Present/Contributors: Graeme Stewart, Liz Sexton-Kennedy, Andrea Valassi, Efe Yazgan, Kyle Knoepfel, Witek Pokorski, Eduardo Rodrigues, Michel Jouvin, Andrea Rizzi, Horst Severini, Teng Jian Khoo, Stefan Roiser, Caterina Doglioni (for a bit), Serhan Mete, Benedikt Hegner, David Lange, Gloria Corti, Kilian Lieret, Pere Mato, Sam Meehan, Sudhir Malik, Daniel Elvira
Apologies: Agnieszka Dziurda
Current Draft:
No additional comments, so this is approved.
This meeting is now set for July 22: https://indico.cern.ch/event/930127/.
July 23, 24 + 27, 28: https://indico.in2p3.fr/event/21698/overview. We were invited to give a 20’ talk on Friday 24 (CERN/HSF/HEP approach to software/code or workflow optimisation), any volunteers? Contact Graeme for details.
We received a first draft of the report from the panel. Largely it was very positive (which is great for the HSF as this is the first time that we have really been reviewed). There was a feeling that some important points have been missed, which were communicated back to Amber.
Andrea Rizzi has been nominated CMS deputy physics coordinator. Congratulations! It means Andrea will not be able to stay as an Analysis WG convenor. Need to start the process for his replacement, to start Sept. 1st. We would have the appointment run until end of 2021.
Discuss with them the format of the training material and possible conversion to Carpentry style (Stefan).
Policy changed: Web_Site_Managers can no longer push directly to master; need at least 1 approval on a PR.
BUT: Few people comment on PRs (even when asked), which is slowing things down a lot.
Possible solutions:
Agreed that we want to keep things running smoothly for people making useful contributions
master branch remains protectedBenedikt volunteered to take a look at these issues and implement some solutions.
The meeting was positive about this for WG pages, so we try to implement that for all WGs.
Announced the next meeting for July 15th. The topic is chiefly B-physics, with some input also on “10%” analyses (i.e. those not fitting the central processing model because of special reconstruction and data handling).
Some topics we’d like to arrange events on:
Held the 3rd topical meeting devoted to use of GPUs on the 24th of June. Included the presentation by the GATE project and 3 lightning talks. Overall, a very interesting set of talks. The general picture one can see is that in order to gain reasonable speedups we need to move substantial (almost all) parts of the computations to the GPU (otherwise the overheads kill the speedup). On the other hand, moving general transport code seems to be very difficult, so one needs to focus on specific problems like optical photos, low energy neutrons, EM shower, etc, complex enough to gain from the use of GPUs but in the same time simple enough to be ‘portable’ to the device.
Organized the Allen meeting + attended a number of seminars and workshops about GPU in trigger and reconstruction. Not much news apart from that for the time being, no meetings in July but will resume in August/September.
PyHEP 2020 workshop agenda finalised and online - a very interesting set of presentations!
We still aim to release a pre-workshop questionnaire as usually done.
As advertised before, a new tag of HSF/prmon is imminent. Among many other improvements, it’ll be the first version that supports NVIDIA GPU monitoring.
Will probably have a meeting next Thursday.
At yesterday’s WG meeting, Ben Wynne from ATLAS presented details on the Athena multithreading scheduler. The meeting was attended by ~20 individuals; there was good discussion.
Still planning on a presentation from LHCb’s Allen framework, possibly as early as next week.
There’s a PR to setup the activity from Nathan, to be reviewed.
Graeme was contacted by FASER to ask about licensing and passed on the HSF’s advice. (CERN experiment, so relatively easy case, following the LHC Experiment models,)
Very positive that we get these spontaneous contacts.
Reminder that the proposed September workshop in Lund is again postponed. Next possible event is another virtual workshop in November.
Stefan has deployed some scripts that manage the synchronization of the HSF Indico category with the HSF Community Calendar (thank you!). This is now working and deployed.
So any (public) events created in Indico will generate an entry then created in the HSF Calendar within 15 minutes (cron job time!).
The script is smart: if you update or cancel a meeting it will change or disappear accordingly.
This removes an extremely annoying manual step to get things into the HSF calendar. However, do be aware that anything created in Indico it will appear in the calendar more publicly than if it was just lurking. (You can set the visibility of events in Indico on calendar feeds though.)
The scripts will work against any Indico instance.