EIRENE-NGM development, Code Camp 2022, KU Leven

21 Nov 2022, 09:00 → 24 Nov 2022, 18:20 Europe/Berlin

KU Leven (In persion / ZOOM )

Dmitriy Borodin (FZJ), Wouter Dekeyser (KU Leuven)

Dinner.pdf

Photo_db.jpg

Photos-code_camp.zip

Monday 21 November

09:00 → 13:45 Arrival, getting started, bilateral talks, etc.

DB:
Wouter has started sending formal invitations from KUL as a host - please take care to contact him if you need one!

DB:
Many have already applied and received EUROfusion missions. For ACH I do not know, however by TSVV budget we have reached the limit. Petra's mission is on hold for now.

KUL:
Advice on local accomodation etc. (maybe copy-paste from SOLPS training)...

13:45 → 14:00 FORMAL START. Greetings, agenda approval, misc.

14:00 → 16:00 A1: Version and simulation case(s): A1

Taking decisions on main code brances (which are to be followed, which are to be used for e.g. parallelisation or certain validation cases) and upcomming master release.

Convener: Yannick Marandet (CEA)

What we agreed on (YM recollection ...) :

1. What the release will be (still 4 branches to merge) and when it comes out, after full tests - January 2023
2. Everybody uses the Jülich repo for development, downstream branches for each 'husband' code
3. A review comity is setup for merge requests into develop with representative of every child branches -  (following the DCOC agreed upon some time ago)
4. How to include the full user community is not yet clear since FZJ will not grant access to everybody (set up an external bug tracker ?, …)
5. More discussion needed this week on the licence (what does it give access to : master + beta develop ? a snapshot ? how technically ?) & connection with the timing of the release  TBD

NB there is a new major release of SOLPS-ITER planned in the first months of 2023, would naturally include the new EIRENE release.

14:00 Overview of the EIRENE versions

10-15 min talk + questions

Speaker: Petra Boerner

Eirene_branches_2022.pdf

16:00 → 16:20 Break (20 min)

16:20 → 18:20 B1: Grouping of variables into structures: B1

(Including in a view of IMAS)

Convener: Jorge Gonzalez (DIFFER)

Outcome of the discussion (more focused on streamlining the parametrization of A&M models)

- agreement on the benefit of revising MODCLF/MODCOL for code readability and maintenance

- first work on the setup of the data structure until end of January (departure of Jorge), JG/PB/YM

- then see who implements (potentially AMU)

Yannick (preparation)

B1 there the idea is we sit down and go through a selected number of modules to agree on what a proper grouping would look like ? I think we need to prepare upfront and agree beforehand on the targeted modules (comprt, …)

     implementation comes when the new master version is out but we could nevertheless (as for folneut below) code at least the module, without replacing all the instances of different variables in the code

Short-circle meeting is needed (YM, JG, PB, DB,...) - organized by JG!

16:20 Main policies and characteristic examples

Speaker: Jorge Gonzalez (DIFFER)

2022-11-21 Grouping Variables.pdf

modcol_by_reaction_types.odt

Tuesday 22 November

09:00 → 10:40 Streamlining the main loop: A2

Convener: Yannick Marandet (CEA)

streamlining_yannick_2022.pdf

YM:

A2 here we have the prototypal folneut routine we designed last December (getting rid of the gotos, using a big select case). This has not been tested so far but could potentially be done on our side, at least to some extent.

     We could identify other critical targets (folion is the obvious one; but it is very similar to folneut). I have to look into locate etc…

     Then there is the unification of the grid description between structured and unstructured grids - to be discussed with Wouter whether there have been progress on this. We could spend quite some time on this ...

>> Presention on the status (PB)
>> indentify further routines
>> Contact Wouter on progress with unification between struct/unstruct grids

10:40 → 11:00 Coffee break

11:00 → 12:30 Joint session with ACH: C3

Convener: Dmitriy Borodin (FZJ)

• Docker can be used alternative to Gateway. This is a way to IMAS.
• Upcomming HTTP access to IMAS will not substitute Docker (mandatory for productive runs)
• Wrapper solution can be run at this environment instead of Gateway

 

Feedback from TSVV:

1. The instalastion unstructions should be more reliable (tested), large commands should be alisases or scripts.
2. We need to have it not just for Windows shell, but rather for MinGW/WSL/WSL2 envioronments. 
3. We need clear instructions for basic cases with examples
   • How to download a signal XX for JET pulse XXX
   • How to fetch the run XX?
   • How to upload the run?..
4. Invetigate how much of WSL/upstream linux one can accessed from insider docker sandbox
5. How to mount local resources from Docker container? We would enjoy couple of standard setup scripts.

 

Feedback from ACH:

• What should be inlcluded into the docker container?

11:15 Gateway Docker demonstration

Speaker: Dmitriy Yadykin (ACH IM)

IMAS_Docker_example_ACH_Poznan.pdf

12:30 → 13:45 Lunch

13:45 → 15:00 Discussion/report/feedback on the coding advancement

Speaker: ALL

15:00 → 16:20 IMASification (joint with ITER): B2

New IMAS tools available

• IMASViz (-> "viz" to launch) - getting the profiles
• md_summary
• ParaView -> VTKGGD - full plasma solutions independent on the  particula codes. One can load 2 BGs simiultaniously, operations are to expect.
• https://confluence.iter.org/display/IMP/Scenarios+for+Diagnostic+Development
• DivGeo (manipulating the surfaces), dg2ids, uinp. One can create the run with amendments from the existing.
   

SimDB will include all the files. Still, EIRENE iRods (EUDAT) based system is needed:

1) ITER will not accomodate all the runs, but just the selected
2) We should not loose flexibility from technical side

3) We not necessarily wish to publish for all ITER community all of our runs.

 

Additional points:

• HDF5 formatted data is preferrable over the MDSPlus
• EIRENE should create the run from the IDS

 

 

16:20 → 16:40 Coffee break

16:40 → 18:00 IMASification (joint with ACH): B3

Conveners: Piotr Chmielewski (Institute of Plasma Physics and Laser Microfusion), Dr Yurii Yakovenko (IPPLM)

EIRENE_IMASification_status&questions_v2.pptx

• Additional output is to be provided (covering generic grid subsets), which is to be referred in the input
• Anything of intrest should also be inluded into that output
• The "triangle_structure" type can be a proxy for a generelized grid type (class). It can be parential to the chindren classes for particular grid types.

 

Wednesday 23 November

09:30 → 10:50 CRMs and A&M data: D2

Convener: Francesco Cianfrani (FZJ)

CCCRMsession.pptx

YM:
Wednesday : personally I would suggest using the two morning sessions into one CRM and A&M data - the parallelization of the call to the setamd routine (prior to the main loop but a heavy bottleneck when you will use an actual CRM in every cell) we are currently working on fits in there. the remainder of the parallelization work, on the main loop, goes through EIRON at the time being - optimization of the OpenMP layer and investigation of domain decomposition schemes (in fact, the idea of having talliers processes goes beyond what we initially had in mind). So this part could be to the topic for Wednesday afternoon, as we discussed this afternoon

DB: 
>> HydKin progressa
>> ColRad progress
 

FC:

_quick intro existing A&M databases

_HYDKIN: restyling and perspectives

_quick review slreac, tab2d, colrad

_load datas from external databases (read from json, tabular, xml):

                 --through HYDKIN??

                 --use tab2d and define new type in eirmod_comxs

_update colrad.

10:20 CRM and photon tracing module in EIRENE

20+10 min

Speaker: Ray Chandra (VTT)

TSVV_Code_camp_CRM_rchandra.pdf

10:50 → 11:10 Coffee break

11:10 → 12:30 Parallelisation and CRMs: A2

Convener: Yannick Marandet (CEA)

parallel_CRM_yannick.pdf

12:30 → 13:45 Lunch

13:45 → 14:30 Discussion/report/feedback on the coding advancement

Speaker: ALL

Discussion on Geometry generalisation

• Can we adapt the type from IMAS convesion for that? Use it as a basis?
• What alreratives to use of OOP do we have to emulate polymorthism (e.g. calling functions by reference)? What about using pre-processor?
• How this is compatible with domain decomposition?
• Introducing versioning for the grid type (e.g. for matching the IMAS evolution)

 

 Additional discussion.

• Versioning of EIRENE input files (using the power of JSON...)
• CMake use documentation
• Documentation on isntallation/compilation (already started)

14:30 → 15:00 Magnum-PSI modelling: CRMs and FEM for PSI

(20 + 10) min

Speaker: Jorge Gonzalez (DIFFER)

2022-11-23 MagnumPSI.pdf

15:00 → 16:20 EIRON status and aplications

Convener: Oskar Lappi (ACH-VTT)

Eiron__Code_Camp_2022-1.pdf

16:20 → 16:40 Coffee break

16:40 → 18:00 Special session on EIRENE hybridisation

Convener: Giovanni Samaey

20221123_TSVV5CodeCamp.pptx

A talk is really wishful - Who?! 15-20 min + questions/guidelines for discussion

Result - summary of the discussion...

 

Wouter/Niel/Wim: may be contribute?...
- short presentation and/or providing few slides and/or points for discussion

==> DEMO simulation cases - chalenges, constrains, status...

Thursday 24 November

09:00 → 09:30 Detachment studies in JET unseeded H, D, T and DT L-mode plasmas

(20+10)min

Speaker: Mathias Groth (Aalto University)

MGroth_Update_JET_LmodeExp_24Nov22.pdf

09:30 → 10:15 Discussion/report/feedback on the coding advancement

Speaker: ALL

10:15 → 12:30 Summary talks and discussions

For each line of the code camp should be a summary and a proposed list of action points to be proposed and agreed during the discussion

10:15 Topic A: Code streamlining and refinement

Speaker: Yannick Marandet (CEA)

summary_section_A.pdf

10:35 Topic B: Imasification and data structures

Speaker: Jorge Gonzalez (DIFFER)

2022-11-24 Conclusions.pdf

10:55 Topic C: interaction with ACH incl. parallelisation

Speaker: Dmitriy Borodin (FZJ)

SummC_ACH_dbor_v2.pptx

11:15 Topic D: CRMs and A&M data

Speaker: Francesco Cianfrani (FZJ)

summary0.pptx

FC:

_quick intro existing A&M databases

_HYDKIN: restyling and perspectives

_quick review slreac, tab2d, colrad

_load datas from external databases (read from json, tabular, xml):

                 --through HYDKIN??

                 --use tab2d and define new type in eirmod_comxs

_update colrad.

12:30 → 13:45 Lunch

13:45 → 14:35 Wrap-up and future plans

1) This code camp was absolute success. It is already 2nd experience proving that this kind of in person communication is really needed.  We suggest to put a high priority to have such event every year.

2) We need new joint publication at FEC conference 2023 - we decide to focus on CRMs this time.

3) We need seemless and consequent automatic versioning of i/o formalts, code versions etc interlinked. Git hashtags to be linked as well, but main versions are to be defined by the developers. All should be grouped in a dedicated compact input file similar to one of SOLPS-ITER.

14:35 → 15:00 Fixing the meeting points and adjorn

Speaker: Dmitriy Borodin (FZJ)