Geant4 EM Physics Code Review Weekly Phone Meeting


Date: November 15, 2013 at 10.00 AM US CDT (Fermilab local time)
Place: 1-630-840-6737 (Soon's office)


Soon Yung Jun
Boyana Norris

Krzysztof Genser
Andrea Dotti


Status report from Boyana

1. collect performance data and need to put them on wiki

2. preparing python script to make instrumentation/measurement easier

3. describing the procedure of TAU instrumentation.
- working with the TAU group to resolve a hang-problem during TAU instrumentation
(manually can kill the process and continuously proceed the process when the hang is detected,
but want to have a clean instrumentation).
- request to add one line to prevent from a parse error (committed to the g4emreview/g4.9.6.r10 repository by Soon)

--- a/g4.9.6.r10/geant4.9.6.r10/source/processes/electromagnetic/dna/molecules/management/include/G4MoleculeID.hh
+++ b/g4.9.6.r10/geant4.9.6.r10/source/processes/electromagnetic/dna/molecules/management/include/G4MoleculeID.hh
@@ -58,8 +58,9 @@ public :

     static G4MoleculeID Create()
-        fLastValue ++;
-        return G4MoleculeID(fLastValue);
+        fLastValue++;
+        int tmpLastValue = fLastValue;
+        return G4MoleculeID(tmpLastValue);

Status report from Soon

1. investigated the inheritance of G4VEnergyLossProcess and confirmed that all derived classes except
G4eIonization and G4IonIonization should be DiscreteProcess (i.e., AlongStepGPIL and AlongStepDoIt
are not necessary). The isIonisation flag is set to true in the constructor of G4VEnergyLossProcess,
but it was overriden by SetIonisation(false) in (with )

processes/electromagnetic/standard/src/, line 99
processes/electromagnetic/muons/src/, line 87
processes/electromagnetic/muons/src/, line 94
processes/electromagnetic/polarisation/src/, line 71
processes/electromagnetic/highenergy/src/, line 66
processes/electromagnetic/highenergy/src/, line 67 

So, AlongStepGPIL and AlongStepDoIt of those processes called for doing no meaningful works


1. Boyana is interested in TBB applications on Geant4 and discussed common interests on parallel applications
with the Intel MIC. Soon also mentioned the on-going effort of FNAL porting the GPU prototype to MIC for
different programming models, which includes applications with openMP, TBB, CilkPlus and openCL.

2. Automatic conversion C++ code to device codes (CUDA, openCL)
- Soon will provide a simple C++ code out of the G4 magnetic integration codes and Boyana will try to generate
a conversion code for an initial test.

3. We shall meet again in three weeks (10am Central on Friday Dec. 6) due to Super-computing and Thanksgivings,
but communicate through emails. We send out a notice to call a written update on every Thursday.