Meeting minutes 2019 11 01 - Geant-RnD/ExascaleProxyDocuments GitHub Wiki

2019-11-01

Attendees: Philippe, Guilherme, Soon, Seth.

Philippe: Finished the presentation for CHEP 2019 to be given by Liz.

Guilherme: Working on getting VecGeom navigation to work on GPU, level locator in particular and Safety calculations. I building on the testing code and finding a few deficiencies and I have solved those. I am moving on to Safety. Since the shapes are properly vectorized.

Philippe: Make sure to focus on the scalar interface and ‘just’ record that the vector interface has bugs.

Soon: Connection with Materials?

Guilherme: I am not sure. Let’s look into it after the meeting.

Soon: I don’t see any problem with material cuts.

Philippe: Should we hold off on that?

Soon: I am working on building the interface for physics.

Philippe: I see, then we do need “some material cuts” but not necessarily the full fledge version.

Soon: In track state, we have ‘material’ information that still need to be flushed.

Soon: About Processes and Models, we don’t need to decide things “now” but I want to put something in to move forward. There is currently a base process used in the tuple thing and it has ‘generic’ GetProposedInteractionLength in derived class then there is specialization for Particle type and Process type … the generic function should move to the base to avoid duplication. I want to try one model and one process completely, and have it run on the GPU.

Soon: What is the plan for moving the mini-proxy to GPU?

Philippe: Not sure how far did Jonathan go. Need to check with him

Soon: I am also worried about validation.

Seth: No technical progress. Got side tracked with conversation with Jonathan. What about Physics related track state? Would we need variable size state?

All: discussion on when/what need to be copied back and forth between CPU/GPU.

Seth: next I am planning on sketching out a Secondary generator.

Philippe: On my plate magnetic field support and global variable access.