ESCOMP
some SCAM IOP's are broken
What happened?
Running FSCAM in debug mode, I am getting error's for IOP cases dycomsRF01, gateIII and bomex:
shr_file_mod.F90 912
This routine is depricated - use shr_log_setLogUnit instead -139
forrtl: severe (408): fort: (8): Attempt to fetch from allocatable variable MED2MOD_AREACOR when it is not allocated
Image PC Routine Line Source
cesm.exe 000000000A4F637F Unknown Unknown Unknown
cesm.exe 000000000532CA83 ice_import_export 712 ice_import_export.F90
cesm.exe 0000000004D43B53 ice_comp_nuopc_mp 1098 ice_comp_nuopc.F90
libesmf.so 00002B67186A12D6 _ZNK5ESMCI13Metho 377 ESMCI_MethodTable.C
libesmf.so 00002B67186A273A _ZN5ESMCI11Method 563 ESMCI_MethodTable.C
libesmf.so 00002B67186A0C9B c_esmc_methodtabl 317 ESMCI_MethodTable.C
libesmf.so 00002B6718AC97AF esmf_attachmethod 1287 ESMF_AttachMethods.F90
libesmf.so 00002B671A472A94 Unknown Unknown Unknown
arm97 and twp06 both run fine. So it's not all IOP's, just some.
What are the steps to reproduce the bug?
set compset to FSCAM and --user-mods-dir to the IOP directory you want to run. Set DEBUG=T.
What CAM tag were you using?
cam6_3_115
What machine were you running CAM on?
CISL machine (e.g. cheyenne)
What compiler were you using?
Intel
Path to a case directory, if applicable
/glade/scratch/aherring/cam6_3_115_FSCAM_twp06_T42_T42_L58dev_1pes_230712_run94_debug
Will you be addressing this bug yourself?
Any CAM SE can do this
Extra info
No response
I looked into this and it is a problem with the cice nuopc implementation for SCAM. I do have a fix for ice_import_export.F90 that I will forward to the ice group. Until this change goes in SCAM should only work for the land only grid points and IOPs. I will look into twp06 as I believe that should have failed as well.
@jtruesdal do you know when this came in? And I am looking to run togaII and arm95 also, if they are unaffected no problem. But togaII is an ocean IOP of course. Thanks!
I think now that we can only use a more recent cam_development tag on derecho, that precludes us from using SCAM then right? Either way this is pretty important for development and it should be a high priority. We stress this is a high priority to the cice group.
I've started working with Dave Bailey to get a fix to the cice core code through the consortium. The bug was in the NUOPC driver code for CICE and popped up when we switched from MCT to NUOPC. It affects all runs over ocean.