Core Status 24 October 2001

Previous status report; Joanne's meeting  minutes

Enabling G4 as a Gismo alternative: what we (still) need

• A CMT package: either an interface package to the external G4 installation, perhaps the current Geant4 (now at v3r2), or one that builds G4 itself. In the latter case, all G4 source code must be imported to the SLAC cvs repository.
• A new package that uses the above, corresponding to the current GismoGenerator with an algorithm like its GismoGenerator that:

1. Instantiates the LAT geometry description in Joanne's xmlUtil package ( see especially flight.xml), using a visitor based on Riccardo's detModel . Riccardo is refactoring the visitor code in detModel to improve this interface. Presumably the reputed G4 visitor will be then rewritten and submitted to cvs.
   
   Sensitive detectors have to be identified to Geant4, and appropriate call-backs set up to process energy deposited in ACD tiles, TKR Si planes, and CAL logs and diodes. An existing candidate for this is the provided by the  instrument package. One slight issue is the requirement to interpret distance units in  use mm and energy units in MeV. An advantage is that display code exists.  The current persistence and irf  mechanisms would not be used: the first is not needed, and the second would be replaced with ROOT digi classes. But note that the long-term goal is instead to create an intermediate structure containing only hit information, and to have a separate hit-to-digi translation.
    
2. Connects to the FluxSvc service to get the a particular flux of particles for its "gun".  I have prepared a document showing the GismoGenerator code using it.
    
3. For each event:
   • obtain a new particle (by name, position, kinetic energy, direction and elapsed or absolute time) from the FluxSvc: load it into the gun and cock it. (See the document.)
   • Fire the gun, collecting all hits in sensitive detectors, and finally generating digitizations in the ACD, TKR, and CAL systems, using the call-backs set up during initialization.
   • Save the digitization information in the TDS for subsequent transfer to persistent store, or reconstruction.  (If we use instrument, this is handled already. )

Memory leaks

TkrRecon leak was found by Tracy. Hooray!  No one has time to follow up and install tools, or look into the list that Karl sent us. The catastrophic leak seen by Steve and Toby has not been understood.

PDR

Karl has used v7 to run 1M all_gamma, as 100 jobs, with a merged root tuple file available at:

/nfs/farm/g/glast/u07/flight2/data/all_gamma1M

The chained tuples are in a file called: all_gamma_1M_TUP.root

Even newer CMT: v1r10 about to be released.  A major new feature is that "use" statements are now scoped as private or public. Thus a package that needs another package only to build binaries, but not as part of its interface, can hide the other package.

T. Burnett Last update: Wednesday, 04. August 2004 15:42:08 -0700