You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Test suite does not compare fully global output #356
TARGET_IO_PES:
description: | When AUTO_MASKTABLE is enabled, target number of IO PEs. If the given target number of IO PEs is not achievable, the target number of IO PEs is set to the nearest smaller number of PEs that is achievable.datatype: integervalue:
$OCN_GRID in ["tx2_3v2", "tx2_3v3"] and $TEST: = - ( - $NTASKS_OCN // 256)$OCN_GRID == "tx2_0v1" and $TEST: = - ( - $NTASKS_OCN // 256)$OCN_GRID == "tx0.25v1": = - ( - $NTASKS_OCN // 128)
output is being split across multiple files. For example, in an SMS.TL319_t233.G1850ABIOMARBL_JRA.derecho_intel this value is set to TARGET_IO_PES = 4 so each stream is broken into four pieces:
The test suite finds a list of files matching a regex, and then picks the last file to use in baseline comparisons. This is supposed to correspond to output from the final day of the test, but instead matches the final block of the final day:
$ cd <BASELINE_ROOT>
$ ls SMS.TL319_t233.G1850ABIOMARBL_JRA.derecho_intel/mom6.*
SMS.TL319_t233.G1850ABIOMARBL_JRA.derecho_intel/mom6.h.bgc.z.0001-01-05.nc.0003
SMS.TL319_t233.G1850ABIOMARBL_JRA.derecho_intel/mom6.h.native.0001-01-05.nc.0003
SMS.TL319_t233.G1850ABIOMARBL_JRA.derecho_intel/mom6.h.ocean_geometry.nc
SMS.TL319_t233.G1850ABIOMARBL_JRA.derecho_intel/mom6.h.rho2.0001-01-05.nc.0003
SMS.TL319_t233.G1850ABIOMARBL_JRA.derecho_intel/mom6.h.sfc.0001-01-05.nc.0003
SMS.TL319_t233.G1850ABIOMARBL_JRA.derecho_intel/mom6.h.static.nc.0003
SMS.TL319_t233.G1850ABIOMARBL_JRA.derecho_intel/mom6.h.z.0001-01-05.nc.0003
SMS.TL319_t233.G1850ABIOMARBL_JRA.derecho_intel/mom6.ic.0001-01-01.nc
Given CIME limitations, I think we need to stitch the blocks back together before the test suite generates or compares baselines
Activity