HSF Weekly Meeting #167, 6 June, 2019
Present/Contributors: Graeme Stewart, Pere Mato, Witek Pokorski,
Daniel Elvira, Andrea Rizzi, Michel Jouvin, Martin Ritter, Paul Laycock,
Gloria Corti, Caterina Doglioni
Apologies: Agnieszka Dziurda
News, general matters
- LHCC this week, Graeme gave the software presentation update:
- Thanks to everyone for comments and suggestions.
- Feedback during the meeting:
- Generators - emphasise to the generator developers that
negative event weights have a huge adverse effect (are
there ways to minimise this by tackling the phase space
- A. Pass this message on via EG WG.
- ML Simulation of calorimeters - will the work be portable
to different detectors?
- A. We understand this is the intention of the
OpenLab/SFT efforts, but not yet there; for future
- Accelerators - do people develop the code in portable
ways, avoiding tuning to a particular GPU architecture?
- A. This is an issue people are aware of - try to
separate physics code from device code. R&D starting
on abstraction layers to insulate us a bit more. (CMS
showed some work with Kokkos and Alpaka that looked
- Referees’ Recommendations (with emphasis for software issues):
- 2021 requirements should be firmed up, based on the latest
understanding of the likely machine performance.
- A timetable for Phase-2 computing model reviews should now
- Discussions with funding agencies should continue,
regarding the provision of resources when future cost
estimates are very uncertain.
- The discussion around the use of HPC should continue.
How can we adapt to use HPC resources and/or have more
influence over how they are setup?
- Work on hardware acceleration (GPU, FPGA) holds the
promise of a big leap in performance. This work should be
encouraged and coordinated across experiments.
- Simulation on accelerators caught people’s eye for sure. We have to
take this seriously and do R&D, but not allow inflated
expectations. x100 speed-ups are not going to come and the costs
of the re-coding will be enormous. We should discuss
this seriously because HSF can help to coordinate answers here.
Google Summer of Code 2019
- Coding period has started.
Google Season of Docs
Activity and Working Group Updates
- Nothing particular to report. Discussions ongoing in multiple
threads on the google groups
- Topical meeting next
week on geometry.
- DD4hep, Geometry optimisation in ATLAS, VecGeom.
- There is a topic of CAD->GDML tool, but probably no time for it
- Foresee another geometry meeting after the summer, but with a
different focus, e.g., visualisation of geometry.
Reconstruction and Software Triggers
- Yesterday’s topical meeting:
- Topic: Algorithmic approaches and data structures to efficiently
exploit accelerators and many-core architectures
- LHCb presented their experience: reached the milestone of 30MHz
after SoA/SIMD event model changes.
- Meeting very well attended ~50 people.
- Presentations very well received - a lively discussion (meeting
- Request to follow up SoA/AoS with ATLAS/CMS
- Live notes:
(still needs to be filled and cleaned in some places)
- To-do: need to add indico/lives notes to the website
- Organising meeting following up on potential meeting topics with
- New training Indico category created at HSF top level:
- Need to follow up on possible training events at CERN.
- Meeting 29 May, minutes available:
- Next meeting planned for 3 July.
- Reminder that call for nominations is open until tomorrow.
- So far 10 nominations received.
- Some people who are HSF Coordinators, so we will establish a
search committee again to propose from the nominees.
Pre-CHEP (2-3 November)
- Organising meeting happened this week:
created to track decision and progress.
- Proposed title: Analysis Systems: From Future Facilities to Final
- Main high-level items for the HSF and DOMA sides identified: we
have a good list!
- Still have to group them into sessions: goal is to mix HSF and
DOMA topics in each session so that the full audience is
motivated to attend the whole event.
- Drafting a description of the workshop for the next CHEP bulletin
planned before the end of June.
- Will also start to fill blocks in Indico and refine them as
soon as we progress with session definitions.
- Saturday will be 1-6pm, Sunday will be 9am - 1pm.
Next HSF/WLCG Workshop
- We are making a formal call now for interest in hosting this
event. Parameters are:
- Facilities for 1 whole week, in May or June 2020
- Monday morning -> Friday lunchtime
- Known weeks to try to avoid so far: w/b 18 May (Ascension Holiday);
w/b 1 June (LHCC); w/b 22 June (ATLAS Week)
- 1 plenary room, capacity min 250
- 2 additional parallel rooms, capacity min 100 each
- Vidyo connection in these 3 rooms for the week
- Optional breakout rooms ~3x30 people would be nice (vidyo
- Low costs favoured (€200 ideal, €300 acceptable)
- Lunch options to be specified - on site and included in
- One reception event on Monday, costs included; one dinner,
could be billed separately if necessary
- Some indications of typical local accommodation costs and
- A European location is favoured this time
- Please express interest by 12 July to
- We will need a small organising team for HSF.
- ECFA special session at EPS-HEP on “Toward the Update of the
European Particle Physics Strategy” (13 July).
- Graeme invited to give the talk on Computing and Software
- AIDA++ follow up call for expressions of interest has been made.
There is a f2f meeting at CERN on 17-18 June:
- Contact Witek, Frank and Graeme to make an EoI (due beginning
of July) and plan to attend that meeting if possible as it
will help to build a consistent program.
- Note, you can attend the meeting even if you do not have an EoI
to present. General input from the community is important to
define the right programme.
- Having coordination meetings every two weeks seems to be a good
rhythm at the moment.
- Should we formalise this and adjust the meeting planning from
- It frees this slot for any WG that wants it.
- We can always have an additional “odd” coordination meeting
again as needed.
- Agreed that we try this cadence for now.
- -> Next meeting 20 June.