Present/Contributors: Graeme Stewart, Paolo Calafiura, Serhan Mete, Attila Krasznahorkay, Caterina Doglioni, Jose Benito Gonzalez, Kyle Knoepfel, Efe Yazgan, Eduardo Rodrigues, Paul Laycock, Witek Pokorski, Horst Severini, Michel Jouvin, Andrea Valassi, Stefan Roiser, Agnieszka Dziurda, Ben Morgan, Bernhard Gruber, Chris Jones, Gloria Corti, Liz Sexton-Kennedy, Paolo Calafiura, Pere Mato, Philippe Canal, Sam Meehan, Stefan Roiser, Teng Jian Khoo, Savannah Thais, Andrea Rizzi
Our CERN colleagues (Jose Gonzalez, Tim Smith) asked for a letter of support for their EuSSI proposal, a H2020 call. The project abstract is here and they aim at building a Europe-wide initiative for Research Software with the goal of software becoming a first-class citizen in research. The initiative will provide guidelines for good software sustainability practices and reproducible science and integrate with many of the similar existing local initiatives (NLeSC, SSI, etc.).
This looks very well aligned with the HSF and we have prepared a draft letter of support.
Q. What about software standards? What constitutes “well written software”? A. Yes, there is a work package on sustainability and quality. There will be a body that tries to improve this area and also training is a significant work package.
The meeting approved the letter of support for EuSSI.
Paolo and Savannah (Exa-TrkX) proposed creating a “project-neutral” repository and website to collect R&D ideas and results on graph algorithms for tracking.
A possibility would be to have a GraphTracking (name TBC) repository under the HSF umbrella. The repository would collect code/results/datasets in areas such as Graph NNs (including trigger apps, and distributed training/inference), metric/similarity learning, clustering/partitioning, and global optimization methods like annealing. Traditional graph algorithms (e.g. min-cut and maxflow) would also be in scope.
Caterina - this is a good idea. How would the HSF repo relate to other repositories? Would need a gatekeeper process (Reco WG coordinators?). Can live side by side with other repos. Can have an open list in the Reco WG page of established connections to different projects. (Technical point, there is a good example of how to host multiple WG pages from the Training WG, which Killian authored.)
Is this then maintained by the HSF? Specific instructions (e.g. Conda instructions) and limited support. What about code that gets abandoned later? At least the price of entry should be working CI (Travis, GitHub Actions). Everyone that contributes would be expected to maintain their piece.
Meeting agreed that we should proceed with this by baby steps, starting with the Exa-TrkX material from Savannah.
Some of us met with the Snowmass Computing Frontier (CF) convenors (Oli, Ben, Steve) on Monday. We discussed how the HSF contributed to the EPPSU and what we could do in Snowmass. Seven working groups are being formed who will look at particular topics.
Existing HSF documents (e.g. LHCC input) are welcome already as a submission, no need to pay too much heed to formal dates or LoIs, we are asked to submit as soon as possible.
Call for CF working group members will be open and an international flavour is encouraged (there will be smaller writing teams, but the WGs can be quite large).
36 student projects got slots. Now in Community Bonding period, until end of the month. A meeting with the students will be announced closer to the beginning of the coding period. The TODO for this period was sent to the mentors, to be followed by another regarding the Coding Period at the end of the month.
3 project proposals (AllPix2, Rucio, ROOT). Waiting for the announcement of Google of the accepted Orgs (May 11). The program targets starting effectively the technical writer projects on Sep 14, but the interaction mentors-potential writers starts as soon as projects get contacted, second half of May.
The HSF Common Tools and Community Software document was submitted on time for the review. It is also published on Zenodo. Some light updating can happen if people spot mistakes or new references are to be added (we know of a few). Please make a PR against the HSF Documents repository: https://github.com/HSF/documents.
Many thanks to all who made that happen!
In addition to the HSF document our generators group and the ROOT team provided additional input.
The WLCG and DOMA documents are public (will go on the WLCG website soon).
As far as we understand ATLAS and CMS foresee making their documents public after the review.
About 5 people per paper can attend the review. We will send Graeme and Liz from HSF Coordination, plus one person per WG: Andrea for Generators; Gloria for Simulation; Caterina for Reconstruction; Paul for Analysis.
Review is Tuesday 19 May, with a feedback session on Wednesday 20 May.
We organised a meeting yesterday prompted by Intel deprecating their low-level tbb::task API. Goals were to assess the state of thread management in each framework and project, as well as to ensure we remain coherent going forwards (it’s important that each piece of the full HEP application uses threads coherently).
Very useful summary of status and plans. Main conclusions:
task_group / task_arena - all in the TBB universe. Work so far is promising.Follow-up in the HSF Frameworks Group.
prmon (see issue #105). The main idea is to initially support NVidia and use nvidia-smi for collection. Then extend it to support other hardware as well.HSF WLCG Virtual Workshop on New Architectures, Portability, and Sustainability is next week.
We ask that people register so that we can gauge the level of interest and best organisation options.