ROOT Meeting 26th June 2023
1. News
- new summer students — welcome everyone!
- Victoria — project storage layer in RNTuples with Jacob and Javi
- Silia — RDataFrame with Vincenzo
- Uri — SOFIE with Lorenzo
- Stefan — Graph NN with Lorenzo
- Daniel W. — benchmark of the statistical tools with Jonas
- Daniel A. — high performance computing with Jonas
- Axel - vacation starting next Tuesday (July 4th)
- summer student course on ROOT — Wednesday the 28th July, or July 27th
- PPP meeting is back! (this week - Thursday, 29th June, 4pm, email will be sent out soon, chairs: Vincenzo and Marta)
- topic for this week: new feature in distributed RDataFrame by Vincenzo
- we want it to become more regular - new developments, new design proposals, benchmarks etc.
2. Shift handover
- Olivier —> Marta
- nothing special happened last week
- forum ok, report on nightlies in MM, PRs and issues are assigned
- MacOS beta - brew installed - Olivier will look into this
- for the future reference: tag Axel for cling related questions
3. Meetings from last week
- I/O
- status updates — xrootd (remote data access) —> software penalises MT application over multi-process applications — fix from xrootd developers, temporary fix
- TMVA
- RooFit
- Planning / Godparents /...
- open bug reports from LHCb —> no reproducers available
- CMS —> ARM CPUs from the UK grids - new MT issues in the I/O, new ARM machine to reproduce the issues maybe needed? (Philipe and Jonas H. should look into this)
- LIM
4. Special Presentation —> Making web canvas the default [Sergey]
Discussion/Q&A:
- Jonas R.
- 0.7s to generate image - what was the time before? —> 1000s of images per second — degradation of the performance now if they do it sequentially in the web , so needs special API for these cases (shouldn’t be a big issue)
- Axel:
- chances not making it default for batch but only for interactive, are canvases different on screen and in pdf? — not much difference, unless you count every single pixel - then that may be different, but generally not a big difference
- web canvas vs. old canvas (I tweak things in web canvas, then I run in batch, then I want a pdf, will there be any differences) —> there should be no new features in web canvas that cannot be represented by the old TCanvas
- do they offer the same set of features? — old canvas — two new classes, small new features can be done in parallel, for the moment should not be any differences
- complex experiment examples — 500 line per macro — how close are these between the web canvas and TCanvas? — mainly the same, maybe some latex rendering would differ, but fill style, markers etc — the same
- TBrowser vs RBrowser (ROOT 7 browser, it will always use web canvas)
- when RBrowser was only in master (and not in release) — any feedback there? — not really, feedback after release,
- web default for interactive, TCanvas default for batch —> in master
- web for batch — new API needs to be discussed in the future
- Vincenzo:
- which features can be switched off and on and when?
- TWebCanvas — why this name? — implementation — TCanvasImp - implementation of the canvas, WebCanvas - class derived from TCanvasImp, user shouldn’t worry, automatically done with “root --web”
- for ROOT 7 — how much will have to be reworked, there is different development there - RCanvas
- re-use as much as possible from this implementation into the RCanvas
- feature compatibility: how many use cases would be broken by using the web canvas? - shouldn’t be worse than usual
- Jacob
- save file from web canvas — file dialog in web GUI - file saved on the remote folder on lxplus for example
- make it available for GUI power users — testers — activate with “—web” option
Conclusion:
- try in master but not in the release
Action Item:
- Sergey — mention this on the forum/announce/blog post — so people test it
5. Roundtable - skipped for today - tomorrow morning at 9am (27.06.23) on zoom
Roundtable 27.06.23
Olivier
- on shift last week - a lot of work on the forum
- now: work on mac mini
Bertrand
- windows problem with interpreter and visual studio
- lvlm patch
- will work on cmake
Axel
- work on CI
- cling and clang issues
- converge on open PR’s
- this week: help LHCb - their datataking problem
- patch release once back from vacation — first half of August
Vincenzo
- last week: concluded bunch of PRs
- work on RNtuple data with Grand Analysis Challenge
- preparing Summer Student projects and courses (new repo)
- met with a power user from Spain - detector physics experiments framework based on ROOT, maybe use RDataFrame there, they also want to organise school, maybe they will invite us too
Florine
- last week: support for the standard sets, using collection proxies — it almost works now, ready for the reviews soon
- start work on support for std::maps
Jolly
- last week: ported CUDA histograming to SYCL
- will look into GPU versions of de-compression
Jonas
- last week: patches to lvlm and clang
- yesterday: got rid of one of ugly cling patches
Marta
- summer student course preparation
- work on progress bar
- discussed with Enrico and Vincenzo about the physics side of the refactored Grand Analysis Challenge
There are minutes attached to this event.
Show them.