Porting CESM2.13 to ARCHER2 memory allocate issue

[m_mineter@ed_ac_uk]

Hello!
Archer2 is the new Cray HPC for UK research.
It has nodes each with 128 cores and 256GB and uses SLURM.

I wondered if you can guide me on this problem that has arisen as I try porting CESM 2.1.3 and running a small case I had used before on another computer:
--compset F2000climo --res f19_f19_mg17

The model builds and begins running but quickly fails in an allocate as shown in the cesm log file referred to in CaseStatus .
(The error message from allocate is I assume not being set so its returning noise after the error arose)

I imagined it might be some memory-related option I missed in the Cray compiler.
ftn --version
Cray Fortran : Version 10.0.4

cc --version
Cray clang version 10.0.4 (ffb772459f6195ccd74395703e557b253fc8ee27)
Target: x86_64-unknown-linux-gnu
Thread model: posix

I attach:

• archer2_version.txt: - output from the describe_version script
• dotcime.tar: the .cime xml files
• sacct_output.txt: the output after teh job failed by running sacct
• CaseStatus
• cesm log file
• tar of the xml files in the case directory

Kind regards
Mike Mineter

 

[jedwards]

Hi Mike,

Although we have a definition of the cray compiler in config_compilers.xml and other locations in the release, we have never been successful in getting it to work. If you have access to an intel compiler I would recommend starting there. I don't see the cesm.log file in your attachments.

 

[m_mineter@ed_ac_uk]

Many thanks Jim.
oops. cesm log now attached.
On Archer2 we have gnu and AOCC. But not Intel
Is there for example a cheyenne gnu config we might reasonably start from?
Regards
Mike

 

[jedwards]

Yes gnu is tested regularly on cheyenne and other systems. I would be interested if you make any progress with the cray compiler.

 

[m_mineter@ed_ac_uk]

I think in the short term we need to get something to work so will hope to return to cray compiler in a while if we have more people-time.