Global Tracking Software Status

1
Global Tracking Software Status
H. Greenlee
Run II Meeting May 12, 2000
2
Outline

• Current tracking performance
• Efficiency
• Resolution
• Timing
• Global tracking review
• Tasks
• Milestones
• Active projects
• Future projects

3
Acceptance
Zmm 0 mb events
Found Tracks
Missed Tracks
4
Acceptance
Zmm 5 mb events
Found Tracks
Missed Tracks
5
Tracking Performance

• Current (t87) tracking performance can be found
  at the following web page
• http//www-d0.fnal.gov/global_tracking/results/lat
  est/

27apr00 t00.87.00 ----------

run bg ev CPU CLK mem swp tm rf
pe pTt mc eff misr fake fcsq mcsq nmsm ---- --
---- ---- ---- --- --- -- -- --- --- ----- ----
---- ---- ---- ---- ---- pzmm 0 1000 15 15 65 80
0 0 6 0.5 7714 .743 .051 .026 1.49 9.05 20.58
pzmm 0 1000 zmu 1625 .892 .017 .004 1.12 5.72
23.06 pzmm 1 1000 48 48 77 93 0 0 25 0.5 20151
.701 .078 .039 1.68 10.19 19.36 pzmm 1 1000 zmu
1560 .897 .026 .010 1.20 5.77 23.09 pzmm 2 1000
99 102 291 307 1 0 52 0.5 32826 .695 .091 .049
1.74 10.60 18.94 pzmm 2 1000 zmu 1593 .876 .034
.026 1.23 5.74 22.94 pzmm 3 1000 134 139 308 324
4 0 69 0.5 39223 .695 .096 .049 1.77 10.58 18.84
pzmm 3 1000 zmu 1599 .884 .038 .029 1.24 5.62
22.71 pzmm 4 1000 284 289 325 341 3 0 212 0.5
58888 .691 .115 .059 1.86 11.00 18.66 pzmm 4 1000
zmu 1581 .871 .033 .040 1.29 6.22 22.76 pzmm 5
1000 437 473 429 445 11 6 321 0.5 71383 .690 .133
.068 1.94 11.17 18.47 pzmm 5 1000 zmu 1598 .852
.057 .091 1.39 6.15 22.60 pzmm 4 250 281 282 261
277 1 0 44 0.5 14689 .686 .119 .058 1.85 11.08
18.66 pzmm 4 250 zmu 382 .869 .039 .042 1.33 6.37
23.19 pzmm 5 250 434 470 412 430 3 4 83 0.5 18092
.693 .135 .069 1.95 11.11 18.57 pzmm 5 250 zmu
401 .873 .063 .077 1.42 6.32 22.67
27apr00 t00.87.00 ----------

run bg ev CPU CLK mem swp tm rf
pe pTt mc eff misr fake fcsq mcsq nmsm ---- --
---- ---- ---- --- --- -- -- --- --- ----- ----
---- ---- ---- ---- ---- pzmm 0 1000 15 15 65 80
0 0 6 0.5 7714 .743 .051 .026 1.49 9.05 20.58
pzmm 0 1000 zmu 1625 .892 .017 .004 1.12 5.72
23.06 pzmm 1 1000 48 48 77 93 0 0 25 0.5 20151
.701 .078 .039 1.68 10.19 19.36 pzmm 1 1000 zmu
1560 .897 .026 .010 1.20 5.77 23.09 pzmm 2 1000
99 102 291 307 1 0 52 0.5 32826 .695 .091 .049
1.74 10.60 18.94 pzmm 2 1000 zmu 1593 .876 .034
.026 1.23 5.74 22.94 pzmm 3 1000 134 139 308 324
4 0 69 0.5 39223 .695 .096 .049 1.77 10.58 18.84
pzmm 3 1000 zmu 1599 .884 .038 .029 1.24 5.62
22.71 pzmm 4 1000 284 289 325 341 3 0 212 0.5
58888 .691 .115 .059 1.86 11.00 18.66 pzmm 4 1000
zmu 1581 .871 .033 .040 1.29 6.22 22.76 pzmm 5
1000 437 473 429 445 11 6 321 0.5 71383 .690 .133
.068 1.94 11.17 18.47 pzmm 5 1000 zmu 1598 .852
.057 .091 1.39 6.15 22.60 pzmm 4 250 281 282 261
277 1 0 44 0.5 14689 .686 .119 .058 1.85 11.08
18.66 pzmm 4 250 zmu 382 .869 .039 .042 1.33 6.37
23.19 pzmm 5 250 434 470 412 430 3 4 83 0.5 18092
.693 .135 .069 1.95 11.11 18.57 pzmm 5 250 zmu
401 .873 .063 .077 1.42 6.32 22.67
6
Efficiency Fakes
7
CPU Time by Step
8
Historical CPU Time
9
Global Tracking Review

• A review of global tracking software was
  conducted by Marc Paterno and Jim Kowalkowski,
  beginning on Feb. 4, 2000.
• The review report was made available on Feb. 29,
  2000.
• http//cdspecialproj.fnal.gov/d0/
• GlobalTracking/index.html
• Review and report concentrated on efficiency and
  speed issues.

10
Global Tracking Review Problems/Recommendations

• Problems/recommendations.
• Excessive copying.
• Return by value instead of by reference.
• Converting objects from edm to trf format.
• Use of activate/deactivate.
• Use of STL.
• Choice of container class (e.g. mapltgt).
• Using STL vector without reserve().
• STL for fixed size objects.
• Duplication of utilities (e.g. linear algebra).
• Trf ptrltgt smart pointer class.
• Can be made lighter weight.
• Get better locality of reference by storing
  reference count with object.

11
Problems/Recommendations (cont.)

• Parameters.
• Obs files have no bookkeeping.
• Migrate user settable parameters in rcp.
• Hard-coded constants.
• Profiling.

12
Profile Results

• Profiling was first successfully done in t86.
• Works on sgi or linux (using special build).
• Results
• One thing that was learned is that trf
  initialization was taking 25-50 of the cpu time,
  depending on event complexity.
• In t89, trf initialization overhead has been
  reduced by a factor of six by combining the six
  global tracking framework packages into a single
  framework package.

13
June Milestones
14
Active Projects

• Cosmic ray tracking (A. Goussiou).
• Track refitting/interacting propagator (V.
  Kuznetsov/D. Adams).
• Track-finding using 2D clusters in CFT (A.
  Kharchilava, J. Krane, H. Greenlee).
• Fixing high-pT inefficiency (H. Hildreth, B.
  Knuteson).
• Tracking in non-uniform magnetic field (O.
  Peters/K. Bos).
• SMT tracking optimization (S. Kulik).

15
Cosmic Ray Tracking
A. Goussiou

• Needed for cosmic ray commissioning. Not
  necessarily needed for production release.
• Trf modifications needed to allow propagation of
  tracks that traverse the detector (done).
• Next step is construction of cosmic tracking
  paths.

16
Track Refitting and Interacting Propagator
V. Kouznetsov

• Uses of interacting propagator.
• Refit to get optimal parameters at each end of
  track.
• Can get optimal fit parameters at any point
  outside the tracking volume by propagating (no
  refitting).
• Can get optimal parameters at any point inside
  the the tracking volume by refitting.
• Can better account for passive material and
  material that is not thin or not close to a
  sensitive surface.
• Status
• On by default for CFT central only in t89.
• SMT turned off because of speed.
• Will be turned on for all steps when speed
  problem is fixed (perhaps next release).

17
Parameter Resolution and Pulls at DCA
18
Parameter Resolution and Pulls at DCA
19
Parameter Resolution and Pulls at DCA
20
Parameter Resolution and Pulls at DCA
21
Parameter Resolution and Pulls at DCA
22
Chisquare at DCA
Fit c2/dof
Match c2
23
Parameter Resolution and Pulls at Outer Layer
24
Parameter Resolution and Pulls at Outer Layer
25
Parameter Resolution and Pulls at Outer Layer
26
Parameter Resolution and Pulls at Outer Layer
27
Parameter Resolution and Pulls at Outer Layer
28
Chisquare at Outer Layer
Refit fit c2/dof
Refit Match c2
29
Track finding in CFT using 2D clusters
A. Kharchilava, J. Krane

• Combine axial and stereo CFT clusters into 2D
  clusters (effectively space points).
• Already done in SMT.
• Should dramatically speed up tracking in overlap
  region by concentrating on CFT hits near ends of
  detector.
• Will also be tried in CFT central region may or
  may not speed things up.

30
CFT 2D Cluster Project Status

• Define class for CFT 2D cluster (released).
• Define CFT 2D cluster chunk (released).
• Write CFT 2D cluster framework package
  (released).
• Define TRF 2D cluster and hit classes (released).
• TRF CFT detector description in terms of 2D
  barrels (released).
• Convert CFT clusters into TRF clusters and add to
  TRF detector description (in cvs).
• Overlap region cluster filters (in cvs).
• Initial fitters (not done).
• TRF paths (not done).

31
Tracking in Non-uniform Magnetic Field
O. Peters, K. Bos

• cvs packages
• mag_field
• Contains abstract magnetic field interface and
  several concrete magnetic fields.
• tim
• Numerically solves equations of motion for
  non-uniform field motion using Runge-Kutta
  method.
• Borrowed from ATLAS.
• tim_interceptor, tim_interface
• These packages adapt ATLAS propagation code for
  D0 and trf interfaces.
• Status
• tim code can currently propagate track parameters
  and error matrices between any two surfaces used
  in D0 tracking (code in cvs).
• Still being verified.

32
Future Tasks

• Tracking with inefficiencies.
• Tracking with non-ideal geometry.
• Need to add hooks to tracking code.
• Do alignment.
• There will be a continuing effort at optimization
  and speed up.
• Tracking in gap region (between overlap and
  forward).
• Tracking in muon system and (possibly) other
  detectors.