Geant4 Profiling and Benchmarking

1) Profiling Tasks Current

2) Profiling Results

Geant4 Version Application Performance Summary
10.0 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
10.0.c01 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
10.0.c00 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.r10 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.r09a SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.r09 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.r08 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.r07O2g*** SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.r07 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.r06** SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.r05 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.r04 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.r03 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.r02* SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.r01 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.p02 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6.p01 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
9.6 SimplifiedCalo Simple Profiler Memory Profiler CPU MEM
*) the first version starting the multithread (MT) migration
**) the reference tag 9.6.r06 corresponds to the published 10.0-beta
***) switching the cxx flag from O2 -DNDEBUG to -O2 -g -fno-omit-frame-pointer

Old Profiling Results: 9.4 9.5

3) CPU per Event: Summary Plots by Versions

SimplifiedCalo PYTHIA H->ZZ electrons pions protons anti-protons

4) Total Memory Count: Summary Plots by Versions

SimplifiedCalo PYTHIA H->ZZ electrons pions protons anti-protons

5) CPU Summary Plots by Physics Lists: 9.6

6) Geant4 MT Performance

Geant4 Version Application Performance
10.0 cmsExpMT Summary Open|SpeedShop HPCToolkit
10.0.c01 cmsExpMT Summary Open|SpeedShop HPCToolkit
10.0.c00 cmsExpMT Summary Open|SpeedShop HPCToolkit
9.6.r10 cmsExpMT Summary Open|SpeedShop HPCToolkit
9.6.r09a cmsExpMT Summary Open|SpeedShop HPCToolkit
9.6.r09 cmsExpMT Summary Open|SpeedShop HPCToolkit
9.6.r08 cmsExpMT Summary Open|SpeedShop HPCToolkit
9.6.r07 cmsExpMT Summary Open|SpeedShop HPCToolkit
9.6.r06 cmsExpMT Summary Open|SpeedShop HPCToolkit
9.5.p01mt cmsExpMT Summary Open|SpeedShop HPCToolkit

7) Useful Links for Performance Tools

  1. Fast: FAST is a set of tools for collecting, managing, and analyzing data about code performance.
    Instructions for use of the FAST toolkit is available at FAST project page

  2. IgProf: Ignominous Profiler is a simple tool for measuring and analysing application memory and
    performance characteristics. For more information, see IgProf home page

  3. HPC Performance Tools: HPCToolkits Open|SpeedShop TAU

8) Other Test Results

  • Performance studies: 9.6
  • g4p_9.5.p01mt callers of __tls__get_addr