Notes from Feb. 7th, 2012 Tracking Discussion with Hall-B
From GlueXWiki
The following are notes from a discussion with some Hall-B folks on charged particle tracking that took place on Feb. 7th, 2012. These were sent to Volker and Eugene on Feb. 15th.
Hi All, I wanted to summarize our tracking discussion from Tuesday since I feel it addresses the topic of sharing tracking software between halls B and D. Something we were asked to meet and discuss by Rolf. Please have a look at the summary below and send me any corrections you may have. My intent is to forward this to Eugene and Volker once there is consensus. ======================= Attending: Maurizio U., Veronique Z., Dennis W., Mac M., Mark I., Simon T., David L., Dave H., Vardan G., Ken H. CLAS12 has some existing tracking code, but Maurizio and Veronique are endeavoring to construct a new package that would allow them to more easily modify and maintain it. Maurizio had previously contacted David and Simon to ask about code Hall-D used for helical fits to see if it would be appropriate for CLAS12. Prior to the meeting Maurizio and Veronique were able to take a look at the two algorithms currently deployed in the Hall-D tracking code. Some discussion was had about details of how the helical fits were done (two stages, each fit to a a set of 2-D points and each yielding 2 parameters for a total of 4 fit parameters). The code for this was fairly general and could likely be adopted with minimal modifications for CLAS12. The subject of sharing code directly was discussed briefly. Hall-B was interested in possibly developing a package with the general use utilities that both hall's tracking codes would need. Hall-D felt that in their experience, it was extremely difficult to maintain generality when developing the tracking code and it might be more efficient for each hall to maintain their own development stream, but keep close communication so knowledge and algorithms could be shared easily. Hall-B will move forward with development of a set of general use tracking utilities while developing their software. Hall-D may refactor its code in the future to adopt these generic tools at a point when it seems appropriate to do so. Mac suggested that we try and look for common terminology and, wherever possible, use similar configuration parameters for both tracking packages. This would provide a level of continuity between the two halls, that might make communication and comparison easier. Hall-D went on to describe some other details of how the tracking was done in their code. This touched on the following areas: - How candidates are formed and how they are linked between the CDC and FDC detectors - If, when and how clones and ghosts are handled - Mechanism for hit selection is done - Kalman Filter - SIMD instructions and linear algebra packages - Hough transforms and Riemann fits Maurizio and Veronique were going to get started by trying to implement the Hall-D helical fitter and would stay in close contact with Simon and David as the development progressed. ================