Framework is capable of running Geant4 simulation

[knoepfel]

DUNE US S&C R&D item 075

[knoepfel]

The primary complication is that Geant4 uses static thread-local variables that do not map cleanly to oneTBB tasks. However, @olantwin of the SHiP experiment at CERN has demonstrated that concurrency gains can be achieved in Phlex by using asynchronous C++ facilities for running Geant4.

A subsequent discussion between Oliver, CMSSW developers, and Phlex developers suggested a more efficient option: do an explicit thread "hand-off" between Geant4 and oneTBB. CMSSW does this already. The Phlex developers believe this will be achievable through the limited-resources system, which has a fairly mature design but is not yet implemented within Phlex.

[olantwin]

The new approach inspired following from our discussions is implemented here: https://github.com/ShipSoft/aegir . We're now verifying and benchmarking against the SHiP legacy framework.

I can prepare some additional documentation for a follow-up discussion, also to make sure I understood/implemented things correctly.