Status of LIT tracking

Andrey Lebedev1,2

Claudia Höhne1

Ivan Kisel1

Gennady Ososkov2

1 GSI Helmholtzzentrum für Schwerionenforschung GmbH, Darmstadt, Germany2 Laboratory of Information Technologies, Joint Institute for Nuclear Research, Dubna, Russia

15th CBM Collaboration meetingApril 12-16, 2010 GSI, Darmstadt

2/11 A.Lebedev Status of LIT tracking 15th CBM Collaboration meeting

Muon system

TRD

3/11 A.Lebedev Status of LIT tracking 15th CBM Collaboration meeting

Track reconstruction algorithm

1km km1km

1kx

km

kxkx

kx

kx

kx

kx

1kkx

1kkx

1kx

0x

Two main steps:– Tracking

– Global track selection

Track propagation• Inhomogeneous magnetic field:

‒ solution of the equation of motion with the 4th order Runge-Kutta method• Large material budget: ‒ Energy loss (ionization: Bethe-Bloch, bremsstrahlung: Bethe-Heitler, pair production) ‒ Multiple scattering (Gaussian approximation)

Tracking is based on• Track following

• Initial seeds are tracks reconstructed in the STS (fast Cellular Automaton (CA), I.Kisel)

• Kalman Filter

• Validation gate

• Hit-to-track association techniques

‒ nearest neighbor: attaches the closest hit from the validation gate

Global track selection• aim: remove clone and ghost tracks

• tracks are sorted by their quality, obtained by chi-square and track length

• Check for shared hits

4/11 A.Lebedev Status of LIT tracking 15th CBM Collaboration meeting

Optimization of the algorithm• Minimize access to global memory

o Approximation of the 70 MB large magnetic field map

5 degree polynomial in the detector planes parabola between the stations

• Simplification of the detector geometryo Problem

Monte-Carlo geometry consists of 800000 nodes Geometry navigation based on ROOT TGeo

o Solution Create simplified geometry by converting Monte-Carlo geometry Implement fast geometry navigation for the simplified geometry

• Computational optimization of the Kalman Filtero From double to floato Implicit calculation on non-trivial matrix elementso Loop unrollingo Branches (if then else ..) have been eliminated

All these steps are necessary to implement SIMD tracking

5/11 A.Lebedev Status of LIT tracking 15th CBM Collaboration meeting

Parallelism

• Parallel programming is a mainstream! • SIMD – Single Instruction Multiple Data

• KF track fitter: fully SIMDized• NN track finder: partially SIMDized

• Multithreading• In event parallelization of the track finder• Event level parallelism

6/11 A.Lebedev Status of LIT tracking 15th CBM Collaboration meeting

Performance of the track fit

Computer with 2xCPUs Intel Core i7 (8 cores in total) at 2.67 GHz

Time [μs/track] Speedup

Initial 1200 -

Optimization 13 92

SIMDization 4.4 3

Multithreading 0.5 8.8

Final 0.5 2400

Residuals Pulls

X [cm]

Y [cm]

Tx *10-3

Ty *10-3

q/p *10-3

[GeV-1]

X Y Tx Ty q/p

0.38 0.39 9.1 8.7 3.4 1.02 0.99 1.08 1.08 0.92

Track fit quality

Speedup of the track fitter

Throughput: 2*106 tracks/s

7/11 A.Lebedev Status of LIT tracking 15th CBM Collaboration meeting

Performance of the parallel trackingSimulation:• 1000 UrQMD events at 25 AGeV Au-Au collisions + 5 μ+ and 5 μ- embedded in each event

Time [ms/event] Speedup

Initial 730 -

Optimization 7.2 101

SIMDization 4.8 1.5

Multithreading 1.5 3.3

Final 1.5 487

Initial version Parallel version

Efficiency [%] 94.7 94.0

Computer with 2xCPUs Intel Core i7 (8 cores in total) at 2.67 GHz

Speedup of the track finder

8/11 A.Lebedev Status of LIT tracking 15th CBM Collaboration meeting

Event level parallelism• Scheduler based on TBB tasks was used

• 1000 AuAu central events at 25 AGeV per task

• 1600 events per second can be reconstructed on lxir039

Computer lxir039 with 2xCPUs Intel Core i7 (8 cores in total) at 2.67 GHz

Further steps: start investigations with PROOF.

9/11 A.Lebedev Status of LIT tracking 15th CBM Collaboration meeting

MUCH1: 6x3 GEMs

MUCH2: 5x2+1x3 GEMs

MUCH3: 3x2 GEMs + 3x3 straw

tubes

MUCH4: 5x2 GEMs +

1st TRD station

MUCH5: 3x2 GEMs + 2x3 straw tubes+ 1st

TRD station

STS+MUCH efficiency for muons

91.6 92.1 91.9 90.8 84.6MUCH efficiency for muons

93.8 94.4 93.8 92.8 86.8

Physical analysis of those MUCH layout variants are needed to evaluate their advantages and feasibility

Test of different MUCH layouts

10/11 A.Lebedev Status of LIT tracking 15th CBM Collaboration meeting

10

position resolution [mum]position resolution [mum]

STS+TRD efficiency TRD efficiency

0.3 mm smearing in ”good” directionSmearing in ”bad” direction is different

allreference

allreference

resolution 0.3mm 0.5mm 1mm 3mm 5mm 1cm 3cm 5cm 10cm 30cm

TRD all 89.2 89.3 89.5 90.2 90.5 90.1 85.4 80.5 69.4 43.3

TRD ref 92.4 92.5 92.9 93.8 93.9 93.1 88.2 83.6 73.4 47.3

ghosts 5 6 7 9 11 15 35 58 109 232

TRD tracking efficiency

11/11 A.Lebedev Status of LIT tracking 15th CBM Collaboration meeting

Summary• Tracking is further optimized and tested on lxir039 computer

• First tests on event level parallelization shows its scalability

• TRD tracking was tested for different position resolution

• Different MUCH layouts were tested, further physics analysis are needed