Performance Timing Test

Did a timing test of Luca's analysis code on the Mac/SLC4 and on CERN/lxplus. I used the same source file, header file, and data file, and made five successive runs. The timing started in the "run" function of the code, before the histograms were created, and ended after all plots were made.

Mac/SLC4 2.2 GHz, 1 Gbyte (Linux) RAM, OSX.4, 128 Mbytes VRAM
Real Time (s) CPU Time (s)
93.4 19.0
43.2 16.7
45.5 17.1
41.0 16.6
41.2 16.8

Mac/SLC4 2.5 GHz, 1 GByte (Linux) RAM, OSX.5, 512 Mbytes VRAM
Real Time (s) CPU Time (s)
46.0 15.9
15.5 11.3
14.5 11.8
15.2 12.9
13.1 10.6

CERN/lxplus
Real Time (s) CPU Time (s)
203 14.8
194 14.1
198 13.6
199 14.9
201 14.2

Discussion: The slowness of lxplus Real Time was due to the time it takes to pop X windows over the internet, for the plots. The comparison is also somewhat uncontrolled since the loading of the two machines was different (see image below). It is really just a qualitative illustration that analysis on a laptop is feasible. The Mac CPU is only used at the 40% level for this version of the code, presumably the running is dominated by i/o since the analysis is pretty simple so far, not many calculations.

* Screen shot of Mac/SLC4 and CERN/lxplus system loads:
timetest.gif

-- WillBrooks - 24 Jun 2008

Truth Containers in DPD files

Comment: Looking for "Truth" containers in Luca's 10,000 event file. Using the command checkFile.py, find:

Size: 341301.356 kb
Nbr Events: 10000

Mem Size Disk Size Size/Evt items (X) Container Name (X=Tree or Branch)
27645.582 kb 1457.248 kb 0.146 kb 10000 (T) DataHeader
606.968 kb 39.731 kb 0.004 kb 10000 (B) TruthParticleContainer_p5_SpclMC
594.233 kb 140.887 kb 0.014 kb 10000 (B) MissingET_p1_MET_RefFinal
6000.142 kb 1691.074 kb 0.169 kb 10000 (B) MuonContainer_p1_StacoMuonCollection
28206.896 kb 2290.039 kb 0.229 kb 10000 (B) EventInfo_p2_McEventInfo
10683.812 kb 2291.983 kb 0.229 kb 10000 (B) Rec::TrackParticleContainer_tlp1_StacoTrackParticles
3264.956 kb 2323.832 kb 0.232 kb 10000 (B) MissingEtTruth_p1_MET_Truth
9226.639 kb 5336.529 kb 0.534 kb 10000 (B) TrackRecordCollection_p2_MuonEntryLayerFilter
25721.989 kb 6564.581 kb 0.656 kb 10000 (B) ElectronContainer_p1_ElectronAODCollection
32789.641 kb 12236.253 kb 1.224 kb 10000 (B) egDetailContainer_p1_egDetailAOD
155019.156 kb 25521.351 kb 2.552 kb 10000 (B) ParticleJetContainer_p1_Cone4H1TopoParticleJets
192143.969 kb 38237.322 kb 3.824 kb 10000 (B) ParticleJetContainer_p1_Cone4TruthParticleJets
154408.950 kb 56172.657 kb 5.617 kb 10000 (B) Rec::TrackParticleContainer_tlp1_TrackParticleCandidate
323847.991 kb 84230.634 kb 8.423 kb 10000 (B) TrigDec::TrigDecision_p1_TrigDecision
272859.932 kb 97964.390 kb 9.796 kb 10000 (B) McEventCollection_p3_GEN_DPD
1243020.856 kb 336498.511 kb 33.650 kb 10000 TOTAL (POOL containers)

Found ./ atlas/ PhysicsAnalysis/ TruthParticleID/ McParticleEvent/ src/ TruthParticleContainer.cxx . Contains, e.g.,

TruthParticleContainer::TruthParticleContainer( const TruthParticleContainer& rhs ) : DataVector( rhs ), m_genEvent ( rhs.m_genEvent ), m_etIsolations( rhs.m_etIsolations ), m_particles ( rhs.m_particles ) {}

How to set branch address? Tried:

#include "McParticleEvent/TruthParticleContainer.h"
...
TBranch* br_mc;
TruthParticleContainer* MC;
trans->SetBranchAddress("SpclMC",&MC,&br_mc);

and it compiled

How to look at root file directly with TBrowser? Get "no dictionary" warnings, e.g.,

Warning in <TClass::TClass>: no dictionary for class Navigable<JetCollection,double,vector<pair<ElementLink<JetCollection,DataProxyStorage,ForwardIndexingPolicy >,double> > > is available Warning in <TClass::TClass>: no dictionary for class Navigable<Rec::TrackParticleContainer,double,vector<pair<ElementLink<Rec::TrackParticleContainer,DataProxyStorage<Rec::TrackParticleContainer>,ForwardIndexingPolicy<Rec::TrackParticleContainer> >,double> > > is available Warning in <TClass::TClass>: no dictionary for class Navigable<ElectronContainer,double,vector<pair<ElementLink<ElectronContainer,DataProxyStorage,ForwardIndexingPolicy >,double> > > is available Warning in <TClass::TClass>: no dictionary for class Navigable<Analysis::MuonContainer,double,vector<pair<ElementLink<Analysis::MuonContainer,DataProxyStorage<Analysis::MuonContainer>,ForwardIndexingPolicy<Analysis::MuonContainer> >,double> > > is available

-- WillBrooks - 29 Jul 2008

Topic attachments
I Attachment History Action SizeSorted ascending Date Who Comment
GIFgif timetest.gif r1 manage 63.2 K 2008-06-24 - 18:29 WillBrooks Screen shot of Mac/SLC4 and CERN/lxplus system loads
Edit | Attach | Watch | Print version | History: r5 < r4 < r3 < r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r5 - 2008-07-30 - WillBrooks
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback