problems using pio2 or pio1 with pnetcdf
mvertens opened this issue · comments
I am running the following compset:
SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap
in my work to introduce an ungridded time dimension in the import/export fields.
If I use PIO2 with pnetcdf, the time variable in the clm history file is garbage.
If I use PIO1 with pnetcdf, I get a crash in DUSTMod with the following traceback
0:cesm.exe 0000000000AA28FF abortutils_mp_end 50 abortutils.F90
0:cesm.exe 00000000006FF6AF dustmod_mp_dustem 336 DUSTMod.F90
0:cesm.exe 00000000004D6D75 clm_driver_mp_clm 670 clm_driver.F90
0:cesm.exe 00000000004BC01B lnd_comp_nuopc_mp 823 lnd_comp_nuopc.F90
With netcdf and PIO1 - everything runs fine.
In addition, if I compare the times using PIO2 versus PIO1,
The LND Run Time with PIO2 is 2.390 seconds
The LND Run Time with PIO1 is .909 seconds
Please provide the information necessary to reproduce the problem. Can I do this on the current nuopc-cmeps head?
On the head of nuopc-cmeps
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap CREATE_NEWCASE
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap XML
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap SETUP
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap SHAREDLIB_BUILD time=270
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap MODEL_BUILD time=54
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap SUBMIT
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap RUN time=30
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap MEMLEAK insuffiencient data for memleak test
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap SHORT_TERM_ARCHIVER
I also ran from ~mvertens/src/UFSCOMP.jan19/cime/scripts/
cheyenne3: /glade/u/home/mvertens/src/UFSCOMP.jan19/cime/scripts
:) qcmd -- ./create_test SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap CREATE_NEWCASE
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap XML
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap SETUP
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap SHAREDLIB_BUILD time=260
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap MODEL_BUILD time=56
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap SUBMIT
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap RUN time=19
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap MEMLEAK insuffiencient data for memleak test
PASS SMS_Vnuopc_Ln5.f45_f45_mg37.I2000Clm50SpNuopc.cheyenne_intel.clm-nuopc_cap SHORT_TERM_ARCHIVER
I've pushed a fix to:
To https://github.com/mvertens/ctsm.git
eb07bf2..8ab1d37 mvertens/nuopc_cap -> mvertens/nuopc_cap
this also fixes an issue with flds_scalar in debug mode