Status meeting May, 10th, 2012¶
Update on implementing electron's physics process (Soon, Dongwook)¶
DongWook has made progress on the EM physics implementation (one code path; Brem-stralhung (sp?), which needs cross-section which comes either from lookup table or calculated on the fly in a 8x11 floating array which are constant (function static in the original code).
Update on CUDA development (Soon)¶
Soon presented our progress report at the Concurrency bi-weekly meeting.
Question of how to evaluate the cost of running on the GPU.
Question of how to we populate the tracks.
Soon: I am starting a simple geometry because the next step is how to make more calculation on the GPU side. We need copy the G4 geometry classes.
Philippe: I was hoping we could limit the amount of duplication.
Soon: Yes, but we actually need a large fraction of the code.
Jim: In addition, we met developer working on geometry and they are doing things differently (voxelization, triangles, etc..). They are eager to collaborate with us.
Philippe: A question is how to get to the simple example?
Soon: I think in a couple of week.
Discussion of the Geant4 workshop in DC.¶
We meet with DOE HEP program manager and the DOE ASCR program manager. The idea was that G4 is likely to be the 'slowest'/'biggest' application that can benefit from parallellisation and thus from a collaboration with computer science.
Scientific computing has increased funding. There was a mixed of attendees (more or so applied) and review of the need of G4 and the tools and skills that can/could come form scientific computing.
Their will be a white paper to request funding for G4 in the context of scientific computing and hopeful make up some of the funding that the lab has lost.]
Expect to be writing proposals in September. Some worry that the ASCR program manager might be too focus on (non-existent!) potential large performance gain obtainable in a short time.