Draft: make examples use periodicity and SFC box meshes #1128
Conversation
| @@ -77,6 +77,7 @@ CEED_QFUNCTION(Setup)(void *ctx, CeedInt Q, const CeedScalar *const *in, CeedSca | |||
| const CeedScalar A33 = J11 * J22 - J12 * J21; | |||
| const CeedScalar detJ = J11 * A11 + J21 * A12 + J31 * A13; | |||
|
|
|||
| if (detJ <= 0) printf("%s: detJ %g\n", __func__, detJ); | |||
There was a problem hiding this comment.
@jeremylt Do you have thoughts about how we should report such a thing to the user? This is especially noisy (will generate millions of lines on real problems) so not suitable for merging, but it would be good to have a clear message to the user about inverted elements (usually represents a mesh IO problem or issues like this that I'm pretty sure we've stumbled into).
There was a problem hiding this comment.
Making a debug macro for qfunction use would be easy enough, but reporting only once would be harder. No obvious solution comes to mind
There was a problem hiding this comment.
We could make a mesh checking gallery QFunction that uses context to flag if the mesh has inverted elements and only report once
jedbrown
force-pushed
the
jed/sfc-periodic
branch
from
1689496
to
e365d64
Compare
jedbrown
force-pushed
the
jed/sfc-periodic
branch
from
e365d64
to
4ca9f30
Compare
|
With #1529, we officially have a |
|
Thanks @jrwrigh, it's there anything we should salvage in this branch or just close it as superseded? |
|
Not that I can think of, unless you were wanting to keep the conversion of gaussianwave to 3D or the inverted element checks. |
This depends on https://gitlab.com/petsc/petsc/-/merge_requests/5411 so won't be ready for merging soon/ever, but I wanted to put examples out there for discussion.