opm-simulators icon indicating copy to clipboard operation
opm-simulators copied to clipboard

Enable New Saturation Function Consistency Checks

Open bska opened this issue 1 year ago • 6 comments

Currently opt-in by the new parameter CheckSatfuncConsistency (command line option --check-satfunc-consistency), this PR hooks up the SatfuncConsistencyCheckManager<> pass to the FlowProblem. I hope to make the parameter be "opt-out" at some point, but for now the new pass breaks too many regression tests.

bska avatar Sep 10 '24 16:09 bska

I'm creating this in draft mode for two reasons

  1. The PR depends on, and contains, the earlier PR #5592
  2. I'd like to refine the new parameter, ideally to not exist at all, but that will probably require adjustments to the underlying systems to avoid breaking the regression tests.

bska avatar Sep 10 '24 16:09 bska

jenkins build this please

bska avatar Sep 10 '24 17:09 bska

jenkins build this please

bska avatar Sep 11 '24 18:09 bska

jenkins build this please opm-tests=1226

bska avatar Sep 11 '24 18:09 bska

jenkins build this opm-tests=1226 please

bska avatar Sep 11 '24 18:09 bska

jenkins build this please

bska avatar Sep 12 '24 19:09 bska

This functionality has been tested in Equinor and I'm looking forward to see this going into master.

tskille avatar Sep 26 '24 11:09 tskille

The upstream PR was merged into the master branch. I've pushed an update here that adds a second new command line parameter, --num-satfunc-consistency-sample-points, in addition to --check-satfunc-consistency that was here when the PR was first created. The --num-satfunc-consistency-sample-points parameter allows the user to request more or fewer sample points than the default value of five and I've updated the PR description to reflect this fact.

I'm marking this PR as "ready for review" now.

bska avatar Oct 16 '24 10:10 bska

jenkins build this please

bska avatar Oct 16 '24 10:10 bska

jenkins build this please

bska avatar Oct 16 '24 10:10 bska

I've pushed an update which adds a few more comments to explain the reasoning and I've moved the MPI barrier closer to where the simulation stops in the case of "critical" consistency failures.

bska avatar Oct 16 '24 13:10 bska

jenkins build this please

bska avatar Oct 16 '24 13:10 bska

With the latest changes this is good to go. Merging!

atgeirr avatar Oct 16 '24 13:10 atgeirr