Offline Leads 8/7/25 meeting notes

Attendees: 

Tracy Usher, Tom Junk, Giuseppe Cerati, Erica Snider, Katherine Lato

Agenda:

1. Spack V1 release
   1. Where we are for experiments, ready to migrate?
2. SimPhoton Change
3. Migration to Phlex
4. Geant4 and Genie Upgrade validations
5. Also – start discussing 2026 plans, scheduling meetings
6. Round table

1. Spack
   • Still working on verification. No show-stoppers yet. v1.0 has the features needed and requested, but still resists the model they have been trying to support. Until recently, were, trying to replicate some aspects from UPS, in particular with reusability of libraries. Spack currently doesn’t support this, so have abandoned this approach. The implication is that each experiment build will be different.
   • Using LArSoft as part of the verification process, so will have that available when done
   • Tom: What’s new in 1.0? I thought you could reuse stacks already in version 0.
     • Erica: There have been changes needed to recipes to accommodate differences. Able to use binaries from lower level things, but the stack from the different experiments will be different. Reusability seems more limited. Believe it has to do in part with re-writing rpaths
     • T: point of LArSoft to reuse?
     • E: Can reuse code, but there will be different builds. They’re still working on it.
   • E: Will have another meeting when Spack is released to talk about plans for experiment migrations and other details
2. SimPhoton Change – want people to be aware of change coming in.  https://github.com/LArSoft/larsim/pull/156. Pertains to change in rounding of times.

• • Didn’t think it warranted a LArSoft coordination meeting talk since it’s a small change. 
  • From Tracy:  it fixes a minor bug discovered while doing checks on the simulation timing. 
    • Basically, SimPhotons have their times rounded to an integer in simulation_time, which means any sub-nanosecond information is lost. As of right now, this is not vital to us: our single-photoelectron response is discretized at 2 ns, so we cluster SimPhotons into wide bins and notice no difference downstream.
    • It would become relevant only if we decided to use finer binning (0.5 ns), which is something Gianluca and I have discussed as a possibility… but with no specific plans to act. The affected variable is already a float, so no change in the data product size. The only adverse effect is the ROOT file compression, so that the file size increases by ~3%.
    • ICARUS advocates fo the change because:

• • • • File size increase is small;
      • If you wanted something light, you’d be using SimPhotonsLite already (where the rounding is done on purpose);
      • Variable already expects a float (ie, this feels like a copy-paste error from the SimPhotonsLite implementation);
      • Either way, this is not essential for us right now to resolve any of our MC issues.

• • • Erica:  If no concerns, we’ll add it to Approvals in Progress.

3) Migration to Phlex – no new information. Working in the background at a slow rate to see if we can get AI assistance with the refactoring needed to make migration easy. Pretty straight-forward coding, but can’t script it so AI might help.

4) Geant4 and Genie Upgrade validations 

• • We have been sitting on new versions of Geant4 and Genie for a while
  • They are blocking updates to art.
  •  Are people thinking of looking at these releases?
  • After the Spack conversion, there will be work to make the Genie releases independent of the LArSoft version, so that this is no longer an issue (aside from the connection to the art version)
  • Geant4 will remain a blocking issue for  further art updates. 

5) Start discussions with experiments on  2026 plans. Expect us to reach out scheduling meetings

6) Round Table:

• DUNE Tom: 
  • For DUNE things have been moving. Have a person working on Spack (V Hewes). Gave a presentation on plans. I couldn’t get the upstream to work, but was able to build the entire thing. Understand we’ll have to build more than we used to. 
  • Have a release manager, Wuyoung ?? (Wooyoung Jang?), looking for a way to fund. It’s on the priority list, but not the top priority. [@Tom, can you please clarify?]
  • LArSoft hasn’t released in five weeks and pull requests are piling up. We need to crank our releases out, this isn’t a LArSoft issue. Will LArSoft releases come regularly in the future?
    • Erica: As is the case now, new releases will be demand driven. At the moment, there is only one PR (from 3 weeks ago) that is approved and waiting to be integrated into a release, with 3 under discussion waiting for author responses, and 1 with approval in progress,, so it is not really a situation of PRs piling up on our end. That said, Lynn has been away a little bit, and  Patrick is deeply mired in work on Spack, so may not have the cycles to devote to weekly release building. If there’s an immediate need, however, we can get a release out. 
    • Tom: We can handle our own releases, just noticed that there hasn’t been a release in a while. We knew DUNE releases would outnumber LArSoft ones, so don’t need to bother you with that.
  • Geant4 doesn’t seem to depend on ROOT, but Genie does.
  • One of our users had a circular dependency and it builds, perhaps because we built everything. So possibly a cmake issue? It was in a DUNE PR, so not a LArSoft issue. Presumably, Spack will catch these things. 

• ICARUS: Tracy
  • Nothing pressing, busy pushing out ICARUS only. Started production on both data and Monte Carlo which will chew up resources. No one looking for anything new except bug fixes. Wirecell undergoing lots of changes as it gets stressed in new ways.

• SBN: Giuseppe
  • Experiment side eager to make the transition to Spack. Experiment work is already done. (Running the production code stack at ANL with LArSoft built under Spack)

Please email Katherine Lato or Erica Snider for any corrections or additions to these notes.