vmc-project / vmc Goto Github PK

We are currently investigating some (internal) redesign. For this we consider multiple inheritance. Simplified:

• Class M1 inherits from nothing
• Class M2 (basically FairModule) inherits from TNamed
• Class D1 inherits from M1 and TVirtualMCSensitiveDetector
• Class D2 (basically FairDetector) inherits from M2 and TVirtualMCSensitiveDetector

classDiagram
  M1 <|-- D1
  TVirtualMCSensitiveDetector <|-- D1
  TNamed <|-- TVirtualMCSensitiveDetector

  TNamed_ <|-- M2
  M2 <|-- D2
  TVirtualMCSensitiveDetector <|-- D2

Case D1 is simple and will just work.

But case D2 has a problem. There are two ways to reach TNamed. I think this is called diamond inheritance. Currently D2 will have two copies of TNamed (TNamed and TNamed_ in the diagram). Not really, what we want.

1. The C++ answer to this problem is called virtual base classes. These come with some performance overhead when accessing TNamed member variables (might or might not be relevant. We haven't checked).

   classDiagram
     M1 <|-- D1
     TVirtualMCSensitiveDetector <|-- D1
     TNamed <|-- TVirtualMCSensitiveDetector
   
     TNamed <|-- M2
     M2 <|-- D2
     TVirtualMCSensitiveDetector <|-- D2
   

2. Another question is, whether vmc, geamt4_vmc and geant3 actually depend on TVirtualMCSensitiveDetector deriving from TNamed? If none of them depend on this base class, it maybe could be dropped from TVirtualMCSensitiveDetector? This would solve D2 and even improve D1, because the whole TNamed thing would be gone.

   classDiagram
     M1 <|-- D1
     TVirtualMCSensitiveDetector <|-- D1
   
     TNamed <|-- M2
     M2 <|-- D2
     TVirtualMCSensitiveDetector <|-- D2
   

What do you think?

cc: @dennisklein

In TGeant3 AddIons() is called before InitGeometry():
fApplication->AddIons();
fApplication->InitGeometry();

On the contrary, the order of calling these functions in TGeant4 is different:
fApplication->InitGeometry();
fApplication->AddIons();

Is it intended behavior?
Can it be changed so that the functions are called in the same order in both TGeant3 and TGeant4?

Hello.
We found some difference in particle acceptance using GEANT3 and 4. After some study I found possible source of difference.
I simulated proton and anti-p with flat momentum distribution. They fly through magnetic field until detector. I built momentum of the particles accepted in the detector. These distributions in GEANT3 and 4 are different for particles with momentum lower 2-3 GeV/c. I changed the medium description for GEANT3, and can reproduce GEANT4 distribution. I tried to explain my study a few slides. I would like to ask: could you please check, that all medium parameters concerning magnetic field are proper calculated or converted from input file to the GEANT3, and they are the same as for GEANT4.

Best regards
Anna Senger
G3_G4.pdf

The compilation of the legacy code now fails. This is fixed by adding the options to Geant3BuildLibrary.cmake

<-fallow-argument-mismatch -fallow-invalid-boz>
I don't know whether this needs to be conditional on gfortran_version>9 or whether it will be OK for older versions as well.