HSF Weekly Meeting #171, 12 September, 2019
Present/Contributors: Graeme Stewart, Serhan Mete,
Concezio Bozzi, Witek Pokorski, Stefan Roiser, Eduardo Rodrigues,
Caterina Doglioni, Michel Jouvin, Agnieszka Dziurda, Andrea Valassi,
Apologies: Pere Mato
News, general matters
- LHCC Referees Presentation summarising software
matters was given on Tuesday:
- Questions from referees about C++ training (is
it too late?).
- Closed session quickly touched on “modern”
languages. Need for C++ training was stressed again, as
performance is critical to us.
- EP R&D on Experimental Technologies:
- This R&D activity has been funded by CERN,
including a software work package
- Four projects will be supported:
- Turnkey Software Stacks (optimised and
integrated stack aimed at detector developments).
- Reconstruction at High Pileup (trackers
and high granularity calorimeters).
- Faster Simulation (machine learning
algorithms for speed-up).
- Analysis Facilities (new ROOT RDataFrame
and object store integration).
- This work will ramp up in 2020 and will run
until at least 2024.
- If you are aware of people who might like to
work on these projects, as CERN Fellows, please let Graeme
Google Summer of Code 2019
- It’s done! 29 out of 33 students successfully
completed their projects.
- Well done to mentors, students and, most of all, our coordinators: Javier, Andre and Antoine.
Activity and Working Group Updates
- Not a lot of news yet - starting to think about
topical meetings for the next few months.
Reconstruction and Software Triggers
- Last week 2nd session with cross-talks:
- Next meetings planned for October.
- We’re going to have a seminar on
Tuning and Analysis
(TAU) from Sameer Shende (Director of the Performance Research
Lab, NIC, University of Oregon) on Monday October 14th at 5pm
- Room: 40/S2-D01
- The idea is to have an installation of TAU on
CVMFS so that people can play with conveniently. How best to
- Depending on the feedback, can have a hands-on
tutorial when Sameer is in Europe (December or later).
- Serhan and Graeme will discuss with Jakob the
/cvmfs/hsf.org setup to install the software.
- First topical WG meeting on fitting tools took
place yesterday, see agenda
- Good discussions at the meeting and continuing
in the Gitter channel.
- A second meeting will probably happen later in
- PyHEP 2019 workshop
- Agenda basically finalised at this
- Registration deadline is approaching very
- There are just over 50 participants.
- Sponsorship from the Python Software
Foundation granted, and highly appreciated!
- Planning Software Carpentry event at CERN, 27-29 November.
- Rooms are booked!
- We are now looking for people to be helpers
for the students - requirement is intermediate level skills in
Python, git, shell
- We need:
- A poster to raise the profile (anyone good
- Decide the final curriculum and find the
- Slow restart after the summer, no meetings
- Met Wednesday 11 September… good chat about Spack
based on the work that Hobbs Willet (CERN Summer Student)
- Next meeting planned for 9 October,
- New convenors appointed:
- Chris Jones, FNAL, CMS.
- Kyle Knoepfel, FNAL, Neutrino
- Attila Krasznahorkay, CERN, ATLAS.
- Had a discussion meeting yesterday to plan first
- Idea is to have a couple of initial meetings
by CHEP to build the WG community.
- Liz, Graeme and Michel met and decided it was not
worth pursuing with FastJet developers in the short term, not very
collaborative on the topic and not really available. The
perspective remains a new jet clustering framework optimized for
parallelism (US project) : details not available yet but will be
done with proper credit to algorithm contributors.
- Problem is FastJet license is GPL, but ATLAS
and CMS want to have their offline code under an Apache2
- Andrea: it would be best to inform Fastjet
developers well in advance that a US group is considering
rewriting a Fastjet-like software using their ideas. We must
be extra careful not to alienate people (especially from the
theory community) even more. Andrea will send a message to Liz
- Andrea - HEPiX benchmarking WG would like to know
what the licensing is for packaged experiment software.
- To be checked but not necessarily a problem if
only the binaries are distributed
- The people following this in the HEPiX WG are
Michele Michelotto and Domenico Giordano. Andrea will tell
them to get in contact with Michel and Graeme (thanks to both
for volunteering to help!).
Next HSF/WLCG Workshop
- Three bids: DESY, Sweden, Romania.
- Bid documents are attached to the
- Swedish bid has in fact 2 locations (2
- DESY only possible May 4-8
- Dates highly constrained: May 4-8 or May
- Short presentations at the GDB on
- Open to responses now, or you can contact Graeme, Michel and David L. until early next week, or email email@example.com.
Pre-CHEP, Analysis Systems: From Future Facilities to Final Plots (2-3 November)
- Planning now well advanced.
- Saturday we have a series of talks to inspire
- Run 3 Analysis Challenge.
- Run 4 / HL-LHC Analysis Challenge.
- 2 talks about facilities, including
IRIS-HEP work around DOMA (confirmed: X. Espinal).
- Analysis workflow.
- Machine learning (as part of the
- Analysis preservation and its impact on
analysis workflow (confirmed: Lukas Heinrich).
- Sunday we plan interactive sessions around 4
- Facilities - how to organise for
- Interactive Analysis - what are the needs
- Machine Learning - how to manage training
and inference; what is the scale of training and who does
- Analysis Model - what are viable models
for the future? Key points to improve or change.
- We have a very large level of participation, 160
people (makes Sunday a challenge to organise, but we think we can
Latin American Workshop on Software and Computing Challenges in HEP (Mexico, Nov 20-23)
- Some progress in getting some speakers (e.g.
Gordon Watts). Looking for another IRIS-HEP speaker.
- Will contact HSF WG convenors as well.
- Still looking for sponsorship.
- Registration is open. Ask HSF Coordinators to
circulate information to their experiments.
- Stefan Roiser will go to the UK RSE conference
next week, presenting software institute ideas.