Meeting minutes 2019 08 23 - Geant-RnD/ExascaleProxyDocuments GitHub Wiki

Attendees: Philippe, Soon, Guilherme, Yunsong, Jonathan.

Yunsong: Roofline model and CPU and memory time. GeantV was 4 times more memory than Geant4 benchmarks.

Soon: Which version are you using?

Yunsong: Master as of one month ago.

Soon: I am a bit surprised at the result, there was significant improvement in that area recently.

Soon: You should redo the test but wait for a little bit as we are debugging the physics quality at the moment.

Soon: Long break, back in action soon. Before that I was looking at implementing a general sampler class for the Final State. But this would depend on which physics model we are integrating in the prototype.

Philippe: Stefano’s start date is in Limbo due to work Visa issues.

Guilherme: Working on validation of GeantV. Number of secondaries, track length and energy deposit are surprising. We are adding finer grainer statistic gathering. Number of Electrons and Gammas is different. There is some structure in the G4 plots, I am updating the GV test/library to create the same plot.

Jonathan: RAS. About to hit a milestone with TiMemory (adding GPU roofline - for CRD at Berkeley Lab) - CUPTI is challenging to tame.

Philippe: I will be attending the US Atlas and we should probably swap our talks.

Philippe: For the next Hackathon, should we wait on Stefano for the next one?

Jonathan: I don’t think so.

Philippe: I propose the week of October 14th or 21st.

Jonathan: Both works for me.

Soon: There is the GeantV committee meeting on the 15th.

Philippe: Indeed, we go with the week of October 14th we will need to start on Wednesday.