Online Track Reconstruction

in the CBM Experiment I. Kisel, I. Kulakov, I. Rostovtseva, I. Kisel, I. Kulakov, I. Rostovtseva, M. ZyzakM. Zyzak (for the CBM (for the CBM

Collaboration)Collaboration)E-mail: M.Zyzak@gsi.de@gsi.de

Deutsche Physikalische Gesellschaft e.V.

Münster 11

Tracking ChallengeTracking Challenge

Fixed-target heavy-ion experiment 107 collisions/s 1000 charged particles/collision Non-homogeneous magnetic field Track reconstruction and displaced

vertex search required in the first trigger level

Track Finder w.r.t. Detector Track Finder w.r.t. Detector InefficiencyInefficiency

Track Finder w.r.t. Detector Track Finder w.r.t. Detector InefficiencyInefficiency

Detector efficiency, %

100 97 95 90 85 80

x, μm 12 13 13 14 14 15

y, μm 57 60 61 65 69 73

tx ,mrad 0.35 0.36 0.37 0.38 0.40 0.42

ty ,mrad 0.60 0.61 0.61 0.63 0.64 0.66

p, % 1.22 1.25 1.28 1.34 1.41 1.48

• The algorithm is stable

• Slight efficiency degradation with detector efficiency decreasing

• Resolution of track parameters becomes slightly worse because of the smaller number of hits

Scalability of the Track FinderScalability of the Track FinderScalability of the Track FinderScalability of the Track Finder

2 CPUs Intel X5550, 4 cores per CPU, HT, 2.7 GHz

4 CPUs AMD E6164HE, 12 cores per CPU, 1.7 GHz(in collaboration with Julien Leduc/CERN openlab)

Strong many-core scalability for large groups of minimum bias events is observed.

ConclusionsConclusionsConclusionsConclusions• For track finding a CA based algorithm is used.

• The algorithm is fast and efficient.

• The algorithm is robust with respect to the detector inefficiency.

• The algorithm shows strong many-core scalability.

• The investigation of 4D reconstruction has been started.

Deterministic Annealing FilterDeterministic Annealing Filter11Deterministic Annealing FilterDeterministic Annealing Filter11

Hit displacement

unshifted 5 σhit 10 σhit 20 σhit

MVD 1 0.4 0.4 0.4 0.4

2 0.7 0.7 0.7 0.7

STS 1 0.3 0.3 0.3 0.3

2 0.4 0.4 0.4 0.4

3 0.4 0.7 0.8 0.5

4 0.5 43.9 85.0 98.7

5 0.5 1.6 1.6 0.8

6 0.6 0.6 0.6 0.6

7 0.6 0.6 0.6 0.6

8 0.1 0.1 0.1 0.1

Task: reduce an influence of attached distorted or noise hits on the reconstructed track parameters.

Percentage of rejected hits depending on the distance from the shifted hit on the 4th STS station to its Monte-Carlo position has been measured.

• A weight is introduced to each hit

• Algorithm is iterative• With each iteration estimation

of the hits weight is improved• Based on SIMD KF track fit

benchmark2

4D Reconstruction for the CBM 4D Reconstruction for the CBM ExperimentExperiment

4D Reconstruction for the CBM 4D Reconstruction for the CBM ExperimentExperiment

CBM will have:• Free streaming data• 4D measurements (x, y, z, t)• Track reconstruction prior

event recognition

First idealized 4D STS reconstruction with CA track finder has been investigated. Discrete time have been used.

The same efficiency Slight increase of the processing

time with larger size of the time slices

Will be further investigated within the CA track finder.

1 R. Frühwirth and A. Strandlie, Track Fitting with ambiguities and noise: a study of elastic tracking and nonlinear filters. Comp. Phys. Comm. 120 (1999) 197-214.2 S. Gorbunov, U. Kebschull, I. Kisel, V. Lindenstruth and W.F.J. Müller, Fast SIMDized Kalman filter based track fit, Comp. Phys. Comm. 178 (2008) 374-383

Track ReconstructionTrack ReconstructionTrack ReconstructionTrack Reconstruction

• Cellular Automaton (CA) based track finder algorithm

• Kalman filter track fit

• Highly optimized code

– Single precision calculations

– Magnetic field approximation

– Reconstruction in several iterations

• Highly parallelized code

– Data level (SIMD instructions, 4 single-precision floating point calculations in parallel)

– Task level (ITBB, parallelization between cores)

0. Hits

1. Segments

1 2 3 42. Counters

3. Track Candidates

4. Tracks

Detector layers

Hits

Cellular Automaton:1.Build short track segments2.Connect according to the

track model3.Tree structures appear, collect

segments into track candidates

4.Select the best track candidates

Cellular Automaton advantages:

• Local w.r.t. data• Intrinsically parallel• Extremely simple• Very fast• Perfect for many-core

CPU/GPU

Track Reconstruction EfficiencyTrack Reconstruction EfficiencyTrack Reconstruction EfficiencyTrack Reconstruction Efficiency

Efficiency and ratios, %

Reference set 97.8

All set 87.6

Clone 0.8

Ghost 12.8

Tracks/ev 733

Time/ev, s 1.4

All set: p ≥ 0.1 GeV/cReference set: p ≥ 1 GeV/cGhost: purity < 70%

Reconstructable track:Number of consecutive MC points ≥ 4

Computer with two Xeon X5550 processors at 2.7 GHz and 8 MB L3, 1 core is used.

Au+Au central events at 25 AGeV, 8 STS and 2 MVD stations. Au+Au central events at 25 AGeV, 8 STS stations.