Port to stampede2 for XSEDE. skip-ci

[MinsukJi-NOAA]

Description

Port to stampede2 for XSEDE users.

Issue(s) addressed

#460
NOAA-EMC/global-workflow#280

Testing

No testing is required.

On stampede2, the following regression tests have been built and run:

COMPILE | SUITES=FV3_GFS_2017_coupled,FV3_GFS_2017_satmedmf_coupled,FV3_GFS_v15p2_coupled,FV3_GFS_v16_coupled S2S=Y WW3=Y
RUN | cpld_bmark_wave
RUN | cpld_bmarkfrac_wave
RUN | cpld_bmarkfrac_wave_v16

Dependencies

No

[DeniseWorthen]

Will we be maintaining a full baseline on Stampede?

[MinsukJi-NOAA]

Will we be maintaining a full baseline on Stampede?

I strongly doubt that. The tests I did on Stampede were just completing the simulations, not comparing with baselines.

[DeniseWorthen]

I see. But the use will be primarily Prototype-like runs by groups outside of EMC using a workflow. Is that right?

[MinsukJi-NOAA]

@DeniseWorthen , that is my understanding.

[climbfuji]

@DeniseWorthen , that is my understanding.

I "ported" the ufs-weather-model to stampede twice in the past. If regression tests are not run, who makes sure that the code keeps working there? That didn't even work with Cheyenne, even when it was a tier-2 (post commit regression testing) platform.

[MinsukJi-NOAA]

@climbfuji. Good point. As Denise mentioned, groups outside EMC will use the global workflow. Maybe any changes specific to stampede should be in the global-workflow repository? For sure, the changes in rt.sh and fv3_slurm.IN_stampede are not needed. I am not sure if modulefiles/stampede/fv3 is directly used by global workflow either (probably not).

[junwang-noaa]

Code changes look good to me, assuming they were tested on stampede2.

[MinsukJi-NOAA]

Changes merged in via #477. Thanks @DusanJovic-NOAA!