EQP #173
EQP #173
Conversation
…umbers of snapshots (resulting from conditions for the first snapshot). Eliminating RHS basis in EQP case. Refactoring the EQP setup.
|
Is it ready for review and merge? |
@chldkdtn it is working in some cases, but I am planning to test more simulations and possibly fix some issues for parallel tests. |
|
@dylan-copeland Let's keep in mind that we need regression tests for EQP. |
| @@ -163,6 +164,8 @@ int main(int argc, char *argv[]) | |||
| Array<double> twep; | |||
| Array2D<int> twparam; | |||
| ROM_Options romOptions; | |||
| int nnlsWindow0 = -1; | |||
There was a problem hiding this comment.
Maybe we need a documentation for these new ideas, e.g., in the Overleaf document.
There was a problem hiding this comment.
I agree that some high-level documentation outside the code would be very helpful. Of course, more comments in the code are always needed as well.
rom/laghos.cpp
Outdated
| @@ -1310,7 +1349,7 @@ int main(int argc, char *argv[]) | |||
| for (int curr_window = numWindows-1; curr_window >= 0; --curr_window) | |||
| basis[curr_window]->writePDweights(pd2_tdof, curr_window); | |||
| } | |||
| if (!romOptions.hyperreduce_prep) | |||
| if (romOptions.hyperreduce) | |||
There was a problem hiding this comment.
If I remember correctly, we intentionally leave the option of online ROM via Galerkin projection but NOT using hyper-reduction, which I believe is because we want to see the fact that hyper-reduction is needed for speed-up. But we ended up having these strange conditions !romOptions.hyperreduce_prep. Should we get rid of this consideration from now on?
There was a problem hiding this comment.
This change looks like a mistake. It has nothing to do with this PR. I will revert this change, see b648ac1. You still have a valid question, which I think should be addressed outside this PR.
rom/laghos.cpp
Outdated
| @@ -2082,6 +2131,22 @@ int main(int argc, char *argv[]) | |||
| << sqrt(tot_norm) << endl; | |||
| } | |||
|
|
|||
| if (romOptions.hyperreductionSamplingType == eqp) | |||
There was a problem hiding this comment.
Actually, it will be interesting to be able to see how the energy conservation is like for any kind of hyper-reduction instead of limiting it to EQP. How about we make an option like calculate_energy?
There was a problem hiding this comment.
It turns out this code is not reached, so I will remove it, see b648ac1. The energy change is output at the end, around line 2523. This PR added more details about the energy change there.
No description provided.