Jed Brown (c06f0325) at 17 Mar 18:59
ci - use new CodeCov uploader
Jed Brown (d6a7779a) at 17 Mar 18:43
ci - use new CodeCov uploader
Jed Brown (c443f4ad) at 16 Mar 22:20
rust - clearer exporting of CARGO_CEED_OPT_FLAGS
... and 1 more commit
Jed Brown (09194783) at 16 Mar 18:06
Mirrors policy update in libCEED and micromorph/ratel!1305.
I've added all the Authors in AUTHORS.md except for Ken, who's Gitlab account I can't find right now. But free to ask anything or simply approve the MR.
Closes #49
CHANGELOG.md
Jed Brown (09194783) at 13 Mar 23:28
Merge pull request #1940 from CEED/jeremy/rust-ceed-opt
... and 3 more commits
Jed Brown (78eb7e18) at 13 Mar 20:41
cov - missing excl marker
Mirroring libCEED's policy: https://github.com/CEED/libCEED/pull/1933
I have listed all Ratel authors (in AUTHORS.md) as reviewers. We'd like to invite comments (or a simple 'approve') of this proposed policy.
Closes #N/A
CHANGELOG.md
Removes the leading 4 spaces from TSAdapt monitor print messages. The correct tabbing is already handled by PetscViewerASCIIAddTab(). Change stems from discussion here: !8882 (comment 3109804544)
I also updated the test output even though the change doesn't impact the tests from passing. Just did so for consistency.
Seems fine. If you run with -ts_monitor -ts_adapt_monitor, you now get the following, which is a little ragged, but acceptable.
123 TS dt 0.036 time 4.428
TSAdapt none ssp 0: step 123 accepted t=4.428 + 3.600e-02 dt=3.600e-02Jed Brown (05b5fa34) at 13 Mar 18:22
rust - minor lifeline clarification
Jed Brown (3e7e75fe) at 13 Mar 18:12
rust - allow setting ceed OPT flags for rust, mostly for CI/cov
Jed Brown (a285975a) at 13 Mar 18:01
Jed Brown (a285975a) at 13 Mar 15:43
Merge pull request #1939 from CEED/jeremy/dealii-build
... and 3 more commits
Jed Brown (4d7a563f) at 13 Mar 15:34
make - force dealii serial build from top make
Jed Brown (16baf82b) at 13 Mar 15:15
LLM/Gen-AI Policy (#1933)
As a practical matter, the diagonal version is still desirable because it accesses so much less memory than existing matrix formats. Once upon a time, we discussed having a MATDIAGONAL, and we still could implement it that way.
The problem being addressed here is that state vectors \bm x (and increments \Delta \bm x) are often dimensional (e.g., pressure in Pa or bar, velocity, temperature, energy), but it would be desirable to have a way to make the arc-length method \pi-invariant (invariant to choice of units). With a state-dependent matrix, one could make the arc-length method invariant to choice of state variables (e.g., primitive, conservative, or entropy variables for compressible flow). That is, measuring arc-length could effectively transform from whatever the state variables are to whatever variables one wants to measure length in. But it would more memory-concise to implement that as a nonlinear arc-length calculation (no matrix in sight), and I think it's out of scope for this PR.
The synthetic text you link is pretty incoherent. The cylindrical coordinates snippet would be if the state variables of the system was position in cylindrical coordinates (not an FEA method in cylindrical coordinates) and then the description is elicited from general coordinate transformation/metric tensor stuff (unrelated to arc-lengths). I know it's nicely-shaped, but that stuff just sows confusion.
Jed Brown (8cbca1ea) at 12 Mar 23:17
dealii - set cmake min to 3.10