GEANT4 performance studies

Slides:



Advertisements
Similar presentations
1 ttbar events with Atlantis   A lot of jets in this event !   Bjets (1 and 2) close in phi   Muon from semilep decay of b-jet1   Tau jet   Pb.
Advertisements

Test with cosmic rays at LNGS - final results - S. Bastianelli, L. Degli Esposti, R.Diotallevi, G. Mandrioli, P. Righini, G. Rosa, M. Sioli OPERA meeting,
GAMOS tutorial Histogram and Scorers Exercises
June 6 th, 2011 N. Cartiglia 1 “Measurement of the pp inelastic cross section using pile-up events with the CMS detector” How to use pile-up.
Pion yield studies for proton drive beams of 2-8 GeV kinetic energy for stopped muon and low-energy muon decay experiments Sergei Striganov Fermilab Workshop.
1 Vertex fitting Zeus student seminar May 9, 2003 Erik Maddox NIKHEF/UvA.
Increasing Field Integral between Velo and TT S. Blusk Sept 02, 2009 SU Group Meeting.
ACAT ’02 CMS GEANT4 Sim. & Detector Desc. Pedro Arce(CERN/CIEMAT) 1 Simulation Framework and XML Detector Description for the CMS Experiment ARCE Pedro.
D 0  K -,  + reconstruction with CBM STS detector I.Vassiliev (GSI) CBM collaboration meeting 06-Oct-04 Simulation tools (cbmroot) & geometry Signal.
G EANT 4 energy loss of protons, electrons and magnetic monopole M. Vladymyrov.
HPS Test Run Setup Takashi Maruyama SLAC Heavy Photon Search Collaboration Meeting Thomas Jefferson National Accelerator Facility, May 26-27,
User Documents and Examples I Sébastien Incerti Slides thanks to Dennis Wrigth, SLAC.
1 Luminosity monitor and LHC operation H. Burkhardt AB/ABP, TAN integration workshop, 10/3/2006 Thanks for discussions and input from Enrico Bravin, Ralph.
Tracking at LHCb Introduction: Tracking Performance at LHCb Kalman Filter Technique Speed Optimization Status & Plans.
Geometry and Field: New features and Developments T. Nikitina For Geometry Working Group Geant4 workshop, Hebden Bridge september 2007 G4Paraboloid.
JUNE 1997PHOBOS Review Mark D. Baker PHOBOS Review Physics Simulations Mark D. Baker Physics PHOBOS design is sound. –( & PHOBOS Computing Project is on.
Pedro Arce Introducción a GEANT4 1 GAMOS tutorial Plug-in’s Exercises Pedro Arce Dubois CIEMAT
Field propagation in Geant4 John Apostolakis 25 th May 2005 Ver. 
Field propagation in Geant4 John Apostolakis, CERN Ecole Geant June 2007, Paris 25 th May 2005 Ver. 
Detector Description: Sensitive Detector & Field
Hadronic Interaction Studies for LHCb Nigel Watson/Birmingham [Thanks to Silvia M., Jeroen v T.]
Oct. 22, G.Lima1 Delving Deeper into Geant4 Guilherme Lima DHCal Meeting October 22, 2003.
1 Status and Plans for Geant4 Physics Linear Collider Simulation Workshop III 2-5 June 2004 Dennis Wright (SLAC)
Beam MC activity A.K.Ichikawa for beam group For more details,
1 Performance of a Magnetised Scintillating Detector for a Neutrino Factory Scoping Study Meeting Rutherford Appleton Lab Tuesday 25 th April 2006 M. Ellis.
G4 Validation meeting (5/11/2003) S.VIRET LPSC Grenoble Photon testbeam Data/G4 comparison  Motivation  Testbeam setup & simulation  Analysis & results.
G. BrunoOffline week - February Comparison between test- beam data and the SPD simulations in Aliroot G. Bruno, R. Santoro Outline:  strategy of.
Geant4 Simulation of the Beam Line for the HARP Experiment M.Gostkin, A.Jemtchougov, E.Rogalev (JINR, Dubna)
Tracking ions inside PRISMA E.Farnea INFN Sezione di Padova.
Photon reconstruction and matching Prokudin Mikhail.
Geant4 Tracking Test (D. Lunesu)1 Daniela Lunesu, Stefano Magni Dario Menasce INFN Milano GEANT4 TRACING TESTs.
Carbon Target Design and Optimization for an Intense Muon Source X. Ding, UCLA H.G. Kirk, BNL K.T. McDonald, Princeton Univ MAP Winter Collaboration.
Geant4 CPU performance : an update Geant4 Technical Forum, CERN, 07 November 2007 J.Apostolakis, G.Cooperman, G.Cosmo, V.Ivanchenko, I.Mclaren, T.Nikitina,
Adele Rimoldi, Pavia University & INFN – CERN G4usersWorkshop Nov H8 Muon Testbeam Simulation CERN - 14 November, 2002 and the Physics Validation.
Pedro Arce Introducción a GEANT4 1 GAMOS tutorial RadioTherapy Exercises Pedro Arce Dubois CIEMAT
Feb 24, Abnormal Events in HF: TB04, Simulation, and Feb.08 Fermi Testbeam Anthony Moeller (U. Iowa) Shuichi Kunori (U. Maryland) Taylan Yetkin (U.
07/24/07 Francisco Carrion OPTIMIZING THE TRACK DETECTOR Francisco Javier Carrión Ruiz Mentor: Hans Wenzel 07/08/09 SID CONCEPT FOR THE INTERNATIONAL LINEAR.
Charged Particle Multiplicity, Michele Rosin U. WisconsinQCD Meeting May 13, M. Rosin, D. Kçira, and A. Savin University of Wisconsin L. Shcheglova.
Tomas Hreus, Pascal Vanlaer Study of Strangeness Production in Underlying Event at 7 TeV 1QCD low pT meeting, 18/03/2011.
ST Occupancies (revisited) M. Needham EPFL. Introduction Occupancies matter Date rates/sizes In particular was data size on links from Tell1 to farm estimated.
Status of the Simulations on Photo Injector Optimization for Low Charges Yauhen Kot BD Meeting,
2005/07/12 (Tue)8th ACFA Full simulator study of muon detector and calorimeter 8th ACFA Workshop at Daegu, Korea 2005/07/12 (Tue) Hiroaki.
Eunil Won/Korea U1 A study of configuration for silicon based Intermediate Trackers (IT) July Eunil Won Korea University.
N_TOF EAR-1 Simulations The “γ-flash” A. Tsinganis (CERN/NTUA), C. Guerrero (CERN), V. Vlachoudis (CERN) n_TOF Annual Collaboration Meeting Lisbon, December.
Progress on Simulation Software Hai-Ping Peng(USTC) Xiao-Shuai Qin(IHEP) Xiao-Rong Zhou(USTC) Yu Hu(IHEP) 2014 STC Workshop (ITP) Hai-Ping Peng.
CHEP ’06 GEANT4E 1 GEANT4E: Error propagation for track reconstruction inside the GEANT4 framework Pedro Arce (CIEMAT) CHEP 2006, Mumbai, 13-17th February.
Monthly video-conference, 18/12/2003 P.Hristov1 Preparation for physics data challenge'04 P.Hristov Alice monthly off-line video-conference December 18,
Dmitry Sorokin (GSoC 2016), John Apostolakis (CERN)
Beam Energy-Loss measurement
X. Ding, UCLA MAP Spring 2014 Meeting May 2014 Fermilab
Upgrade Tracker Simulation Studies
Use of interpolation and FSAL property for integration in field
Systematic uncertainties in MonteCarlo simulations of the atmospheric muon flux in the 5-lines ANTARES detector VLVnT08 - Toulon April 2008 Annarita.
The LHC collider in Geneva
(CMS GEANT4 simulation)
STAR Geometry and Detectors
Damping Ring Kicker Tests at AØ
Field propagation in Geant4
Pedro Arce (CERN/CIEMAT)
Pedro Arce (CERN/CIEMAT) (on behalf of CMS collaboration)
Field propagation in Geant4
The Hadrontherapy Geant4 advanced example
The n-3He Simulation Using Geant4
Summary of dE/dx studies in silicon and MS in muon system
Field propagation in Geant4
GAMOS tutorial Plug-in’s Exercises
Field propagation in Geant4
Use of GEANT4 in CMS The OSCAR Project
Physics event timing Use Pythia to generate hadronic decays at 125 GeV
MEIC Alternative Design Part III
Presentation transcript:

GEANT4 performance studies Pedro Arce (CERN/CIEMAT) GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

GEANT4 preformance studies Pedro Arce (CERN/CIEMAT) Outline GEANT3 vs GEANT4 performance for full events Where time is spent Tracking speed Tuning GEANT4 tracking in field parameters GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

GEANT4 vs GEANT3 Time Performance - Full events in full CMS: H / tte / Z - Cuts in primary particles: Pt : 1 GeV, || < 2.4 (3.0) - Full CMS geometry - 3D TOSCA magnetic field - Production and tracking cuts as in CMSIM (GEANT3) - GEANT4 voxel navigation - GEANT44.4.0.ref02 - Pentium III 850 Mhz - Checked that they are really the same events (see next) GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

GEANT4 preformance studies Pedro Arce (CERN/CIEMAT) G3-G4 events detailed comparison H in all CMS (10 Events) Initial track energy No Tracks MeV Track length No Tracks mm GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

GEANT4 preformance studies Pedro Arce (CERN/CIEMAT) G3-G4 events: detailed comparison Origin of secondary tracks: R vs Z R (mm) CMSIM (442828 tracks) Z (mm) R (mm) OSCAR (427136 tracks) Z (mm) GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

GEANT4 preformance studies Pedro Arce (CERN/CIEMAT) Time performance - H (10 events) CMSIM: 439k tracks 64 sec/evt OSCAR: 400k tracks 117 sec/evt (1.83) - tte (10 events) CMSIM: 1809k tracks 184 sec/evt OSCAR: 1159k tracks 425 sec/evt (2.31) - Z (10 events) CMSIM: 1105k tracks 96 sec/evt OSCAR: 941k tracks 248 sec/evt (2.58) GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

GEANT4 preformance studies Pedro Arce (CERN/CIEMAT) Where time is spent? Profile (´gprof´)10 full pp events in full CMS (GEANT4.3.2.ref02) (numbers are accumulated: spent in the method itself or those invoked) 100% main 1,8% G4RunManager::RunInitialization 0.8% G4SteppingManager::SetInitialStep 0.3% G4ProcessManager::EndTracking 0.2% G4ProcessManager::StartTracking 95,7% G4TrackingManager::ProcessOneTrack 95,7% G4SteppingManager::Stepping 41% G4SteppingManager::DefinePhysicalStepLength 22% G4VProcess::AlongStepGPIL 17% G4VProcess::PostStepGPIL 26% G4SteppingManager::InvokeAlongStepDoItProcs 20% G4SteppingManager::InvokePostStepDoItProcs GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

GEANT4 preformance studies Pedro Arce (CERN/CIEMAT) Where time is spent? 41% G4SteppingManager::DefinePhysicalStepLength 22% G4VProcess::AlongStepGPIL 18% G4Transportation::AlongStepGetPhysicalInteractionLength 4% G4MultipleScattering::AlongStepGetPhysicalInteractionLength 17% G4VProcess::PostStepGPIL 14% G4VContinuousDiscreteProcess::PostStepGetPhysicalInterac-tionLength 1% G4VDiscreteProcess::PostStepGetPhysicalInteractionLength 1% G4VRestDiscreteProcess::PostStepGetPhysicalInteraction-Length GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

GEANT4 preformance studies Pedro Arce (CERN/CIEMAT) Where time is spent? 26% G4SteppingManager::InvokeAlongStepDoItProcs 5% G4ParticleChangeForTransport::UpdateStepForAlongStep 5% G4ParticleChange::UpdateStepForAlongStep 3% G4VeEnergyLoss::AlongStepDoIt 3% G4Transportation::AlongStepDoIt 1% G4MultipleScattering::AlongStepDoIt 1% G4VhEnergyLoss::AlongStepDoIt 20% G4SteppingManager::InvokePostStepDoItProcs 5% G4MultipleScattering::PostStepDoIt 4% G4Step::UpdateTrack 4% G4Transportation::PostStepDoIt 2% G4ParticleChangeForMSC::UpdateStepForPostStep GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

GEANT4 preformance studies Pedro Arce (CERN/CIEMAT) Where time is spent? 25% TRANSPORTATION: 18% G4Transportation::AlongStepGetPhysicalInteractionLength 14% G4PropagatorInField::ComputeStep 8% G4ChordFinder::AdvanceChordLimited 3% G4Transportation::AlongStepDoIt 4% G4Transportation::PostStepDoIt 4.2% LOCATE IN GEOMETRY TREE 1.0% G4PropagatorInField::LocateIntersectionPoint 0.8% G4Navigator::LocateGlobalPointWithinVolume 1.5% G4Navigator::LocateGlobalPointAndUpdateTouchableHandle 1.0% G4Navigator::LocateGlobalPointAndSetup GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

GEANT4 preformance studies Pedro Arce (CERN/CIEMAT) Tracking speed: G3 vs G4 Track charged geantinos in CMS (muon w/o physics as there are no charged geantinos in GEANT3) 1000 tracks at random eta and phi Magnetic field OFF GEANT4 magnetic field is more precise... ... Can we make it less precise and get as fast as GEANT3? ... Or maybe GEANT3 is too unprecise for our needs? Magnetic field ON GEANT3 GEANT4 P=10000 GeV 13.7s 12.62s P=50 GeV P=5 GeV 12.57s GEANT3 GEANT4 P=10000 GeV 24.7s 88.2s P=50 GeV 26.0s 131.2s P=5 GeV 28.2s 174.3s GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

Tracking in field parameters GEANT4 gives you a big freedom to optimise magnetic field tracking. User can select: IntegratorStepper class G4FieldManager::SetDeltaChord(value) G4FieldManager::SetDeltaOneStep(value) G4FieldManager::SetDeltaIntersection(value) G4PropagatorInField::SetMinimumEpsilonStep(value) G4PropagatorInField()::SetMaximumEpsilonStep(value) G4IntegrationDriver()->SetHmin( value ) Select a default and Change systematically one parameter each time 1000 tracks of 50 GeV (same effects observed for 5 Gev & 10000 GeV) Integrator Stepper delta Chord delta OneStep delta Intersection minEpsilon maxEpsilon driverHmin G4ClassicalRK4 0.1 mm 0.25 mm 0.001 mm 1.E-5 0.05 0.01 GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

Tracking in field parameters deltaChord (mm): deltaOneStep (mm): deltaIntersection (mm): minEpsilon 1.E-5 1.E-4 1.E-3 1.E-2 0.1 1. 10. 281.2 151.7 131.2 164.7 320.8 641.6 1710.9 0.001 0.01 0.1 0.25 1. 10. 100. 321.1 321.4 320.8 321.0 316.5 170.8 0.001 0.01 0.1 1. 10. 320.8 306.5 293.7 294.7 295.3 1.E-3 1E-4 1E-5 1E-6 1E-7 134.7 172.7 320.8 630.9 799.9 GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

Tracking in field parameters maxEpsilon: driverHmin: Select some good parameters: and change integrator stepper: 1. 1E-1 0.05 1E-2 1E-3 1E-4 321.2 320.8 321.5 321.1 320.4 0.0001 0.001 0.01 0.1 1. 10. 295.5 295.6 296.4 296.1 291.8 delta Chord delta OneStep delta Intersection minEpsilon maxEpsilon driverHmin 0.01 mm 0.25 mm 0.001 mm 1.E-4 0.05 0.1 ClassicalRK4 CashKarpRKF45 HelixExplicitEuler HelixImplicitEuler HelixSimpleRunge SimpleHeum SimpleRunge 125.2 110.9 55.2 92.4 57.9 99.0 72.0 GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)

Tracking in field parameters Push it to the limit (although not clear how much the optimal value of a parameter depends on others): We can play with the parameters, but first we have to understand How much we can change the parameters without affecting the physics results? What are ´good physics results´ (is GEANT3 precision enough?) Long way to go... Integrator Stepper delta Chord delta OneStep delta Intersection minEpsilon maxEpsilon driverHmin HelixExplicitEuler 0.001 mm 100. mm 1.E-3 0.05 0.01 GEANT3 GEANT4 GEANT4 (opti?) P=10000 GeV 24.7s 88.2s 41.5s P=50 GeV 26.0s 131.2s 48.9s P=5 GeV 28.2s 174.3s 65.2s GEANT4 Workshop 02 GEANT4 preformance studies Pedro Arce (CERN/CIEMAT)