Enable EQP for nonlinear elastic example

LLNL / libROM

Model reduction library with an emphasis on large scale parallelism and linear subspace methods

https://www.librom.net

Other

198 stars 36 forks source link

Enable EQP for nonlinear elastic example #218

Closed axla-io closed 6 months ago

axla-io commented 1 year ago

This PR adds EQP based hyperreduction to the nonlinear elastic example.

Current steps to take:

[x] Implement EQP hyperrreduction in main file
- [x] Sketch implementation following mixed_nonlinear_diffusion
- [x] Make sure CI tests run
- [x] Verify
[x] Implement SetupEQP_snapshots
- [x] Sketch
- [x] Verify implementation
[x] Implement HyperelasticNLFIntegrator_ComputeReducedEQP
- [x] Sketch
- [x] Verify implementation
[x] Fast reduced EQP
[x] Remember to do formatting

Future work

Preconditioning for NNLS
Subsampling

chldkdtn commented 11 months ago

@axla-io is this ready for review?

axla-io commented 11 months ago

@chldkdtn Not yet! I still have to refactor the code so that it maintains good practices for .cpp and .hpp files. I get a compilation error locally for my latest commit, although it seems to compile on GitHub actions. I'll try to do this change within the week.

axla-io commented 11 months ago

@chldkdtn I fixed the .hpp style issue and it should be ready for review soon. I want to make a commit cleaning up the repo a bit and then I'll request a review.

dylan-copeland commented 9 months ago

@axla-io to resolve the conflicts, please do git pull origin master.

axla-io commented 9 months ago

@axla-io to resolve the conflicts, please do git pull origin master.

@dylan-copeland Rebased now! Thanks for the the heads up!

axla-io commented 9 months ago

@dylan-copeland For some reason I can't make astyle 3.1 work on my mac. I formatted using astyle 3.4.10 but it seems to fail the checks... Would you be able to run astyle 3.1 on it to make it pass the checks?

axla-io commented 8 months ago

@dylan-copeland @chldkdtn @dreamer2368 @cval26 @mjkim1001 This PR still needs to be reviewed.

chldkdtn commented 7 months ago

@dylan-copeland has just informed me that EQP implementation for nonlinear elasticity has dependency to FOM scale, but it should be easy to remove the dependency. It would be ideal to remove that dependency in this PR so that we can use the optimized code for hyper-reduction comparison paper as well.

axla-io commented 7 months ago

@chldkdtn Yes, I saw @dylan-copeland s implementation in mixed_nonlinear_diffusion. Let's implement that in a second PR, since this is already a big one. I'm currently focusing on getting a Julia SciML PR in but I can take a look once that merges.

axla-io commented 7 months ago

I just merged the optimized EQP lifting from a sub-branch into this one, so now the PR is complete and ready to merge. @chldkdtn @dylan-copeland

mjkim1001 commented 6 months ago

I was asked to review the PR, and it looks good to me!