Status meeting July 12th, 2012¶
Update on implementing electron's physics process (Soon, Dongwook)¶
I've finished the procedure to find the physical interaction length for the electron bremsstralung process and evaluate the CPU/GPU gain which is about a factor of 30 for lookup table and 80 for on the fly calculation. The incoming particle is assumed to be 100K electrons with 10 GeV and they are spread into 32 blocks with 128 threads on each block on GPU. The gain does not depend on incoming energy. Now I am thinking to implement PostStepDoIt.
Some optimization was done but no explicit tuning to the GPU architecture.
Update on CUDA development (Soon)¶
- merged geometry and em physics codes into transportation
- committed updated codes to the repository
- contributing write-up for the FWP report
- Tried alternate RK with less lookup (which means less computation so less speed up but still less total time).
- Comparing table-based field lookup and the parametrized field lookup but could not yet get a good measurement. Soon will send instruction to Philippe to check.
Shared new shape with John A.
Discussion of the Geant4 workshop report in DC.¶
Bob Lucas and Rob Roser are planning on taking over the editing of the document next week and we are sending a 'last' set of comments tomorrow.
Discussion of the FWP report due at the end of the month.¶
Donwook has passed on his contribution to Soon. Soon currently has too much material and was working on summarizing. Philippe and Jim will take over the summarization. Soon will send it current version tomorrow.