[test] New Slurm setup for Cp2kGpuCheck

[lucamar]

I propose to run the Cp2kGpuCheck with 6 tasks per node and 2 cpus per task, as it is slightly more efficient with respect to the current Slurm setup.
I would also like to test the CP2K modules recently provided with the latest stable release 7.1: however, the new modules are not yet default.
Should we change directly the modules loaded in the existing checks or should we rather create a new set of checks for the new versions with a new variant (e.g. dev) and keep the existing ones for the default modules only?
The latter would be probably a better solution to keep track of the performance of the new vs. old release of applications: the new variant would result in having Cp2kCpuCheck_small_dev, etc...

Please let me know your opinion and I would adjust the checks accordingly.

[pep8speaks]

Hello @lucamar, Thank you for updating!

Cheers! There are no PEP8 issues in this Pull Request!Do see the ReFrame Coding Style Guide

Comment last updated at 2020-06-09 21:12:30 UTC

[lucamar]

@jenkins-cscs retry dom

[lucamar]

The check on dom:mc failed since the nodes were not available:

Reason: job blocked error: [jobid=1015576] job cancelled because it was blocked due to a perhaps non-recoverable reason: 
ReqNodeNotAvail,  UnavailableNodes:nid00[002,038-040,043,404-405,418-419]

[lucamar]

@jenkins-cscs retry dom

[lucamar]

No chance to have 6 mc nodes on Dom:

dom101:~$ scontrol show nodes nid00[000-011,032-043,052-055,404-407,416-419,401-402] | grep -A 5 ActiveFeatures=mc | grep State
   State=MAINT+DRAIN ThreadsPerCore=2 TmpDisk=0 Weight=1 Owner=N/A MCS_label=N/A
   State=MAINT+DRAIN ThreadsPerCore=2 TmpDisk=0 Weight=1 Owner=N/A MCS_label=N/A
   State=RESERVED ThreadsPerCore=2 TmpDisk=0 Weight=1 Owner=N/A MCS_label=N/A
   State=IDLE ThreadsPerCore=2 TmpDisk=0 Weight=1 Owner=N/A MCS_label=N/A
   State=IDLE ThreadsPerCore=2 TmpDisk=0 Weight=1 Owner=N/A MCS_label=N/A
   State=IDLE ThreadsPerCore=2 TmpDisk=0 Weight=1 Owner=N/A MCS_label=N/A
   State=IDLE ThreadsPerCore=2 TmpDisk=0 Weight=1 Owner=N/A MCS_label=N/A
   State=MAINT ThreadsPerCore=2 TmpDisk=0 Weight=1 Owner=N/A MCS_label=N/A

[vkarak]

Hi @lucamar, is this setup valid also for the default CP2K version? If yes, I would suggest to keep the default module and invoke reframe with -M to map to the new version.

[lucamar]

@vkarak Thanks for your comment: yes, the new setup is also valid for the current default module.
Of course we can use the module mapping to test the new module, as I have being doing so far: the only disadvantage is that you need to issue two different commands for gpu and mc checks, since the modules to be mapped are different.
I thought that it would be interesting to keep track of the performance of the new modules of supported applications before changing the defaults: maybe we can find a different way to do that...

[vkarak]

@lucamar What if we parametrized the test on the module name?

[codecov-commenter]

Codecov Report

Merging #1304 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #1304   +/-   ##
=======================================
  Coverage   91.66%   91.66%           
=======================================
  Files          83       83           
  Lines       12673    12673           
=======================================
  Hits        11617    11617           
  Misses       1056     1056           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update dc1d118...c302a4b. Read the comment docs.

[vkarak]

lgtm