TPC SCD: add option to keep all cluster and add MC to unbinned residuals - #15893
Open
matthias-kleiner wants to merge 6 commits into
Open
matthias-kleiner wants to merge 6 commits into
matthias-kleiner wants to merge 6 commits into
Conversation
…opagation fails TrackInterpolation drops the whole track when the propagation of the reference track to a TPC pad row (or to the TRD/TOF anchor) fails, e.g. at scdcalib.maxSnp. That propagation follows the ITS-predicted direction, so the drop selects tracks by the reference's own curvature error and biases the residuals of the surviving sample at low pt. With scdcalib.keepClustersOnPropFail=true the track is kept. TPC clusters without a reference are stored position-only: y, z = cluster position, dy = dz = 0, tgSlp = UnbinnedResid::TgSlpPositionOnly (-0x8000, which the tgSlp packing never produces; UnbinnedResid::isPositionOnly()). extrapolateTrack: all clusters after the failure. interpolateTrack: a row keeps its residual only if both the outward and the inward propagation reached it; a failure of the outer TRD/TOF anchor makes all TPC clusters position-only and suppresses the TRD/TOF residuals. Position-only clusters are excluded from validateTrack and stored in row order. The ITS and PV residuals are unaffected. ResidualsContainer::fill, staticMapCreator.C and TPCResidualReaderSpec skip position-only (and tgSlp-clamped) residuals for the binned voxel fit. Default false: output unchanged (verified on one MC TF, all fields identical). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The interpolation workflow had MC switched off ("not yet implemented"). With --enable-mc (opt-in, since the
workflow also runs on data without --disable-mc) and --send-track-data, a TrackDataMC vector aligned 1:1 with
the TrackData output is sent as GLO/TRKDATAMC: MC labels of the ITS-TPC part of the seeding track and of its
ITS and TPC parts (flag for a fake ITS-TPC match), the truth at the ITS outer parameters (nearest ITS track
reference, propagated with the workflow's material correction and the true mass to the x and alpha of
TrackData::par, with the distance to that reference), the truth at the TPC entrance (first TPC track reference
in time, sector frame) with the distance between it, propagated to the innermost TPC cluster of the track, and
that cluster (flags loopers, wrong legs and fakes), and the PDG code. Only the ITS, TPC and ITS-TPC track
labels are read; the kinematics are looked up event by event and released after each event. The residual
aggregator writes the truth as branch trkMC of the trackData tree with --enable-mc.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
For tracks with TRD residuals (TRD seeds and ITS-TPC map seeds whose most complete track has TRD), TrackDataMC now also holds the truth at the TRD entrance (first TRD track reference in time, sector frame) and, per layer, the true y and z at the x of the tracklet in the tracklet's sector frame, i.e. the frame of the stored TRD residual (nearest TRD track reference propagated with the workflow's material correction), for the true particle of the ITS-TPC part (ideal tracklets). TrackInterpolation records which ITS-TPC-TRD track gave the TRD residuals of each stored track (getTRDGIDsSuccess, aligned with the track data). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… sister tracks) For the particle of the ITS-TPC part TrackDataMC now holds the mother label and PDG code (for primaries the generator-level parent), the production vertex, momentum and process, a primary flag, and sisterIdx: the index of another stored track with the same mother (cycling through all of them if more than two), e.g. to select both daughters of a K0s. pdg is now the PDG code of the particle of the ITS-TPC part, consistent with the origin and TRD fields. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ree the baskets loadTracksForSourceAndEvent deep-copied the vector which ROOT allocated for us (a pointer to nullptr was passed, so it is ours) and deleted the original: +1x the event at peak (373 MB for a PbPb event with 6.5M MCTracks). The vector is now stored directly and the branch address reset. In addition, the decompressed baskets (~1x the event for the split MCTrack branch) are dropped once the event is the last entry of its TTree cluster, i.e. when no later entry can reuse them, which keeps small events sharing baskets as fast as before (QED, 10k events: 0.17 s either way, 0.70 s when dropping after every event). For the PbPb event: peak memory in use 1628 -> 1232 MB, held after releaseTracksForSourceAndEvent 882 -> 443 MB. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
matthias-kleiner
requested review from
davidrohr,
sawenzel,
shahor02 and
wiechula
as code owners
October 3, 2026 16:14
Contributor
Author
|
Hi @sawenzel , |
Collaborator
|
@matthias-kleiner : Ok for the MCKinematicsReader. But it appears that |
The track references of a source were loaded for all events at the first access and never released. They are now loaded per event on demand (with the baskets dropped at the end of a TTree cluster, as for the tracks) and releaseTracksForSourceAndEvent frees them together with the tracks. For a PbPb TF (10 signal events, 1.75M references): +304 MB held from the first access until the end before, now +135 MB at peak and +55 MB after releasing all events. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
Author
Thanks for spotting this. The track references are now also read per event on demand and released together with the tracks in releaseTracksForSourceAndEvent. For a PbPb TF this lowers the peak memory by about 170 MB (+304 MB before vs +135 MB) with no change in speed. |
shahor02
approved these changes
Oct 3, 2026
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.