Architects Forum 3.11.2016 -DRAFT

Attendees

Pere Mato, Gabriele Cosmo, Andre Sailer,
Patricia Mendez Lorenzo, John Harvey,
Walter Lampl, Charles Delort,
Marco Clemencic, Marco Cattaneo, Gerardo Ganis,
Andrea Valassi, Benedikt Hegner

Next Meeting - 8.12.2016

News from EP-SFT

• New staff post in SFT - detector simulation - EM physics modelling. Open for 4 weeks
• HSF organized a community meeting on GeantV R&D last week.
  • More than 40 people in the room, and 10 connected remotely. 10 external reviewers, mainly experts in simulation and core software from the LHC experiments.
  • Very successful in promoting dialogue between developers and users.
  • 10 page report being finalised with O(100) detailed comments and recommendations - available soon.
  • Bottom-line - experiments would like to use what is being produced and want developers to ease the uptake of changes e.g. by integrating changes in future Geant4 releases and facilitating integration in the experiments data processing frameworks. 
• Pere Mato will take over as group leader of EP-SFT on Dec 1st.

Report from SPI/LIM

• Since the last AF meeting two minor LCG releases have been deployed for ATLAS use: LCG_85a and LCG_85b including new versions of the xrootd package that should solve some of the bugs reported by the experiment.
• Getting ready for the next LCG_86 major release that should be trigered as soon as the ROOT team provides a new ROOT tag for v6.08-patches (foreseen for tomorrow Friday). This release will put the emphasis in gcc62 builds which will be provided for both slc6 and centos7, while builds with gcc49 will be reduced to slc6 only as a testing platform. Ubuntu16 builds are also foreseen in this release. In addition, this release will contain an updated version of Boost (1.62), latest version of DD4HEP, and some extra packages requested by LHCb.
• dev2 slot in nightly builds is now stopped.
• dev3 is currently testing the latest ROOT 6.08-patches version (so dev4). dev3 will come back to the HEAD version of ROOT after the incoming release.
• Builds of dev3 with python3 is now completed. Backward compatibility changes have been now included in our master branch

News from Projects

Simulation

• 10.3 preparation;development release two days ago; last before final one in early december
  • Q: what are the major changes? --> improvements in physics, FTF hadronic model brought back to previous tuning (plus minor improvements)
  • porting to new compilers - clang/xcode/multithreading fixed
  • new EM models are faster and accurate, validation by experiments required

ROOT

• Quite a lot of preparations for CHEP, several presentations
  • Root general, Notebooks, SWAN service, C++ interface
• Closing the 6.08 release. A branch was made available for several weeks for experiments to check (was added to dev3 slot). Unfortunately, there is still a couple of problems just discovered this week.  Will try to tag tomorrow if we manage to fix them today.
  • Marco Clemencic: blockers for LHCb are fixed
• Features include:
  • ​Implicit multithreading, TMVA and ML, notebooks, Python3, JSROOT
  • new version of LLVM fixes many problems; ABI compatibility; new implicit multithreading

Q: time scale for Python3 in experiments?
 Marco Cattaneo: hired fellow in LHCb to look into Python3

CernVM

• new patched version of cvmfs; support for sierra needed fixes
• CentOS 7 overlay support needed kernel patch so far; in 7.3 will be gone
• 2.4 work ongoing
• October 1st - a technical fellow working on integration w/ docker started
• New cernvm-launch tool to start images; similar to dockers (work of technical student)
• since CHEP new version of ec2 image

CORAl/COOL

• made new tag 3.1.6 for next LCG release
  • bug fix release
• working on porting to ubuntu16
• tests on ubuntu failing due to a testing infrastructure problem

Experiments

ALICE

not present due to ALICE offline week

ATLAS

• 21 Tier0 release done; NN trainings etc in the process, so that reprocessing can be done
  • No major upgrade for the remainder of Run2 planned
• Migration to new LCG release in a two step procedure - first new release, then move to gcc 6.2

CMS

not present

LHCb

• new release of stack as soon as LCG release is out; polishing of Gaudi ongoing

CLIC

no news