2024.01.09 - https://indico.cern.ch/event/1359196/

Budget

Axel shared the current budget:

• 450k year 1 (150k for EP/SFT, 100k for CMS, 100k for ATLAS, 100k for IT)
• 550k years 2, 3, 4
• 350k year 5
• this does not include about 200k extra for ALICE and LHCb (2x ORIGIN)

This includes everything: overhead (material/travels) and supervision

Note: technically, the budget is split into EP and IT.
For example, for year 1 we have ~550k for EP and 100k for IT.

How is the 10% of “supervision” money spent and accounted ?

• can we ask the supervisee to cover for 10% of the supevisor work ?
• otherwise, does the money go to EP ?
• or to supervisor’s group, e.g. EP/CMD ?

Expressions of interest:

Question: how do we want to organise the work ?

• split by experiment ?
• split by topic/milestone ?

Axel: propose to hire people to work on topics/milestones, and work across experiments.
We still need to identify a person responsible for the delivery of each topic/milestone.

ALICE (David Rohr)

• efficient data structures
• interface to ML inference (possibly funded on 1.2/1.3, still tbd)
• (low priority) novel programming languaes

LHCb (Marco Clemencic)

• efficient data structures
• efficient scheduling
• (personal interest) novel programming languaes

Marco Clemencic: what about high level domain-specific languages ?
For example, a DSL that gets transpiled into C++/CUDA/whatever and scheduled efficiently on each backend. This may sit across between efficient scheduling and programming languages.
Note: LHCb not much interested on cross CPU/GPU scheduling, but they might be interested for the simulation use case.

IT (Stefan Roiser)

• efficient scheduling
• efficient data structures
• (low priority) interface to ML inference

Topics for IT are mainly simulation and especially event generation.

CMS/IT (Eric Cano)

• efficient data structures
• common heterogeneous libraries

CMS (Felice)

• interface to ML inference
• efficient data structures
• common heterogeneous libraries

CMS (Andrea)

• efficient scheduling
• efficient data structures
• (personal interest) novel programming languages

SFT (Axel)

• interface to ML inference (only if Sophie is relevant)
• efficient data structures (based on C++ reflection)
• (not a priority) common heterogeneous libraries

ATLAS (Attila) (after the meeting)

ATLAS is interested in all the topics, except for the novel programming languages:

• efficient scheduling
• efficient data structures
• common heterogeneous libraries
• interface to ML inference

Action Items

• Axel will ask Attila (ATLAS) for their interest
• within each bullet point we have to discuss how to distribute the budget and personpower
• the next meeting will be on Tuesday the 16th at 15:00.