HSF Weekly Meeting #152, 29 November, 2018
Present/Contributors: Graeme Stewart, Pere Mato, Serhan Mate, Witek
Pokorski, Andrea Valassi, David Lange, Jim Amundson, Torre Wenaus, Ian
Collier, Danilo Piparo, Daniel Elvira, Dario Menasce, Tommaso Boccali,
Gloria Corti, Paul Laycock, Frank, Caterina Doglioni, Agnieszka Dziurda,
Ed Moyse
News, general matters
- IRIS-HEP would like to have HSF representation on their steering
committee.
- Pete Elmer (PI) has proposed that Graeme takes this on.
- Fine with Graeme and got support from others.
- Agreed.
- Gerri Ganis gave the software report to the LHCC
referees this week.
- Questions on licenses and CWP review comments.
- Graeme will give a short report on the new working groups and the
generators workshop tomorrow at the WLCG Overview
Board.
HSF/WLCG/OSG Workshop
- Meeting took place on Monday 26th to discuss the workshop
timetabling.
- Very constructive and we were able to arrive at good range of
substantially shared topics, while also earmarking necessary
separate sessions for WLCG/HSF/OSG matters respectively.
- Draft block timetable being populated here:
- Broad division is 2 days for plenaries with WLCG/OSG, 2 days for
HSF parallel sessions.
- Meeting on Thursday 29th to discuss local organisation matters
(immediately after this one).
- Aim today is to fix logistical matters so we can open up
registration well before the holidays.
New Working Groups
- Convenors have been appointed - congratulations to them and thanks
to all who took part.
- Detector Simulation: Heather Gray, Witek Pokorski, Gloria Corti
- Reconstruction and Software Trigger: David Lange, Caterina
Doglioni, Agnieszka Dziurda
- Data Analysis: Andrea Rizzi, Danilo Piparo, Paul Laycock
- Starting the work of the groups
- Update website (please see
instructions)
- Planning for an activity start next year
- HSF “owns” a slot every 2 weeks in 32-1-A24, 17h (odd
weeks):
- 16, 30 January; 13, 27 February
- Sometimes we give over the coordination slot to a topical
meeting, but this room (32-S-C22) is not very good (too
small, no vidyo).
- Pere - equip this room better? Yes please!
- At least one workshop pre-meeting?
- Start to organise a parallel session at the workshop:
- The main theme could be the survey of the existing
activities and planned ones in the community.
- Short plenary from each group?
- Contact experiments.
- Mailing lists:
- Convenors should meet together soon.
- Graeme will circulate other names and inputs that we got during
the process.
- Short report on plans from each working group in 2 weeks time
(December 13).
Generators Software Computing Workshop
- Andrea: Physics Event Generator Computing Workshop took place at
CERN this week
(https://indico.cern.ch/event/751693/timetable/).
Successful workshop, useful discussion and nice atmosphere with a
mix of theorists/experimentalists and a few engineers.
- Many thanks to all the organisers! And to speakers and
participants!
- Talks and discussion Mon-Tue and hackathon on Wed.
- Good attendance Mon-Tue (56 people registered), around 30-40
people in person plus 15-20 via vidyo. Smaller attendance
(less than 10) at the hackathon, but was also useful.
- Live
notes
have been taken (by tasked usual suspects, but not only). Some
take home messages are at the end.
- We will have a post-mortem probably next week to discuss next
steps. A few people (Andrea, Steve) have already agreed to
push on work in this area. We might have a dedicated 5-day
practical workshop in 2019?
- Graeme will also give a WLCG Overview
Board report tomorrow.
We forwarded Federico’s invitation to ACAT to generators
people. We might also give a summary there (Andy Buckley will
be there).
- Some of the take home messages and further work planned:
- Need to continue investigating ATLAS vs CMS differences.
Includes Madgraph vs Sherpa but not only. Some comparison
numbers were presented by Josh in the ATLAS talk. (Note:
LHCb generators mainly concerned with decays rather than
production. ALICE were contacted but did not take part in
the workshop).
- Need to better understand Run4 projected needs and possible
model. Clear that LO can do unweighted events ok, NLO can
do unweighted events now but only with negative weights
(large penalty as many more events must be generated).
NNLO for the moment is only for differential cross
sections, no unweighted events, but 1-2 orders of
magnitude slower for LO->NLO->NNLO is to be
expected.
- Need some metric to evaluate impact of negative weights,
in order to best compare generators.
- The need to coordinate work and reward system between
experiments and theorists was pointed out. Example in the
Herwig talk “improving code performance would be a lot of
work for no recognition”. Also, how to integrate
engineers/experimentalists work in theorists’ work?
ATLAS/CMS started investigation on recognition for work in
generators area.
- Experiment computing infrastructure may help theorists.
Example: regression tests for timing and physics
performance across code versions. Example: large scale
resources for specific NNLO resources.
- Some reports on HPC and GPU work, but not that much work
there.
- David: in the 1.5h HPC session GPUs were not even
mentioned.
- In general the HPC work did not show a real need for
HPC, this work could have been done also on HTCs.
- GPUs only presented by Madgraph and it was mentioned
that at the time they thought the experiments were not
interested (or not able to use GPUs). The work was
done around 5 years ago by authors that maybe moved
elsewhere, but the port should still be functional in
principle.
- Some more technical points, e.g. push for common interfaces
to apply generator level cuts, review of Les Houches
format etc.
- Licensing was mentioned in David’s talk but we kept it
intentionally as a side subject. It may be useful to bring
this up again in the future though.
Activity updates
Licensing
- Matteo Cacciari was not able to come to CERN, so we are exploring
dates early next year to follow up on this. Licensing was
raised at the generators workshop and so we may broaden the
discussion to the whole theory community.
Packaging
- Meeting 21 November,
focusing on Portage tool.
Minutes
available.
- Next meeting planned for 5 December
Software Development
- Meeting on 6 December on
monitoring tools - Servesh will discuss his new Trident tool.
N.B. this will replace the usual coordination meeting next
week.
Software Forum
- Discussion with the Electron Ion Collider Software
Consortium, 21 November.
- Some interesting topics to follow up on (data formats, analysis
workflow) - we could foresee further interactions at the JLab
workshop.
- Next meeting on Spark has
been expanded to also include a talk from ROOT on Spark - should
prove to be fruitful, and we’ll try not to run on too long!
CWP
General Matters and Roadmap
- CWP Roadmap was resubmitted to CSBS.
- Waiting for news from Springer.
Publication status for Individual WG Papers
- Data Organisation, Management and Access
- Ready to resubmit to arXiv.
- Visualization
- Event/Data Processing Frameworks
- To go to final circulation.
- Facilities and Distributed Computing
- Conditions Access
- Paul did a significant revision of the document, hopes to
circulate it this weekend.
AOB
- Update: Should we submit a HSF related paper to the European
Strategy Update?
- Pere’s proposal : write a couple of pages around importance of
taking software & computing into consideration early in the
process of designing new machines/detectors and refer to the
CWP for details.
- Pere discussed with Fabiola at the EP faculty meeting and she
did think this was a good idea.
- Graeme, Pere to work on a draft for next week.