ECHEP Trigger/reconstruction area meeting
Additional related contributions from Edinburgh 2020 meeting
ECHEP Trigger/Reco meeting 30/04/2020
General comments on themes/interests
- General interest for utilising heterogenous computing platforms, e.g. HPCs with GPUs.
- General Interest in co-processor (both GPU and FPGA) acceleration for realtime triggering/reconstruction
- Interest in utilising abstraction layers (both for GPU and FPGA) to separate the code developer from the lower layer device deployment.
- Interest in ML for reconstruction, either low level (LHCb track stub finding for FPGAs) or higher level (Dune).
- Interest in code profiling and optimisation. Work already ongoing or completed in Trigger/Reco WG perhaps of interest to other WGs.
- Potential industrial links for FPGAs (CMS,LHCb,?) (e.g. Xilinx,Intel)
- Interest in exploring if LHCb GPU model (full sequence, multi-event framework) is applicable to GPDs.
- Question on if specific reconstruction algorithms are too Experiment specific to be in scope for ECHEP.
- Question on whether 4D (timing) aware reconstruction is in scope for ECHEP.
From these ideas we need to try and identify specific proposals for R&D areas, that in the context of ECHEP represent experiment neutral areas of study that are also genuinely useful and applicable to specific Experiments, at a later stage. We should start to collect these in
https://docs.google.com/document/d/1lhC5xhiwG1lblhipW-0z3ukobzwCaMFmtJ0GY5mrPKA/edit?usp=sharing
There was also a suggestion that some of the ideas, e.g. code profiling and optimisation, usage of heterogeneous computing resources goes beyond the trigger/reconstruction WG, and as such it should be considered if a refactoring (rebranding) of the WP areas should be considered.