Present: Vito Di Benedetto, Giuseppe Cerati, Lynn Garren, Kyle Knoepfel, Robert Kutschke, Marc Paterno, Gianluca Petrillo, Erica Snider
Remote: David Adams, Robert Hatcher, Alex Himmel, Katherine Lato, Gleb Sinev
Project status report [Erica Snider]
- still working on Tessorflow packaging into UPS product
- discussion about guidelines on code and module documentation
- possible actions: convert existing top-of-the-header documentation into Doxygen format, design a good connection between FHiCL documentation and Doxygen documentation
- [Giuseppe Cerati] is documenting via Doxygen the FHiCL configuration
struct, and having the module documentation point to that struct
- on November 10, 2017, OSX 10.10 ("Yosemite") will not be supported by FNAL any more
- no official builds will be possible after that
- [David Adams] which OSX Yosemite users should update to? [A] OSX 10.12 ("Sierra") recommended (not 10.13 "High Sierra")
Conclusions
- requesting feedback on a 3D reconstruction workshop together with LHC (ALICE)
Introduction to multithreading [Kyle Knoepfel]
- art is hosting meetings to inform and discuss multithreading in the framework
- art is preparing for a version 3.0 supporting processsing multiple events in parallel
- Intel Thread Building Blocks (TBB) are used for the implementation
- if user code needs direct access to parallelism, it is recommended to use TBB as well (parallel algorithms or TBB tasks)
3D track vertexing [Giuseppe Cerati]
- the common vertex from two tracks is computed as the point closest to both tracks, considering uncertainties
- if more than two tracks are present, a vertex is formed from two of them, then improved adding information from one track at a time
- new content is necessary in
recob::Vertex
- fit result information
- impact parameters
- list of tracks associated to the vertex
- ... and more
- [Robert Kutschke] have you considered a vertex object containing the correlations between track momenta as constrained from the vertex, and the vertex position? [A] that information is not produced yet, and there is no room for it as now [Erica Snider] we can think of extending the vertex through facades [Robert Kutschke] this solution might reveal to be "icky"
- [Robert Kutschke] the "valid" flag might be replaced by an enumerator describing that the vertex fit succeeded or how it failed
There are minutes attached to this event.
Show them.