GPU TPC: Tracking Developments #14542

davidrohr · 2025-07-24T12:13:22Z

Will take some time to be stable, but want to check in the CI already. Improvements so far:

Change track fit and track storage to operate on legs, not on looping tracks.
- This avoids problems in the fit when something goes wrong in one leg, not the rest of the track is lost.
Improvements to visualization and QA to benchmark it.
Optimize memory layouts to save memory, and adjust defaults.

In particular:

Change some default settings: consider only edge pad to flag as edge cluster, not 2 pads; reject clusters that come from IFC instead of giving them huge errors.
Improve weights-formula for ambiguity solving for cluster to track association.
Add some optional sanity checks to check that e.g. the track merge graph is non-cyclic. Since these do not trigger I believe the problems we had are fixed, thus removed workarounds for cyclic graphs.
Changed to track fit in projections by default, since there seem to be some problems in the non-projected fit - to be fixed later.
Some code cleanup of obsolete code.
Rejection of clusters based on interpolated track position moved to fit iterations 0 and 1, but rejection is only applied in 2.
fraction of allowed shared clusters is computed outside-in (instead of inside-out), so we have a higher chance to allow shared clusters in the inner pad row, and fraction is more accurate there.
Legs of looping tracks are kept as individual track (segments). All segments are kept separate in the fit, and fit independently. This fixes several problems in the track fit and allows for better parallelization. We just keep the linked index, so we can recover the full track at any time. When storing the final O2 TPC track, only the primary leg is stored anyway, so nothing changes here.
Fix some bugs in extrapolation of looping tracks to find clusters nearby.
Some improvements to event display and QA.

github-actions · 2025-07-24T12:13:30Z

REQUEST FOR PRODUCTION RELEASES:
To request your PR to be included in production software, please add the corresponding labels called "async-" to your PR. Add the labels directly (if you have the permissions) or add a comment of the form (note that labels are separated by a ",")

+async-label <label1>, <label2>, !<label3> ...

This will add <label1> and <label2> and removes <label3>.

The following labels are available
async-2023-pbpb-apass4
async-2023-pp-apass4
async-2024-pp-apass1
async-2022-pp-apass7
async-2024-pp-cpass0
async-2024-PbPb-apass1
async-2024-ppRef-apass1
async-2024-PbPb-apass2
async-2023-PbPb-apass5

alibuild · 2025-07-25T00:37:26Z

Error while checking build/O2/fullCI_slc9 for c79a241 at 2025-07-25 07:33:

## sw/BUILD/o2checkcode-latest/log
--
========== List of errors found ==========
++ GRERR=0
++ grep -v clang-diagnostic-error error-log.txt
++ grep ' error:'
grep: error-log.txt: binary file matches
/sw/SOURCES/O2/14542-slc9_x86-64/0/Framework/Core/src/StepTHn.cxx:330:20: error: statement should be inside braces [readability-braces-around-statements]
++ [[ 0 == 0 ]]
++ exit 1
--

Full log here.

alibuild · 2025-07-25T12:26:43Z

Error while checking build/O2/fullCI_slc9 for 3faeb33 at 2025-07-25 19:30:

## sw/BUILD/O2-latest/log
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMSectorTrack.cxx(278): error: explicit specialization of function "o2::gpu::GPUTPCGMSectorTrack::TransportToX(o2::gpu::GPUTPCGMMerger *, float, float, o2::gpu::GPUTPCGMBorderTrack &, float, __nv_bool) const [with I=2]" must precede its first use (at line 1741 of /sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMMerger.cxx)
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMSectorTrack.cxx(278): error: explicit specialization of function "o2::gpu::GPUTPCGMSectorTrack::TransportToX(o2::gpu::GPUTPCGMMerger *, float, float, o2::gpu::GPUTPCGMBorderTrack &, float, __nv_bool) const [with I=2]" must precede its first use (at line 1741 of /sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMMerger.cxx)
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMSectorTrack.cxx(278): error: explicit specialization of function "o2::gpu::GPUTPCGMSectorTrack::TransportToX(o2::gpu::GPUTPCGMMerger *, float, float, o2::gpu::GPUTPCGMBorderTrack &, float, __nv_bool) const [with I=2]" must precede its first use (at line 1741 of /sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMMerger.cxx)
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMSectorTrack.cxx(278): error: explicit specialization of function "o2::gpu::GPUTPCGMSectorTrack::TransportToX(o2::gpu::GPUTPCGMMerger *, float, float, o2::gpu::GPUTPCGMBorderTrack &, float, __nv_bool) const [with I=2]" must precede its first use (at line 1741 of /sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMMerger.cxx)
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMSectorTrack.cxx(278): error: explicit specialization of function "o2::gpu::GPUTPCGMSectorTrack::TransportToX(o2::gpu::GPUTPCGMMerger *, float, float, o2::gpu::GPUTPCGMBorderTrack &, float, __nv_bool) const [with I=2]" must precede its first use (at line 1741 of /sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMMerger.cxx)
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMSectorTrack.cxx(278): error: explicit specialization of function "o2::gpu::GPUTPCGMSectorTrack::TransportToX(o2::gpu::GPUTPCGMMerger *, float, float, o2::gpu::GPUTPCGMBorderTrack &, float, __nv_bool) const [with I=2]" must precede its first use (at line 1741 of /sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMMerger.cxx)
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMSectorTrack.cxx(278): error: explicit specialization of function "o2::gpu::GPUTPCGMSectorTrack::TransportToX(o2::gpu::GPUTPCGMMerger *, float, float, o2::gpu::GPUTPCGMBorderTrack &, float, __nv_bool) const [with I=2]" must precede its first use (at line 1741 of /sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMMerger.cxx)
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMSectorTrack.cxx(278): error: explicit specialization of function "o2::gpu::GPUTPCGMSectorTrack::TransportToX(o2::gpu::GPUTPCGMMerger *, float, float, o2::gpu::GPUTPCGMBorderTrack &, float, __nv_bool) const [with I=2]" must precede its first use (at line 1741 of /sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMMerger.cxx)
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMSectorTrack.cxx(278): error: explicit specialization of function "o2::gpu::GPUTPCGMSectorTrack::TransportToX(o2::gpu::GPUTPCGMMerger *, float, float, o2::gpu::GPUTPCGMBorderTrack &, float, __nv_bool) const [with I=2]" must precede its first use (at line 1741 of /sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMMerger.cxx)
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMSectorTrack.cxx(278): error: explicit specialization of function "o2::gpu::GPUTPCGMSectorTrack::TransportToX(o2::gpu::GPUTPCGMMerger *, float, float, o2::gpu::GPUTPCGMBorderTrack &, float, __nv_bool) const [with I=2]" must precede its first use (at line 1741 of /sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMMerger.cxx)
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMSectorTrack.cxx(278): error: explicit specialization of function "o2::gpu::GPUTPCGMSectorTrack::TransportToX(o2::gpu::GPUTPCGMMerger *, float, float, o2::gpu::GPUTPCGMBorderTrack &, float, __nv_bool) const [with I=2]" must precede its first use (at line 1741 of /sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMMerger.cxx)
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMSectorTrack.cxx(278): error: explicit specialization of function "o2::gpu::GPUTPCGMSectorTrack::TransportToX(o2::gpu::GPUTPCGMMerger *, float, float, o2::gpu::GPUTPCGMBorderTrack &, float, __nv_bool) const [with I=2]" must precede its first use (at line 1741 of /sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMMerger.cxx)
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMSectorTrack.cxx(278): error: explicit specialization of function "o2::gpu::GPUTPCGMSectorTrack::TransportToX(o2::gpu::GPUTPCGMMerger *, float, float, o2::gpu::GPUTPCGMBorderTrack &, float, __nv_bool) const [with I=2]" must precede its first use (at line 1741 of /sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMMerger.cxx)
ninja: build stopped: subcommand failed.

Full log here.

alibuild · 2025-07-25T22:44:34Z

Error while checking build/O2/fullCI_slc9 for 94013b9 at 2025-07-26 00:44:

## sw/BUILD/o2checkcode-latest/log
--
========== List of errors found ==========
++ GRERR=0
++ grep -v clang-diagnostic-error error-log.txt
++ grep ' error:'
/sw/SOURCES/O2/14542-slc9_x86-64/0/Framework/Core/src/StepTHn.cxx:330:20: error: statement should be inside braces [readability-braces-around-statements]
++ [[ 0 == 0 ]]
++ exit 1
--

Full log here.

alibuild · 2025-07-26T07:58:20Z

Error while checking build/O2/fullCI_slc9 for 2ec5599 at 2025-07-26 09:58:

## sw/BUILD/o2checkcode-latest/log
--
========== List of errors found ==========
++ GRERR=0
++ grep -v clang-diagnostic-error error-log.txt
++ grep ' error:'
/sw/SOURCES/O2/14542-slc9_x86-64/0/Framework/Core/src/StepTHn.cxx:330:20: error: statement should be inside braces [readability-braces-around-statements]
++ [[ 0 == 0 ]]
++ exit 1
--

Full log here.

alibuild · 2025-07-31T11:32:24Z

Error while checking build/O2/fullCI_slc9 for e7b16f9 at 2025-07-31 13:32:

## sw/BUILD/O2-latest/log
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMTrackParam.cxx:761:17: error: expected initializer before 'ries'
/sw/SOURCES/O2/14542-slc9_x86-64/0/GPU/GPUTracking/Merger/GPUTPCGMTrackParam.cxx:762:10: error: 'maxTries' was not declared in this scope
ninja: build stopped: subcommand failed.

Full log here.

alibuild · 2025-08-01T20:06:41Z

Error while checking build/O2/fullCI_slc9 for fb63238 at 2025-08-25 03:36:

## sw/BUILD/o2checkcode-latest/log
--
========== List of errors found ==========
++ GRERR=0
++ grep -v clang-diagnostic-error error-log.txt
++ grep ' error:'
/sw/SOURCES/O2/14542-slc9_x86-64/0/Detectors/TOF/workflow/src/make-parameter-collection.cxx:66:3: error: annotate this function with 'override' or (rarely) 'final' [modernize-use-override]
++ [[ 0 == 0 ]]
++ exit 1
--

Full log here.

alibuild · 2025-08-26T20:43:29Z

Error while checking build/O2/fullCI_slc9 for 222fa1b at 2025-08-28 08:52:

## sw/BUILD/o2checkcode-latest/log
--
========== List of errors found ==========
++ GRERR=0
++ grep -v clang-diagnostic-error error-log.txt
++ grep ' error:'
/sw/SOURCES/O2/14542-slc9_x86-64/0/Detectors/TOF/workflow/src/make-parameter-collection.cxx:66:3: error: annotate this function with 'override' or (rarely) 'final' [modernize-use-override]
++ [[ 0 == 0 ]]
++ exit 1
--

Full log here.

alibuild · 2025-08-29T08:40:49Z

Error while checking build/O2/fullCI_slc9 for d971cac at 2025-08-29 10:40:

## sw/BUILD/o2checkcode-latest/log
--
========== List of errors found ==========
++ GRERR=0
++ grep -v clang-diagnostic-error error-log.txt
++ grep ' error:'
grep: error-log.txt: binary file matches
/sw/SOURCES/O2/14542-slc9_x86-64/0/Detectors/TOF/workflow/src/make-parameter-collection.cxx:66:3: error: annotate this function with 'override' or (rarely) 'final' [modernize-use-override]
++ [[ 0 == 0 ]]
++ exit 1
--

Full log here.

davidrohr · 2025-09-04T13:51:08Z

Marking as draft: had passed the CI, will collect future tracking changes in separate PR.
This contains all the changes to the track fit for now. Once this is validated, it can be merged in principle, but other changes will come in few weeks.

…ssages to the console

…Position to avoid killing some tracks when some first clusters are bad

…more shared hits at inner rows

…ing of extrapolated track segments

…e results

…k Fit

…cked, but should be removed eventually)

alibuild · 2025-09-11T10:24:02Z

Error while checking build/O2/fullCI_slc9 for ac854e3 at 2025-09-15 17:36:

## sw/BUILD/o2checkcode-latest/log
--
========== List of errors found ==========
++ GRERR=0
++ grep -v clang-diagnostic-error error-log.txt
++ grep ' error:'
grep: error-log.txt: binary file matches
error: error: clang-tidy: Unknown command line argument '--load /sw/slc9_x86-64/o2codechecker/v20.1.7-local2/lib/libclangTidyAliceO2Module.so --extra-arg=--gcc-install-dir='.  Try: '/sw/slc9_x86-64/Clang/v20.1.7-1/bin-safe/clang-tidy --help'
error: error: clang-tidy: Unknown command line argument '--load /sw/slc9_x86-64/o2codechecker/v20.1.7-local2/lib/libclangTidyAliceO2Module.so --extra-arg=--gcc-install-dir='.  Try: '/sw/slc9_x86-64/Clang/v20.1.7-1/bin-safe/clang-tidy --help'
++ [[ 0 == 0 ]]
++ exit 1
--

Full log here.

alibuild mentioned this pull request Jul 24, 2025

Please consider the following formatting changes to #14542 davidrohr/AliceO2#125

Closed

davidrohr force-pushed the dev branch from c79a241 to 3faeb33 Compare July 25, 2025 10:58

davidrohr force-pushed the dev branch from 3faeb33 to 94013b9 Compare July 25, 2025 17:27

alibuild mentioned this pull request Jul 25, 2025

Please consider the following formatting changes to #14542 davidrohr/AliceO2#126

Open

davidrohr force-pushed the dev branch from 2ec5599 to e7b16f9 Compare July 31, 2025 08:48

davidrohr requested review from a team, shahor02 and wiechula as code owners July 31, 2025 08:48

davidrohr force-pushed the dev branch from e7b16f9 to fb63238 Compare July 31, 2025 14:31

davidrohr force-pushed the dev branch from fb63238 to 222fa1b Compare August 26, 2025 14:31

davidrohr force-pushed the dev branch from 222fa1b to d971cac Compare August 29, 2025 07:54

davidrohr force-pushed the dev branch 3 times, most recently from 62dee1a to 9ee262c Compare September 4, 2025 10:45

davidrohr marked this pull request as draft September 4, 2025 13:50

davidrohr mentioned this pull request Sep 4, 2025

GPU TPC: Tracking Developments - Part 2 - DO NOT MERGE #14651

Open

davidrohr added 4 commits September 10, 2025 19:57

GPU QA: Add track t0 QA

431ade5

GPU Display: Block until display started and print meaningful info me…

1399253

…ssages to the console

GPU: Use [[maybe_unused]] to silence compiler warnings

9f359ee

GPU: Improve timing messages for GPU Display and GPU QA

bd684b6

GPU QA: Proper fix for fetching timebins of MC data

06b06b9

davidrohr force-pushed the dev branch from d78f81c to db75072 Compare September 10, 2025 21:48

davidrohr added 21 commits September 11, 2025 08:23

GPU TPC: Implement looper following with propagator for segmented tracks

a9d0c1e

GPU: Add comments for customizable kernel parameters

870ea2e

GPU TPC: Require minimum NDF for mergerInterpolateRejectAlsoOnCurrent…

ddcbe95

…Position to avoid killing some tracks when some first clusters are bad

GPU TPC: Make mergerNonInterpolateRejectMinNDF configurable

b4b2a69

GPU TPC: TrackletSelection: Count shared hits from outside, to allow …

1aa6fa3

…more shared hits at inner rows

GPU QA: inputHistogramsOnly inplies noEvents

ef6e9a0

GPU TPC: Better formula for cluster weights

0cb5fc5

GPU: Improve debug dumps

8494996

GPU TPC: Deterministic (and faster since not relying on atomics) link…

cf38ef2

…ing of extrapolated track segments

GPU TPC: Do not interpolate with too few NDF

f094c9f

GPU TPC: Don't constrain SinPhi between inward/outward refits

2cab32d

GPU: Improve some debug messages

64491cd

GPU: Remove obsolete code

46d2447

GPU TPC: Fix setting of FitInProjection and PropagateBzOnly

0cc2a8f

GPU TPC: Fix applying tpc.trackFitCovLimit

518ddf4

GPU TPC: Do Interpolation rejection in TrackParam.cxx

22ff2c5

GPU: Temporarily disable with without projections since it gives wors…

3818b8d

…e results

GPU: Improve some debug messages

9e34c7a

GPU TPC: Make Looper Merging Afterburner work with new Segmented Trac…

37030d8

…k Fit

GPU TPC: Avoid some code duplication

c426151

GPU TPC: Make workarounds for cyclic merge graphs optional (to be che…

138c794

…cked, but should be removed eventually)

davidrohr force-pushed the dev branch from db75072 to 3f85db6 Compare September 11, 2025 06:56

GPU: Add some more optional sanity checks

ac854e3

davidrohr force-pushed the dev branch from 3f85db6 to ac854e3 Compare September 11, 2025 07:22

davidrohr marked this pull request as ready for review September 11, 2025 07:22

davidrohr changed the title ~~GPU TPC: Tracking Developments - DO NOT MERGE~~ GPU TPC: Tracking Developments Sep 14, 2025

davidrohr merged commit ac46fff into AliceO2Group:dev Sep 16, 2025
10 of 11 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

GPU TPC: Tracking Developments #14542

GPU TPC: Tracking Developments #14542

Uh oh!

davidrohr commented Jul 24, 2025 •

edited

Loading

Uh oh!

github-actions bot commented Jul 24, 2025

Uh oh!

alibuild commented Jul 25, 2025 •

edited

Loading

Uh oh!

alibuild commented Jul 25, 2025 •

edited

Loading

Uh oh!

alibuild commented Jul 25, 2025

Uh oh!

alibuild commented Jul 26, 2025

Uh oh!

alibuild commented Jul 31, 2025

Uh oh!

alibuild commented Aug 1, 2025 •

edited

Loading

Uh oh!

alibuild commented Aug 26, 2025 •

edited

Loading

Uh oh!

alibuild commented Aug 29, 2025

Uh oh!

davidrohr commented Sep 4, 2025 •

edited

Loading

Uh oh!

alibuild commented Sep 11, 2025 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

GPU TPC: Tracking Developments #14542

GPU TPC: Tracking Developments #14542

Uh oh!

Conversation

davidrohr commented Jul 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Jul 24, 2025

Uh oh!

alibuild commented Jul 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

alibuild commented Jul 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

alibuild commented Jul 25, 2025

Uh oh!

alibuild commented Jul 26, 2025

Uh oh!

alibuild commented Jul 31, 2025

Uh oh!

alibuild commented Aug 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

alibuild commented Aug 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

alibuild commented Aug 29, 2025

Uh oh!

davidrohr commented Sep 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

alibuild commented Sep 11, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

davidrohr commented Jul 24, 2025 •

edited

Loading

alibuild commented Jul 25, 2025 •

edited

Loading

alibuild commented Jul 25, 2025 •

edited

Loading

alibuild commented Aug 1, 2025 •

edited

Loading

alibuild commented Aug 26, 2025 •

edited

Loading

davidrohr commented Sep 4, 2025 •

edited

Loading

alibuild commented Sep 11, 2025 •

edited

Loading