Draft Plan for G4 Integration Draft

The Plan

For some time, we have had a plan in the works for the data flow from geometry and flux generator through to reconstruction. See Traudl's original description (esp. p 7) for the starting point. Here are the major points in that plan:

• simulator algorithm takes input from FluxSvc and detModel and its outputs are the McEvent, and Hits structures as described here. Drafts of the Hits classes are in cvs; the McEvent classes are already in use.
• A separate step of digitizations gets the Hits off the TDS and returns Digis to the TDS.
• Recon works from those new Digis.

Intermediate Step for G4

We believe we should take intermediate steps to get to The Plan. 

• create an algorithm and package, G4Generator, analogous to GismoGenerator.
• interface FluxSvc to G4
• over-ride the G4 event loop in favour of the Gaudi event loop
• use the instrument package to intercept G4's score mechanism.
  The results can be saved to an irf, and read back. In the latter case, the GlastDetector objects can be instantiated from an XML file, glast.prs, containing ids and geometry info.
  A converter, LdGlastDataCnv in GlastSvc, will fill TDS structures in TdGlastData, defined in GlastData.

These steps will allow us to deliver G4 output to Recon. These additional steps can be taken to complete the G4 part:

• at the same time, fill in Hits structure on TDS
• create persistent version of Hits

Note: "detModel" is a place-holder for interfaces provided by the detModel and xmlUtil packages. And it is expected that other packages, namely the recons, will have connections. Please see Joanne's note for a clarification.

Finally, the addition of Digi algorithms and converters for standard Recon input will finish off The Plan.

How to Get There from Here - Intermediate Step

Based on various discussions with people, here is a starting point for who does what:

• G4Generator package and algorithm - Toby (done)
• FluxSvc interface to G4 - Francesco, Nicolo, Johann
• detModel/xmlUtil interfaces, conversion to one or more Service's - Riccardo, Joanne, Marco
• over-ride the G4 event loop - Riccardo, Francesco
• interfacing instrument to G4 - Riccardo, Toby
• filling McEvent, Hits structure on TDS - Claudia, Monica
• PDS version of Hits - Heather

Note that this intermediate step would give us an early look at the Hits structure to see if it is feasible (data volume may be a problem).

For reference, here are the talks from Toby & Francesco from the November '01 workshop on this topic.

To There - Final Step

The Digi phase will involve completing the Identifiers scheme, to convert volume ID's to Digi ID's as well as writing code to reroute our current digi algorithms to eat the new Hits structure. Then we can embark on implementing new digi algorithms, à la Bari for TKR.

R.Dubois & T. Burnett:  Last Modified: 11/19/2001 20:13