Key4hep Discussion

→ Europe/Zurich
Zoom

Zoom

Zoom Meeting ID
95444210901
Host
Andre Sailer
Useful links
Join via phone
Zoom URL

# Key4hep Live Notes

Date: Nov 3, 2026

Agenda: https://indico.cern.ch/event/1748780/

Connected: Andre, Fangyi, Joshua, Mahmoud, Andreas, Thomas, Scott, Juraj, Sanghyun
Apologies: Brieuc, Juan

## Github project

- https://github.com/issues?q=is%3Aopen+is%3Aissue+user%3Akey4hep

## Discussions
[Github Project Discussions](https://github.com/orgs/key4hep/projects/2/views/1?visibleFields=%5B%22Repository%22%2C%22Title%22%2C3970156%2C%22Status%22%2C%22Assignees%22%5D)

## Presentations

https://github.com/orgs/key4hep/projects/4/views/1

### Reconstruction workshop recap
* How to do short term detector development samples?
    * Decided on having a sort of "running train model", i.e. on a somewhat fixed schedule tag everything that is ready and build a stack from there.
* Vertexing discussion (w/ Taikan)
    * Related to https://github.com/key4hep/k4Reco/pull/63
    * Proposal to have a core repository with shared resources for LCFIPlus and its Gaudi port. common repository would then be referenced from (existing) LCFIPlus and the part that is referenced from Gaudi.
    * Follow up discussion in Nov 26 Reconstruction meeting (time TBC)
    

## Open PRs / Overview

Open PRs: https://fcc-prs.github.io/
Merged PRs: https://fcc-prs.github.io/merged.html

### Initialization time for certain digitizers reported by Armin & Giovanni
* Found via k4Bench
* Traced back to DD4hep Surface Manager: https://github.com/AIDASoft/DD4hep/pull/1704

### RPATH issue related to setting `CMAKE_INSTALL_LIBDIR`
* See https://github.com/key4hep/k4geo/issues/677
* [x] Fix in k4-project-template as well -> https://github.com/key4hep/k4-project-template/pull/34

## Potential bug in LCFIPlus vertexing
* https://github.com/lcfiplus/LCFIPlus/pull/80 **merged** (and ported to https://github.com/key4hep/k4Reco/pull/63)

## Release
* Tagging most packages this week unless there are comments
    * Tagged (today): 
        - LCFIPlus (v00-11-02)
        - MarlinReco (v01-38-02)
        - MarlinTrkProcessors (v02-12-11)
    * Only PRs from k4RecTracker from Andrea blocking?
    * k4Gen PRs (from David & Joshua)?
    * DD4hep needs a new tag
    * podio could be tagged (but would like to have some of the open PRs in first)
    * EDM4hep?
* Not yet tried to use the "tagged packages" stack
    * Would need to be registered in iLCDirac for production usage
    
## Open items from last time
* [x] Move `k4RecCalorimeter` from `HEP-FCC` to `key4hep` -> https://github.com/key4hep/k4RecCalorimeter
* [x] Need to make sure that `k4-project-template` ships the current defaults for `.clang-format`, etc. -> **Yes**
* [x] Figure out who participant from Rome / Sapienza is (she put down the workshop list as guarantor for the CERN hostel)
    * No response from direct mail to her
    * Tell CERN hostel again, that we don't know her
* [x] Which version of LCFIPlus do we currently build in the nightlies ->latest commit
* [x] Estimate the effort that would be necessary to have only the primary and secondary vertexing ported to k4Reco (without taking the LCFIPlus internal EDM & machinery) -> discuss at dedicated discussion in reco meeting
* [x] Organizing github repositories and creating teams for easier management of access rights to different repositories

### Organisation
* Probably need a mixture of teams and single people for some repositories
* Follow the PS&C organigram?
* Assign k4DetectorPerformance PRs to high level reco team
* Discuss in more depth at in-person meeting

## No nightly on Ubuntu 24 any more
* https://github.com/key4hep/key4hep-spack/issues/945
* who still uses these?
* Most people seem to be on lxplus using the Alma9 nightlies.
    * No one aware of using Ubuntu nightlies (regardless of 24 or 26)

## AOB

### Work started on a general detector optimization framework
* https://github.com/key4hep/k4Opt
* General idea: Have a way to define and identify a detector optimization problem that is also reproducible
    * could be useful also to have a sort of *database* of things that have been tried
* Uses Optuna to optimise
* Can implement evaluators at any level (simulation, reconstruction, etc)
    * Needs to be seen how computing resources deal with that and at which stages it makes sense
* Flexible framework that can be used to tune (almost) anything
* Should be possible to connect k4DetectorPerformance to this and use evaluators / metrics from there.
* Similar effort (conceptually) exists for Pandora optimization in ALLEGRO
    * See: https://indico.cern.ch/event/1694161/contributions/7143590/attachments/3302408/5907773/BBO_FCC_fullsim_Mieskolainen_24062026.pdf
* Definition of actually interesting metrics still needs to happen and be discussed


## Next meetings

* Nov 10, 2026, 09:00 CET

There are minutes attached to this event. Show them.
    • 16:00 → 16:10
      Introduction 10m
    • 16:10 → 17:00
      Discussion 50m
      Speaker: Everyone