Draft: TRD clusterfinder rewriting.
For viewing by @apuntke, @a.bercuci and @p.kaehler.
This is the current state of my work on TRD hitfinders.
The main thing to look at for now is CbmTrdModuleRecR::BuildClusters(). This is my suggestion for efficient building of clusters in 1D. Row merging was removed for now, as the original implementation doesn't work as intended (discussed with @a.bercuci), and should be added as a post-processing step.
Merge request reports
Activity
requested review from @d.smith
assigned to @d.smith
Could you maybe include these tests from MR1744 into your MR? Then we would immediately see if your updated version passes them in the pipeline. They can also be run by you manually by navigating into the
build/reco/detectors/trddirectory and runningmake test. The 2-row cases you can comment out for now.Also the time interval these tests require was doubled to two spadic clockcycles (2*62.5 ns) because we observed some time shifts between CROBs in the current beamtime. We are currently working on correcting these shifts in the unpacker which probably makes this increase obsolete again. Therefore you can adjust the CheckTimeDistanceSmallEnough test again to use digis with a smaller time distance.
added 1 commit
- d1dfd2c4 - Clenup and ROOT removal. Hitfinding now done module-wise (RefId changed).
added 1 commit
- 87ec4cb9 - Started to refactor CbmTrdModuleRec2D by introducing a class for transient hit data.
added 1 commit
- e3302205 - Cleanup and removal of ROOT types. AddDigi and BuildCluster fused in 2D.
added 1 commit
- 3f69ea81 - Moved TRD2D cluster finding to separate functor class.
added 1 commit
- b1855bda - TRD2D cluster finder uses ROOT-free clusters.
added 1 commit
- 64061ffb - Added option for retaining order of modules in hit output tree (ROOT histogram...
added 1 commit
- 6d1227e6 - Preparations for module-wise cluster building.
added 1 commit
- fad82d23 - Cluster algos are now instantiated module-wise.
