Choose timezone
Your profile timezone:
ROOT Team Meeting
ROOT Team Meeting
Zoom Meeting ID
97374667082
Host
Axel Naumann
Alternative hosts
Bertrand Bellenot, Lorenzo Moneta, Danilo Piparo, Enrico Guiraud, Jakob Blomer, Vincenzo Eduardo Padulano
Useful links
Join via phone
Zoom URL
News
Strategic goals
Vassil: I have some comments I will share offline, some things do not read well to me. A lot of time is spend (e.g. when writing proposals) on "deciphering" strategic goals. We need to make clear what the words map to.
6.34 Release
We will know in a few days what CERN mgmt will do with Run 3 (i.e. if they will extend). The date of 6.36 will depend on this.
For 6.34: deadline for changes was missed due to LLVM18 upgrade. Initially the issues were (seemingly) fixed, but CMS builds on ARM64 broke (almost) everything JIT-related. Upgrade is still within reach, but work is needed.
We were blind to this because we don't have ARM in our own builds.
Priority items from CMS
4 high-priority items, need to be addressed.
Web Graphics
We will do a mini-workshop to decide if we want to enable it by default for the next release (and what is needed for this to be possible). Meeting on 11/09 (time/place TBC), follow-up with Serguei the week after.
PoW discussion
Wed 25/09, 15h-18h. Same format as last year (jamboree).
Jakob: How flexible are we with the date for this?
Danilo: Let's discuss offline, we can maybe change.
Proposed speakers can be found in the slides. If in doubt (regarding the assigned topic), speak to Danilo.
Jack: What is "infrastructure"?
Danilo: CMake/build system and the CI. This means talking to me and Bertrand.
Vincenzo: We should also consider packaging, i.e. Conda, pip, Docker, etc.
Stephan: What if Jakob has an idea that he wants to add to the PoW for graphics? Does he have to tell me beforehand or during the session?
Danilo: This has to be always possible. We can formalize this, but in principal both are welcome options.
Vision for ROOT 7
ROOT components table
Based on feedback of many peopl, first proposal. Basically, shows the components up for removal/deprecation/legacy/discussion ("under review"). Note that "legacy" is loosely defined and has different levels of priority (e.g. TTree vs TSpectrum).
Table right now is taken for 6.34 (but will be adapted according to upcoming release). The first 4 points are very uncontroversial.
Stephan: for legacy/depracated/remove, perhaps make it a dropdown since they are mutually exclusive.
Jack: If something ends up in legacy, does that mean it cannot be deprecated anymore?
Danilo: Didn't think this through yet. I can see scenarios where this might happen.
Vicenzo: TLorentzVector should be added to the table as well. It's marked as deprecated in the docs and has a replacement (ROOT::Math::LorentzVector) but many people still use is.
Lorenzo: There are many more classes we can put in this table.
Danilo: This is a start. We need something to collect all of them.
Forthcoming events
- Q3 report. Let Danilo know if you want to discuss anything non-standard here.
- CHEP rehearsals (7/8/9 oct).
- SFT Retreat (1 nov).
- PoW Jamboree (25 sep).
- 2nd Hackathon (25/26 nov, potentially additional day).
- Jakob: I was wondering if we could "glue" the RNTuple metadata workshop/discussion to this event, potentially with some additional time or just focus groups.
Shift handover
From Monica to Olivier.
The only issue is with Win x86 Debug. Failures due to rebuilds during testing and timeouts. Have been disabled by Danilo. Both for 6.30 and Master.
For the rest, everything green, assigned, lots of activity on the forum.
Meeting summaries and plans
Very small I/O meeting, nothing worthy to share.
There is a new graduate starting for NGT (ML-related). Staff in November.
RooFit: smaller meeting off the normal schedule due to vacations, some power users from ATLAS. Robin presented new developments. Mostly talked about restructuring tutorials. Still some blockers on the ATLAS Higgs Factory workspaces.
Topics
Both topics were already discussed during "News"
Round table
Actions: table is done, Olivier proposed PR.
Vincenzo's PyHEP not-trip report
See slides for complete information. Note that they are objective, i.e. only report on what was discussed and the general sentiment of the discussions. Vincenzo also has extended notes.
Tuesday: Fitting (AM), ML and fitting (PM)
JonasR: What does it mean for the future of iminuit?
Vincenzo: They are happy to use it and keep using it, but also want to improve support an credibility of the other minimizers.
Long discussion on HS3, and how it works together with RooWorkspace (i.e. have them live together seperately or somehow make RooWorkspace also serializable). No clear path forward. Motivation for the discussion is the expectation/hope that serialization will inspire cross-experiment knowledge sharing.
Stephan: They are leaving one thing out. RooWorkspace is serializable, within ROOT.
Vincenzo: The reason for reporting this is mainly to understand the participant's point of view and way of thinking.
Ingestion of ROOT data for ML is considered "solved", but still relies heavily on workarounds (i.e. converting to other formats).
Vassil: people want to use whatever they can use with Python and that ML ecosystem.
Wednesday: workflow management (AM), histogramming (PM)
Workflow management is not coupled with ROOT whatsoever. Tools to help you describe and orchestrate your workflow. Everyone uses luigi. It would mostly be nice to know how we could hook ROOT into such a workflow (student project?). There are no examples currently. Not urgent whatsoever.
Histogram serialization: boost.histogram cannot be serialized (but is coming), so has to be converted to e.g. ROOT before storing to disk. Questions whether we would support serializing boost histograms.
Thursday: Analysis at scale (AM)
Question from Vincenzo: are we (the community) happy with Dask? Are we fine by locking ourselves into this solution? Many opinions but everyone agrees that this is current reality.
High-level comments on the atmosphere: more more open, less hostile than it used to be. People were interested in what we are doing and planning, als wrt ROOT 7. Overall, the discussions were high-level and on-point. Clearly biased (because of audience expertise), but not in a hostile way.
JonasR: Was there any discussion on Cppyy?
Vincenzo: None, it is not considered industry standard and therefore not considered a solution.
Half was IRIS-HEP, half from German institutions (~30 people in total).
Vassil: How to build benchmarks that are representative to problems in our field, and use them to compare different "community standards" and decide objectively what solutions to use.
Vincenzo: AGC goes a step in the right direction.
Jakob: There are the Analysis Description Benchmarks as well.
Vincenzo: They were not discussed during the workshop. Something I did not mention before was that some portion of the morning was also dedicated to Julia.
There will be a full, proper (less detailed) trip report during next Monday's SFT meeting.
There are minutes attached to this event.
Show them.
