TRD tracking improvements - #6943
Conversation
shahor02
left a comment
There was a problem hiding this comment.
Hi @martenole
Thanks! Sorry, I did not understand what is the difference between the spec and the tracker in extracting the TPC clusters access? In principle, it would be more logical to have the refit in the tracker itself.
Please see one comment below.
|
Could you remove the Draft tag to trigger CI? |
davidrohr
left a comment
There was a problem hiding this comment.
I think it is OK in the DPL device for now.
Eventually I would move it into the GPU Refit class instead of the GPUTRDTracker class
|
Hi @shahor02,
the only difference is that I need to access it through the GPU IO pointers instead of through the RecoContainer. It should not be a problem to change it. But since I was not sure how to do it and wanted to have a first version ready fast I did it in the device for now. Should I also store the chi2 for the inward/outward refits? The TRD track has one member for its chi2 which is currently the chi2 for the matching of the seed to the tracklets. I will add also the refit for ITS-TPC-TRD and then mark the PR ready for review. Cheers, |
d7382f5 to
056a3a2
Compare
- refits inward/outward for ITS-TPC-TRD and TPC-TRD tracks - calculation of TrackLTIntegral during inward refit
056a3a2 to
e7d318e
Compare
HI @martenole : I think accessing the data through the GPUIOPtr is actually the way to go, since that will allow us to move it to the GPU more easily later on. |
* Revert dummy TRD track base class * Suppress fit parameter range warning * TRD tracking default parameter adjustment * Add TRD refits (temporarily in DPL device) - refits inward/outward for ITS-TPC-TRD and TPC-TRD tracks - calculation of TrackLTIntegral during inward refit * dummy commit to restart fullCI Co-authored-by: Ruben Shahoyan <shahor02@users.noreply.github.com>
Hi @shahor02
the last commit 671a22e adds the inward and outward refits to the TPC-TRD tracks. For the moment I added the method to the DPL device, eventually it should go into
GPUTRDTracker. For that I just need to check how to retrieve the relevant information (TPC cluster access/transformation) from inside that class.But maybe the general functionality can already be checked? Or should I rather try to move the implementation directly into the TRD tracker?
Cheers,
Ole