Framework is capable of running GausHitFinder algorithm in LArSoft

[knoepfel]

DUNE US S&C R&D item 072

[wddgit]

Below is a pdf of a few slides shown at a Phlex development meeting on 11/7/2025. It is just a list of issues that will need to be resolved to migrate the module GausHitFinder to be usable by Phlex.

GausHitFinder1.pdf

[wddgit]

Same as previous post, but instead of a PDF file I just pasted in the content of the slides:

Converting GausHitFinder

1. Uses Ptr and Association
2. beginJob books histograms that are filled (fold? or just delete?)
3. art::ServiceHandle<art::TFileService const> tfs;
4. art::ServiceHandle<geo::WireReadout const>() (needs geometry)
5. Throws exceptions (uses std::exception)
6. Configurable type and number of produced products
7. Contains TBB parallel_for nested in parallel_for (allowed?, optimal?)
8. Uses tools, e.g. art::make_tool<reco_tool::ICandidateHitFinder>(hitFinderToolParamSet);
9. Comments indicate some tool implementations not thread safe
10. Meghna’s work April 2024. Only a small part of what will need to be done. Also, behind by
    12 commits so would need merge.
    11.TH1::AddDirectory(kFALSE); (probably should just delete, not thread safe, why is it there)
11. I have not looked inside of the tools yet…

---

Converting GausHitFinder

1. Some parts can be refactored and separated. A lot of the code in
   GausHitFinder_module.cc there will need to be 2 versions, one for art and
   one for phlex…
2. Some lower level code should be fine as is, e.g. recob::Hit.
3. Depends on fhicl to get parameters
4. art::Globals::instance()->nthreads()
5. Inherits from art::SharedProducer
6. asyncart::InEvent(), not sure what that does
7. art::Event put and getByLabel functions
8. art::ProducesCollector produces is called
9. Uses art::ValidHandle

---

Converting GausHitFinder

1. Uses art::InputTag
2. There are sections of code in GausHitFinder that could be refactored into
   something art independent that would work in both art and phlex, but a
   significant fraction of it cannot be.
3. CMS has the concept of stream cache, run cache, lumi cache and some
   others. It’s complex, but these caching options seem useful.
4. From Meghna’s code changes. Here is an example I am puzzling over. A
   std::function was defined at a low level but is now a top level data member
   of class. Are we able to create data members at HOF level and pass them
   into algorithms?

[wddgit]

In the meeting we discussed the next things to do related to this topic and the decision was that the most productive way forward would be implement a version of GausHitFinder that will run in Phlex and leave out the parts of it that cannot yet be implemented in Phlex. For example, it will not be able to write Assn's because that part of Phlex isn't designed or implemented yet. For the present this will just live on a branch. Enough things will be missing that this will not be usable yet.