MALI update to fix issues from earlier PR causing sporadic test failures #6627
Conversation
|
|
@matthewhoffman -- I had also tested with adding a default value for albanyVelocityError in Registry. I don't think it makes any difference, but we could add it here if you want |
|
It passed the failing ERS.f09_g16_g.MALISIA.pm-cpu_gnu three times in a row with these changes |
|
Also ran ERS_Ld5.TL319_oQU240wLI_ais8to30.MPAS_LISIO_JRA1p5.pm-cpu_gnu.mpaso-ocn_glcshelf successfully three times with no fails. That test has also been having random problems since PR #6559 |
There was a problem hiding this comment.
@jonbob , looks good to me. Thanks for tracking this down. You are welcome to add the addition of the default_value for albanyVelocityError to Registry. Making the default default value explicit in this case isn't a bad idea.
|
Thanks @matthewhoffman -- OK, I've added the default value as well |
|
Sounds like you've solved it. You might also try with more MPI's. Maybe 1024? |
MALI update to fix issues from earlier PR causing sporadic test failures Including a variable that was deactivated in the globalStats stream caused sporadic failures during the second run of some ERS tests on several platform/compiler combinations. That variable is now only included when MALI is using Albany. Also updates a namelist default that had been missed but does not change answers. Fixes #6600 [NML] for configurations with MALI [BFB]
|
Passes, with expected NML DIFFs:
Successfully runs:
which had been failing Merged to next |
|
merged to master and expected NML DIFFs blessed -- plus regular DIFFs for the two tests fixed by this PR |
Including a variable that was deactivated in the globalStats stream caused sporadic failures during the second run of some ERS tests on several platform/compiler combinations. That variable is now only included when MALI is using Albany. Also updates a namelist default that had been missed but does not change answers.
Fixes #6600
[NML] for configurations with MALI
[BFB]