Closed stvdwtt closed 5 months ago
The CI error is unrelated to this PR. I did not know this but fetnat08
and fetnat09
only have four cores while all the others GPU machines have at least six. The error in the CI is from MPI because we are trying toe oversubscribe the machine.
While debugging #270 we realized that we don't have a test where the mesh is allowed to coarsen after the beam. This PR adds a test for that.
Note: Due to the issue with mesh coarsening across active/inactive boundaries (#274) this test case had to be carefully chosen. The fix to #274 should have a similar test to this one but with a shifted scan path to force active/inactive boundaries across children of the same coarse mesh element.