ROOT Review planning

Europe/Zurich
32/1-A24 (CERN)

32/1-A24

CERN

40
Show room on map
Fons Rademakers (CERN)
Present: The ROOT Team, Marco, Peter, Benedikt, Andreas(R), Andrea (R), Bill (R), Valeri (R)

ACTIONS:
 o Alice/LHCb: If time allows provide some indication of the symbols responsible for the memory footprint increase
 o LHCb: If time allows, provide the pyROOT example which shows a slowdown wrt ROOT5
 o Danilo: provide to Philippe a way to easily reproduce the segfault issue in the cmssw environment

Summary:

o Peter
 - Reconstruction of Alice: Performance slightly better in 6
 - Results compatible: investigating further since overall compatibility is not enough when it comes to real physics
 - Memory allocation: RSS increment seen. Less than 10% overall but at startup non negligible (30%).
 - Some 200MB RSS more for reconstruction jobs, some ~70 MB come from the debugging symbols in libCling. To be further investigated
 - Valgrind Massif running now to profile memory usage

o Valeri
 - Terminate LBL contract: will not continue working on ROOT6
 - Track down why reflex-free build failed due this week, the last task

o Marco
 - Situation not so bad: Failures are known and being tackled (all reported in JIRA) also by Axel and Danilo
 - See also the memory usage increase: tests fail since they are also monitoring deltas of RSS during the run and printing on output if a certain threshold is exceeded.

o Bill
 - Fix problems but more coming out, not necessary ROOT issues, but it's like "peeling an onion". Solving issues is in any case (despite the new ones appearing).
 - - ~12 issues, but only one is known to be a problem in ROOT itself
 - The ROOT related one: compatibility issue ROOT5 files cannot be read with ROOT6: typedefs handled differently. A jira issue will be opened.

o Andreas:
 - Relocatability issue fixed
 - Not easy to reproduce in a completely standalone environment the intermittent segfault. See actions.

o Andrea:
 - The situation is much better with beta2
 - Summary on the slides
 - Overload resolution of templated methods seem to need c++ exceptions in pyROOT. This is not just the tests, but of the mechanism of resolution.
There are minutes attached to this event. Show them.
    • 16:00 16:05
      Minutes of Last Week 5m
      Speaker: Fons Rademakers (CERN)
    • 16:05 16:10
      Progress made by the ROOT team 5m
      Speaker: Axel Naumann (CERN)
      Slides
    • 16:10 16:20
      ALICE 10m
      Speaker: Peter Hristov (CERN)
    • 16:20 16:30
      ATLAS 10m
      Speaker: Dr Valeri Emil Fine (Brookhaven National Laboratory (US))
    • 16:30 16:40
      LHCb 10m
      Speaker: Marco Clemencic (CERN)
    • 16:40 16:50
      CMS 10m
      Speaker: Dr William Tanenbaum (Fermi National Accelerator Lab. (US))
    • 16:50 17:00
      COOL 10m
      Speaker: Dr Andrea Valassi (CERN)
      Slides