Log In | Users | Register
COSMO Web
Edit | Attach | New | Raw | Diff | Print | | Tools
You are here: ICON » ICONmeet20190131

ICON meeting January 31st, 2019

Participants:

Colombe Siegenthaler (CS); Katie Osterried (KO); Philippe Marti (PM); Valentin Clement (VC); Will Sawyer (WS); Remo Dietlicher (RD); Xavier Lapillonne (XL) - partially;

* Xavier Lapillonne (XL):

  • Xavier will propose another workflow at the next ICON infrastructure meeting in Hamburg. Some of the propositions will be: to transfer all the ICON branches on a GitLab server hosted by DKRZ, use the issue tracking and code review facilities of GitLab instead of Redmine, and include Buildbot in the GitLab
  • XL tried to compile ICON-ENIAC on mistral with NAG compiler, got a ton of warnings. XL is trying to remove the warnings.

* Will Sawyer (WS):

  • Will will present activities conducted at C2SM and CSCS, as well as DSL developments at the ICON meeting
  • WS will try to get internal support for CSCS branch in Buildbot. XL will write an email to state how this support would be useful for Meteoswiss.
  • WS finished the gpu porting of the dynamical core. It will be in merge_candidate_4 branch.
  • WS extended the GPU branch in buildbot.
  • The buildbot line to test PGI is not done yet.
  • The official PGI version 18.10 should come soon (within the next 3 months). WS will inquire when the official release from PGI will be.
  • DKRZ is now interested by GPUs, as their next machine will be 40% GPU.
  • Thomas Schultess wants to have a full ICON on GPU as soon as possible. He wants to have benchmarks to get some basis for the next machine. WS did an ICON test: run 2.5 km resolution on 1536 nodes: 2.4 times faster P100 against Haswell.

* Valentin Clement (VC):

  • Valentin is testing CLAW in JSBACH. Remaining time estimation is in the order of weeks.

* Katie Osterried (KO):

  • Katie did a big release of Extpar at the end of last year.
  • KO and XL decided on a strategy to adapt the COSMO testsuite for ICON. They will use the ICON output (in Netcdf format), process it in a python script using the cdo package and write out an ASCII file containing mean, min and max of the needed fields. The latter ASCII file will be read by the testsuite. The timeframe for the adaptation of the testsuite is the next 6 months.

* Philippe Marti (PM):

  • Philippe is moving the ECHAM parametrisations (vdiff, nsurf_diag, update_surface) which are ported with OpenAcc into the full ICON. Hopefully ICON containing the ported physics will be ready for the ICON developer meeting in February. PM cannot add the "Graupel physics" yet because the interface calling the graupel is missing in the official meeting. This point will be raised at the next infrastructure meeting.
  • PM's work should be in the merge_candidate_5 branch.
  • PM worked with Cray compiler. PM would need PGI 18.10 which is not available yet.
  • The graupel test crashes sometimes depending on the nproma value.

* Remo Dietlicher (RD):

  • Remo prepared some slides on ICON workflow. You can find them here.

* Colombe Siegenthaler (CS):

  • Colombe is working on installing ICON-HAM on Piz Daint. ICON-HAM is based on an older ICON version (coming from the AES branch) than ICON-ENIAC because people want to compare directly with the version published by MPI.
  • CS is trying to put Bernhard's ICON-LAM case and the ICON DWD tools used to create the boundary files in Jenkins,
  • CS reported that we get no segmentation fault anymore when running ICON-ENIAC on more than 2 nodes with the gcc compiler.
Attach
pdf
version 1 uploaded by ColombeSiegenthaler on 06 Feb 2019 - 13:57
spacer
This site is managed by the Center for Climate Systems Modeling (C2SM).
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors. Ideas, requests, problems? Send feedback!