AF meeting 19th June 2014
Fabritzio Furano [replacing Micolo], John Harvey, Pere Mato,
John Harvey, Rolf Seuster, Marco Clemencic, Gerri Ganis, Marco Cattaneo,
Gabriele Cosmo, Andrea Valassi, Benedikt Hegner, Markus Elsing, Nicolo
Magini, Federico Carminati
- Action Items
- Line Management Issues (John)
- Report from the LIM meeting (Benedikt)
- News from projects
- Experiments Feedback
1 (20/06/2014 @ 9:00)
Management Issues (John)
- TS Committee is now but the SFT group
has no quota this time.
- SFT group provided a contact person for
the FCC computing/software activity
- Discussions on the future of Geant4
- A meeting of the Oversight Board will
be called by CERN management.
- Governance model will come to some
- Meeting with Springer
- Interested on publishing a open book
for students on software for HEP. Practical guide for software
development infrastructure, details of the software stack, different
domains and challenges.
- Contributions will be welcome.
- Reuse material from tutorials, courses
(i.e. CSC, GridKA), and hand-on exercises.
- HEP Software Foundation
- 9 papers has been submitted. John has
started to process them. A meeting will be called soon to present and
discuss the status.
- DIRAC consortium invited SFT to join a
H2020 proposal (INFRA-1 call)
- Interested on CernVM-FS on various
aspects to distribute the software. Developments will be needed (user
mode FS, encryption, etc.)
- AIDA-2 proposal
- Progress report will be given to the
SFT group meeting next week.
from the LIM meeting (Benedikt)
- Release LCG 69 is in the pipeline
- It will be a double stack w/ ROOT 5
and ROOT 6 flavours again.
- The ROOT 6 flavour to come first. The
tags for CORAL/COOL are still in the works and I cannot give an
estimate yet, when the release will be cut.
- Savannah to JIRA migration
- The SFT JIRA instance has a serious
incident. All the data could be recovered expect for the attachments.
Incorrect backup procedures has now been corrected for all
- LHCb, CMS, and ATLAS have been
contacted again recently by Ilias Goulas. The main contact people are
Marco Clemencic, David Lange and Rolf Seuster
- There is a new HEP-SPEC in preparation
and we've been asked for input
- We discussed it during the LIM (see
extract from minutes below 
- Discussed other possible flags (e.g.
sse2, std=c++11, ...) It was agreed to be conservative and not to
speedup the specmark unnecessarily with respect other typical
production jobs of the experiments.
- Agreed to communicate the LIM proposed
flags as final.
- ROOT 6 was released on the scheduled
day (Friday, May 30)
- It provides the all the available
functionality up to this date
- It does not provide the
parsing of headers on demand, which was one of the strategies
for reducing the increased memory consumption of ROOT6 that was
observed by the experiments.
- Released yesterday a patch for 6.00/01
- It fixes an installation problem
with --prefix, dynamic scope issues that affected some tutorials,
performance issues reported by CMS.
- It was requested to add this version
as part of LCG_69. In case it does not work properly we can revert
- Agreed of the release date for ROOT
6.02 that is September the 22nd with these features:
- No MCJIT
- Reduce memory load
- new llvm/clang version
- Later release versions and dates will
be decided soon
- The ROOT team have to see what
we can be added between September and November
- Fons is definitely departed from the
- Preparing the CORAL and COOL tags for
the ROOT5/ROOT6 based releases, but I still need a few days, apologies
to the experiments.
- The preparation of a new patch release
10.0.p02, which Geant4 team plan to release this week, hopefully
- It includes a series of fixes in
different areas, some of them based on feedback we could receive
from the experiments, particularly from CMS.
- The testing phase for the 10.1-beta
release is approaching the end, and we should be in time for having it
out by end of this month as planned.
- Patch releases has been made available
to the CernVM appliance (security releases)
- CernVM-FS 2.1.19. First release with
new functionality on discovering proxies.
- Version 2.0 has a vulnerability when
dealing with big files. Observed last week in the Genser repository.
This is during the copy of the shadow tree to the final repository
that can take very long and if interrupted it may leave
inconsistencies. This is corrected with version 2.1.
- The VO migration to version 2.1 will
be taking place from August 5th to September 2nd.
- LCG_69 being tested and have shown
some problems related to obsolete code. Andrea had removed
deprecated code and produced some problems because the cleanup didn't
- Producing releases for data challenge
- Nothing special.
- Testing LCG_69 (dev4 slot). A number
of hiccups is affecting the build and testing system. Soon they will
have some results.
- Using ROOT 6.00.01 in the Integration
Builds (requested by Bill).
- The performance problems during the
build are due to memory (swapping).
- Multi-threaded framework has shown a
bug in glibc. If you load a library with TLS section after the
threads have started. CMS have a patch for Scientific Linux, but there
could be a formal problem because the author of the patch is not
anymore available. The solution is limit TLS and deploy their own
loader with the patch.
- Garbage collection for CernVM-FS is
preventing progress in streamlining the operation of CernVM-FS. The
migration was scheduled for summer, but it could be accelerated. Gerri
will send a message with the results.
- RH 7 has been released. What are the
plans for adopting it? Surely we will need to wait for another long
- Next meeting on July 17th.
 HEP-SPEC discussion
A new HEP-SPEC 14 is in preparation. AF and LIM have been asked if the
following compiler switches are still state of the art:
We agreed for dropping "-m32" and keeping the rest unaltered. However, we
are interested in the HEP-SPEC approach to vectorization.
- using gcc as compiler
- -fPIC, -O2, -pthread
In addition, it would be beneficial if the performance suite would allow
replacing gcc by clang to give experiments a chance to do some measurements