Project

General

Profile

Geant4 EM Physics Code Review Weekly Phone Meeting

Meeting

Date: December 6, 2013 at 10.00 AM US CDT (Fermilab local time)
Place: 1-866-740-1260 (ReadyTalk line) (Host : Krzysztof)

Participants:

Andrea Dotti
Krzysztof Genser
Soon Yung Jun
Boyana Norris

Discussion

(Summary by Krzysztof)

Krzysztof reported that it did not look like gcc stl has a partial
specialization for vector of pointers, so the fact that our vector of
vectors performance was worse than the one of vector of pointers to
vectors was not due to a special library optimization.

Boyana posted her freshly obtained TAU results (see the main wiki
page).

Andrea mentioned that there should not be many changes in the EM code
between v9.6.ref10 and the just released Geant4 v10

Soon has placed Geant4 v10 at our Redmine repository

Plans:

We agreed to move to Geant4 v10 and continue our review using it.
We should finish our pass through the initially selected functions
and then "re-validate" our previous findings which were based on the
previous version of Geant4.

Boyana is going to expand the TAU results tables to include exclusive
function data and more results including cache misses, as well as provide
documentation on how to use TAU to profile a Geant4 application, using
SimplifiedCalo as an example.

Andrea will look at G4UniversalFluctuation::SampleFluctuations

Soon will profile Geant4 v10.

Krzysztof will look at G4UrbanMscModel

We should aim at writing our final report starting at the end of January

Next meetings: 12/13,20, 1/10