Conversation
|
@tslin2 @RAbolafiaRosenzweig This is a further bug fix for WRF coupling to solve openMP compiler issue. |
|
|
||
| end type NoahmpIO_type | ||
|
|
||
| type(NoahmpIO_type) :: NoahmpIOdom(5) |
There was a problem hiding this comment.
Is the 5 upper limit? Is it possible to take from WRF? or maybe add a error message and stop model to ask for change here and recompile model if running for more than 5
There was a problem hiding this comment.
This is a good point. "5" is an arbitrary max number of nested domains I put for NoahmpIOdom structure. We could link it to max_dom from WRF but that would create additional dependency from noahmp code to WRF high-level model domain file setup in compilation, which I prefer not to do it. Normally people would not use 5 nested domain setup so that's why I put a reasonable value here to avoid allocating too large NoahmpIOdom structure to save memory. We may add an error message later. For now let's see if this bug fix could pass all WRF regression test.
|
The PR resolves the issue of OpenMP, the tests was performed. Bitwise consistent results for gcc serial, smpar and dmpar. the tests for intel dmapr works and show bitwise consistent for restart and continuous runs on my end |
2 participants
This PR is to fix the bug in WRF-NoahMPv5.1 coupling for openMP compiler option.
Tests on Derecho have been successful for both openMP (smpar), serial, and MPI (dmpar) for gcc. Intel compiler tests are also successful.
MMF run and urban run as well as nested run are all successful on my end.