Repository navigation
Send OASIS fields before writing history & restart files - #162
Merged
Merged
Conversation
kvrigor
marked this pull request as ready for review
November 6, 2026 08:07
s-poll
previously approved these changes
Nov 8, 2026
s-poll
left a comment
Member
There was a problem hiding this comment.
Thanks Paul. Nice PR. The send now happens right after lnd2atm, before history and restart output and I confirmed bit-identity.
Two small requests:
- Could you please put the lines, marked in this review, in ifdef statements.
- Could you repeat exactly the same simulations as in the ICON-GPU-eCLM scaling study, with and without this PR to quantify the runtime benefit? On JURECA with the default setup there was a runtime benefit of 2%, but there is a larger benefit expected on heterogeneous setups.
Member
Author
|
Thanks for looking into this @s-poll . Re on the ICON-GPU-eCLM scaling study: I'd prefer to do it after this PR is merged since such benchmarking exercises are costly in terms of time and compute resources. This PR also doesn't claim there would be performance improvements, although one could speculate. More importantly, as mentioned in my first post, this PR tries to simply fix an obvious algorithmic mistake and not introduce any regressions. |
Even though it's unnecessary
s-poll
approved these changes
Nov 9, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Originally, the coupled fields had to wait for eCLM disk writes to finish before being sent to ICON and/or Parflow. Since there's no good reason to do this, it's safe to change the order from
write -> oas_sendtooas_send -> write:I've manually verified the fixes by running a 3 sim. day, fully-coupled EUR-11 simulation. Simulations completed without issues and took around ~25 minutes. No time savings could be observed, but it remains to be seen how the
oas_send -> writeorder affects the runtime of much longer simulations. For now it's sufficient to simply fix an obvious algorithmic mistake and make sure it does not break existing functionality.For those interested in the results, see attached LLview reports and the simulation runs archived at
/p/data1/slts/rigor1/verification/eCLM_PR162_dev-refactor-oasis-sendrecvllview_results_after-refactor.pdf
llview_results_before-refactor.pdf