Presentation is loading. Please wait.

Presentation is loading. Please wait.

Requirements for the O2 reconstruction framework R.Shahoyan, 14/08/2015 1.

Similar presentations


Presentation on theme: "Requirements for the O2 reconstruction framework R.Shahoyan, 14/08/2015 1."— Presentation transcript:

1 Requirements for the O2 reconstruction framework R.Shahoyan, 14/08/2015 1

2 Base track model AliRoot AliExternalTrackParam AliVTrack AliVParticle Tobject  192 B in total, well optimized, fully sufficient for propagation in barrel. The size can be halved by going to Float_t (but keeping calculations in Double_t)  In most algorithms TObject functionality is not used, at most to attach it to TClonesArray (but STL is more efficient) → ideally would prefer class with multiple inheritance, e.g. AliTrackT : public TObject, public TrackParam NON-TObject TrackParam implements track specific functionality (to be used in algorithms). Compatible with HLT needs (which uses 2 versions: with and w/o covariance matrix) AliTrackT is used mostly for persistency. FairRoot FairTrackParamH(FairRoot/trackbase/) FairTrackPar Tobject  640 B in total  Lot of data members which we don’t use FairTrackParam (FairRoot/base/event/) Tobject  Equivalent of AliExternalTrackParam but defined for forward tracks (might be good for MFT, MUON). Both miss propagation methods implementation? 2

3 RecoParam framework Absent in FairRoot? AliRoot AliRecoParam : container for AliDetectorRecoParams for every event type  Can be ported to AliceO2 almost as it is (just change TObjArray with random order to fixed array of AliDetectorRecoParams ) Vertex base class AliVertexESD can be ported as it is Clusterizer, Tracker base classes Absent in FairRoot? In Aliroot used for  Containers/IO of digits/clusters of specific detector: this functionality must be kept also in AliceO2  Generic track propagation routines with material correction: can be moved Utilities class  Access to vertex: in O2 in general will be >1 vertex, can be accessed directly from Event container. → Start with very light base class, used to register the tracker + IO 3

4 Digit, Cluster AliRoot AliITSdigit TObject  112 B in total, most of memory spent on 2 fixed size (10) arrays to store hits and labels references. AliDigit TObject  32 B in total, fixed size (3) array to store labels AliCluster TObject  56 B in total, fixed size (3) array to store labels  Alignment support → Size of 3 labels might be too few for generic digit, size of 10 is too much. → MC labels are used in very small part of digits/clusters algorithm, still take most of memory, shall we consider dedicated object to hold MC references, to which digits/clusters will refer to by index? Slower in access, more efficient in memory usage. FairRoot Digit(Alice02/its/) FairTampStamp FaitMultiLinkData_Interface TObject  128 B B in total, of which 80 is TimeStamp  Labels are stored in std::vector → dynamic objects kill advantages of using TClonesArrays or STL containers  Do we really need to waste so much memory on per-digit labels which are common for large groups of digits?  If yes, can’t we instead use an index of the TimeStamps object describing the whole group. 4

5 Alignment framework AliRoot AliAlignObjParams AliAlignObj  Shift/rotation parametets + matrix representation for single alignment volume.  Interfaced to TGeo access methods (by ID or symbolic name) and alignment mechanisms via TGeoHMatrix AliceO2 MisalignmentParameter (Alice02/its/) FairParGenericSet  Array shift/rotations parameters for multiple volumes set  Mechanism not clear (not implemented yet?) Material query support AliRoot AliTrackerBase::MeanMaterialBudget(xyz0,xzy1)  Very precise but extremely slow TGeo query for all information we need in tracking (needs some fixes) MaterialMap.C: LUT on cylindrical mesh, “demo” development by Andrei, single slice only, identical to LUTs used in detectors classes Need to implement configurable LUT stacks for thick layers. FairRoot FairMesh + FairRadGridManager ?  Not sure intended for reconstruction-time per track query 5

6 Alignment framework AliRoot AliAlignObjParams AliAlignObj  Shift/rotation parameters + matrix representation for single alignment volume.  Interfaced to TGeo access methods (by ID or symbolic name) and alignment mechanisms via TGeoHMatrix AliceO2 MisalignmentParameter (Alice02/its/) FairParGenericSet  Array shift/rotations parameters for multiple volumes set  Mechanism not clear (not implemented yet?) Material query support AliRoot AliTrackerBase::MeanMaterialBudget(xyz0,xzy1)  Very precise but extremely slow TGeo query for all information we need in tracking (needs some fixes) MaterialMap.C: LUT on cylindrical mesh, “demo” development by Andrei, single slice only, identical to LUTs used in detectors classes Need to implement configurable LUT stacks for thick layers. FairRoot FairMesh + FairRadGridManager ?  Not sure intended for reconstruction-time per track query 6

7 7 Output event container  It will take some time to define optimal final event format, also depends on still unknown HW parameters (affecting pile-up expectations)  In near future we need it mostly as a placeholder where trackers / vertexers could write/read their output. → Start with generic container for arbitrary time slice, on which we can exercise even separation: array of tracks + primary vertices + secondary vertices


Download ppt "Requirements for the O2 reconstruction framework R.Shahoyan, 14/08/2015 1."

Similar presentations


Ads by Google