cap update: supporting traditional and ESMF-managed threading#100
cap update: supporting traditional and ESMF-managed threading#100junwang-noaa wants to merge 1 commit into
Conversation
…anaged threading. > > Co-authored-by: Gerhard Theurich <gerhard.j.theurich@noaa.gov>
|
@theurich @jiandewang FYI. |
Codecov Report
@@ Coverage Diff @@
## dev/emc #100 +/- ##
========================================
Coverage 28.62% 28.62%
========================================
Files 252 252
Lines 73515 73515
========================================
Hits 21040 21040
Misses 52475 52475 Help us with your feedback. Take ten seconds to tell us how you rate us. |
|
@junwang-noaa @theurich What I can see is that you removed those open-mp block in cap code, I will ask NCAR to do a pre-test to make sure the change will not break their system. |
|
@jiandewang Thank you for coordinating the code commit! |
|
@junwang-noaa we got greenlight from NCAR, so I am going to add your cap change in my pending MOM6 PR #99, and UFS PR ufs-community/ufs-weather-model#1358 |
|
this is done through #99 |
Code changes are made by Gerhard. The changes are to resolve an inconsistency between traditional and ESMF-managed threading under MOM6. Basically the original threading code tries to do some threading management locally in the cap, and through this essentially turns it off for the traditional approach. ESMF-managed does not suffer from it, because it manages threading on a lower level, and is unaffected by this code. But it does lead to the weird inconsistency that no matter what OMP_NUM_THREADS is set for traditional, MOM6 will always run single threaded. The local threading management code is removed in this PR.
Co-authored-by: Gerhard Theurich gerhard.j.theurich@noaa.gov