Focus on:  Hide Contributions
Login

3rd CORSIKA 8 Focus Week

→ Europe/Berlin
Description

The CORSIKA 8 focus week is a virtual workshop with at least one coordinating phone call each day, with the aim of concerted work on CORSIKA 8 development.

VIDYO room: 

https://vidyoportal.cern.ch/join/oSNNTkKni5
Fallback: 

DFNconf video system: Room Name: CORSIKA, Number: 97958319

With your browser under Windows, MacOS, Linux:
  1. https://conf.dfn.de/webapp/conference/97958319
  2. By telephone (audio only):
Call 0049 30 200 97958319
  3. Via Skype for Business:
97958319@vc.dfn.de
  4. By a mobile smart phone.
You have to install the Pexip App beforehand: Please enter: 97958319@conf.dfn.de

Meeting password (if asked) is 1234

Participants
  • Anatoli Fedynitch
  • Dieter Heck
  • felix riehn
  • Lukas Nellen
  • Maximilian Reininghaus
  • Ralf Ulrich
  • Tanguy Pierog
  • Mon, February 3
    • Discussion Monday

      Welcome to 3rd CORSIA 8 focus week

       

      Summary of Monday discussions:

       

      General motto and philosophy:

      Physics first. Code clean. Optimize and profile next. Watch out for potential "dead end", but go ahead. We cannot avoid to eventually go back and perform major code revisions for further optimizations. Data structures are at least as important as code!

       

      General impression:

      We are going in the right direction. A lot of progress. CORSIKA 8 prototype can already do many things.

       

      - Brief review of status, MR, issues, milestones

      - Have code dependencies in source tree

      - complicated/binary dependencies in containers

      - E.g. for PYTHIA, have helper script to remove all bloat -> create minimal package

      -> Need maintenance project with tools. I suggest in corsika-docker

      - check "git large file system", but license issues. Alternative: git binaries

       

      - discuss paper plans and general strategy for the next time on Friday

      - list of conferences, who is going where?

       

      Anatoli presented a talk about parallel running and scheduling

      EAS -> branching tree.

      CPU time is almost linear function of sub-shower energy

      Need master to distribute work, and organize

      -> scheduling. sorting according to Wall time -> dispatch work tasks

      -> need also sequential mode (at initialization)

      -> advantages: load balancing, monitoring, log on scratch space transfer only if needed

      About 3 parameters needed for scheduler to optimize. Can be easily optimized on any computing cluster.

      - data serialization is critical. Large number of workers -> may need "data concentrator" nodes, or/and concurrent writes to files

      - MPI is important on large machines, but annoying on personal computers and small clusters. Need with and without option.

      - OpenMP corresponds to "thread model"

      - Data transfer between nodes is very minimal

      --> Define abstraction to allow different backend implementations

       

      • 1
        Discussion

        Welcome to the "3rd CORSIKA 8 Focus Week"

        The aim is to make a major concerted step toward first simulations of both hadrons and e.m. cascades.

        Today we should set our priorities for the coming week and also for the upcoming milestone.

        Week3_intro.pdf
      • 2
        Parallel execution, scheduler
        Speaker: Anatoli Fedynitch (DESY Zeuthen)
        200128_CKA8_scheduler_Anatoli.pdf
  • Tue, February 4
    • Discussion Tuesday
      interaction_histogram.pdf
      presentation.pdf
      spectra_mceq_crmc_2.pdf
      spectra_mceq_crmc.pdf

      Today the discussion mainly focused on the PID spectra. We compare:

      - frequency and type of interactions in EAS in C7, C8. Plans are to include CONEX.

      - PID energy spectra in EAS in C7, C8, CONEX, MCEq

      - PID event generator spectra SIB, QII in CRMC, MCEq

       

      One event generator level comparisons are in general good with differences probably linked to different handling of decays.

      -> run new simulations at higher statistics, and switching off decays. E.g. in CRMC there is an excess of pions in QII, but no rho0.

       

      Interactions in EAS indicate a real issue visible in Sigma0 in C7 and SIB2.3c. Even if fix needed, impact will remain small since Sigma0 will decay to Lambda, which then may also interact.

       

      PID energy spectra in EAS indicate major issues between all codes and particles. Neutrons are in particular strange. Max --> need to cross check those plots!

       

       

       

      • 3
        Discussion about PID spectra in EAS

        We need at least one low-energy hadron interaction model. Could be GHEISHA or URQMD (good for comparisons to CORSIKAv7) or HillasSplitting (comparing to AIRES).

        We need to test and finalize the mechanics to switch between high/low energy models for given projectile particles.

  • Wed, February 5
    • Discussion Wednesday
      DPG session Cosmic Rays III

      - Problem in C7 was identified in the logic to decay strange baryons. It is fixed now, but the impact has to be evaluated with some showers. Max has to update his study and plots with this.

      - Updated generator-level simulations for sibyll2.3c03 and qgsjetII04 were done (2e6 events) and decays of important short-lived particles switched off. Anatoli can compare this to his distributions and also send around new plots. Main question is how does this affect the pion spectra differences.

      - Need better documentation and guidelines how to work with git. Workflow. Lukas should provide a page in Wiki. Page is already created, but empty. Issue also exists.

      - Pythia dependence Issue should be closed. Problem in Sibyll decays should be understood and crashes in energy-losses for zero-length particles should be avoided. This has to be more robust.

      - PROPOSAL made significant updates. Moving their interfaces closer to how C8 is designed. This is very nice. Also work on physics was performed. The main restructuring and testing is almost finished.

      - photo-hadronic collisions are sampled, but not right now calculated. It would be very straightforward to do in C8. In C7 there are explicit code for the resonance region, and generator for the coninuum region. All this is basically low-energy. We may look into DPMJET or SMASH for this, or we use what C7 did.

      - maintenance of PROPOSAL and CORSIKA over the long run has to be optimized. We will have to converge as much infrastructure as possible in order to help. We should avoid to double implement complex logic.

      - We discussed experience with GPU in PROPOSAL. But studies are long time ago ~7 years. This has to be updated.

       

      • 4
        Discussion

        We need energy losses and multiple scattering for hadrons and muons.

      • 5
        Status of PROPOSAL integration
        Speaker: Jan Soedingrekso
        PROPOSAL_update_for_CORSIKA.pdf
  • Thu, February 6
  • Fri, February 7
Powered by Indico v3.3.5