Geant4 EM Physics Code Review Weekly Phone Meeting¶
Date: October 11, 2013 at 10.00 AM US CDT (Fermilab local time)
Place: 1-866-740-1260 (ReadyTalk line) (Host : Krzysztof)
Soon Yung Jun
(Summary by Krzysztof)
We discussed Boyana's puzzling results regarding the cache misses and Soon attempts
to use TAU and some of TAU limitations regarding analysis of Geant4 code.
We may need to double check the versions of code which were used in cache misses comparisons.
Boyana and Soon will document the steps of using TAU.
Krzysztof reported on his brief analysis of G4eplusAnnihilation.
The class seems fine and the only possible suggestion would be to get
the the 3 random numbers in one step rather then in the three separate calls.
Soon reported on his initial look at the classes inheriting from G4VEnergyLossProcess
and the fact that some of the functions there are quite long.
We decided to continue looking at that group of classes with Krzysztof joining Soon in it.
Boyana will analyze related classes/functions with TAU:
(based on 50GeV electrons no magnetic field, prof_big_functions_leaf_count_list)
and we "compare the notes" once we know more about the functions.
Soon will look into generating the profiling tables in the csv format
to make it easier to input the information into the profilers.
Regarding the (full) initialization of the objects in the constructors,
we decided that we may not be able to make a general recommendation and
will need to decide on it on a case by case basis.
G4PhysicsVector may be a case where it may be useful to do it in the constructors
to make the containers read only for the MT case,
but than again it may be problematic if the constructor
would need to read the data from external files.
We shall meet again on Friday (10am Central).