step 57 while the Any help or suggestions are appreciated!. |
Did you invalidate and re run? The performance of the computing cluster for testing also varies |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
quick first pass
@GiudGiud Were the timeouts fixed by invalidating? I assumed it was just a slow job/node, our weekly VTB pipelines have been passing on our end. |
Yeah they were |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
let s use the new module name.
We'll have to discuss the transfer setup.
@hapfang Just in case you were not aware of this, I wanted to pass a long a GitHub tip. If you click on the "Files changed" tab, you can accept and then batch commit all suggested changes to a GitHub PR. This reduces the number of requests going to CIVET from a single PR and keeps the commit history tidy. |
Good to know. I was not aware of it. Thanks for sharing. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
let's add a renameBlockGenerator on the two meshes to have names on the subdomains on top of the ids
also the HC input cant be run on its own because the temperature seemingly blows up (thousands of K). I cant run the neutronics (no XS) but I imagine it's not the power density, it's the heat flux BCs that arent realisitically initialized?
Job Test griffin on f0a9725 : invalidated by @hapfang redo the ci tests to confirm errors |
Job Test blue_crab on f0a9725 : invalidated by @hapfang |
About temperature being blowing up, I could run the HC input without a problem. Since it uses the transient executioner while there's no time derivative and no changes in boundary condition, I just calculated three time steps and saw no changes in the solution after then. I see the initial condition for power density and boundary condition for the wall surface temperatures are set without a problem. Heat flux is the resulting quantity of the calculation, not the boundary condition. Can you elaborate more on your observation? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'll need to think of something to explain why we turn off the transfersanity checks. It's clearly because we move data from nodes to centroids, and from centroids to nodes, and that's generally not a good idea.
Alternatively, let's use the machinery you all had me develop last year. So ProjectionAux to move the data from nodes to centroids, then transfer
So, such transfer occurs for heat flux from heat conduction to NekRS. The heat flux computed by heat conduction is an elemental variable and the receiving variable in NekRS is a nodal variable. I assume you suggested the opposite way. So you meant using ProjectionAux to move the heat flux variable in heat conduction from centroids to nodes and then transfer? |
Yes I think the heat flux is affected, and the temperature as well but in the opposite direction. Ideally projecting before transferring will at least minimize the L2 norm of the difference between the fields, and will be reproducible in parallel unlike the NN approach |
Right. Thanks. Let me add your description in the md file. By the way, temperature transfer from NekRS to heat conduction is nodal-to-nodal transfer. So only the heat flux has been the problem. |
@GiudGiud I've addressed your comments except for the test part changing |
NONLINEAR should work. |
…pe to nodal type for transfer using ProjectionAux
…d being halved when being projected to nodal value
… running HC.i standalone
- add testing with more apps - add headers to simulation inputs Update lfr/7pin_cardinal_demo/runjob_sawtooth.sh Update lfr/7pin_cardinal_demo/runjob_sawtooth.sh Bring back delete fls file
Add gold files in the right folder use Newton for HC solve for performance
0da39b7
to
1f4617d
Compare
@hapfang do you know if this error means OOM? or does it mean we need to compile nek ? |
58a7c81
to
882558b
Compare
looks like it was OOM , went a little further with 16 procs |
bb17671
to
9a330c4
Compare
thanks for the contribution! |
Thank you for your effort in getting this merged!! |
This PR is related to issue #359.