[YAKL] Update YAKL and PAM submodules #5925
Conversation
|
@brhillman can you take are of this? |
|
@mrnorman @xyuan @abagusetty I see there are some more recent updates to the main YAKL repo, should we include those here? Seems like some of them might be involved with recent PAM dev? |
|
Might as well upgrade. I can help if you get build errors associated with the template parameter type for |
|
Biggest changes are Intel SYCL backend and CPU performance improvements for YAKL, increasing vectorization by switching to |
|
@abagusetty based on above discussion, YAKL should be updated again on this branch. |
|
@abagusetty why did you merge master into this branch? |
brhillman
force-pushed
the
abagusetty/yakl-submodule-update
branch
from
d1f691f to
6d96997
Compare
|
@brhillman thanks for cleaning up the history with a force-push. You can start merging this. |
|
No problem. MMF tests are currently failing to build with this though, looking into that now. |
|
The MMF2 test fails to build with the updated version of YAKL due to PAM needing updates for compatibility. This will require a commit to the external PAM code, then we will need to pull in the updated PAM alongside the updated YAKL. We should wait to merge this PR until that's completed and can be included here. |
|
PAM PR E3SM-Project/PAM#138 will fix the build error, but now I'm getting runtime errors that I'm working on debugging. |
|
PAM PR is merged, so I updated this branch to bring in the PAM update as well. It looks like this branch should now build all MMF tests, but the MMF2 build fails on next due to the recent cmake changes and incompatibility with SCREAM builds (the MMF2 test uses P3 and SHOC from EAMxx). We should probably wait to merge this PR until the MMF2 build fail clears up on the dashboard, which I think will require fixes to E3SM to fix the EAMxx build on E3SM-Project/E3SM. |
|
Please update the PR title/description to mention the PAM update. |
brhillman
changed the title
|
@brhillman you can start merging this. |
|
This looks like it will be non-BFB for MMF2 on pm-cpu at least. Updated description and tags to show that. |
|
@brhillman non-BFB changes are fine - I'm planning some additional non-BFB changes, and the model stability is unacceptable for real science - so I'm not concerned about breaking it. |
|
To clarify, that wasn't a complaint, I just wanted to document the change in the description for provenance. But since this is non-BFB, it needs to wait until Monday. |
|
This is now part of #6153 |
Downstream merge of scream/eamxx fork into E3SM The build system fixes on the fork are needed for E3SM. Also updates E3SM with latest version of eamxx. Merge remote-tracking branch 'scream/master' into jgfouca/scream_downstream_merge Updates submodules: * ekat * haero * mam4xx EAM updates: * Consolidate shoc_Ck nml settings * New iop_coriolis nml setting (for nudging) * initialize IOP information on the first restart time step * SHOC updates * Add wet/dry qv_sat (saturation mixing ratio), old qv_sat is still available for backwards compatibility * P3 updates (scream version only) This PR now includes #5925 : Update the YAKL submodule, and update PAM submodule to work with the YAKL update. [YAKL] [BFB] - (except for SCREAM/EAMXX and MMF2 cases)
Update the YAKL submodule, and update PAM submodule to work with the YAKL update. Non-BFB only for MMF2.
[non-BFB]
[YAKL]