LHCb Core Software Meeting
Date and Location
03 Dec 2008
10:30 - 12:30
CERN (2-R-030)
Attendees
Gloria,
Hubert,
Joel,
Marco Ca.,
Marco Cl. (minutes),
Markus,
Olivier,
Patrick (EVO),
Philippe
Subjects
News
- Joel: Meeting with Ricardo to discuss the installation of Dirac on Friday.
Software Releases
Gaudi (Marco Cl.)
- Planning a bug-fix release by the end of the week.
This release should be build from the SVN repository.
- Status of the nightly builds:
- dev1
- fake problem due to a change in the standard output of the tests
- dev
- strange problem with std::string (submittes ROOT bug #44917
)
Marco Ca.: Can we build also slc5-gcc34 binaries for next version of Gaudi?
It is a problem because LCGCMT_55b is not ready for it. Marco Cl. will check with Stefan.
LHCb, Boole, Brunel (Marco Ca.)
- Relased last week.
- New changes are coming.
- PartProp package by Vanya (depending only on Gaudi)
- Packing algorithms by Olivier are in the release, but not used yet.
Do we use them in FEST09?
Gauss (Gloria)
- New release of Geant4
- the absence of version directories required a minor change in the script to get the sources
- New DecFiles
- Gauss HEAD build with the latest LHCb, but have problems with the configuration in Python (because of the recent changes).
Proposal of a convention on the tags for SIMCOND
Starting from the attached
slide
, we had a long discussion.
The convention was defined during the
last Gauss meeting
starting from the idea that it has to be human-readable and tell what are the conditions in it points to.
The parameters that Gauss is going to retrieve from the CondDB are a bit on the border between configurations and conditions. In particular, Brunel needs to be configured with different algorithms/tools if the magnet is off or on, and also the Velo position may require different configurations. To address this issue, Marco Cl. was thinking to have a Python function/class to decode the tag (using some table in a file for the parts that cannot be extracted from the tag string) . The idea is not welcomed and it is thought that the parameter should be retrieved from the CondDB at configuration time, but it is not possible because the parameters to configure the access to the database are know at runtime and not a configuration time.
The only workable solution seems to define a mapping between the simulation configuration key (equivalent to the SIMCOND tag, but not necessary) and the configuration parameter, that Gauss will find in the DB and will be used to steer Brunel. The mapping has to be released in SQLDDDB (or equivalent) and generated "automatically" reading the content of the CondDB for the given tag.
During the discussion was also raised the idea that we could use the beam conditions/settings to steer the simulation of spillover. An algorithm (in Python) can use number of bunches and luminosity (implied by the simulation configuration key) to define the number of minimum bias events to generate and to configure Boole accordingly.
Round Table
- Marco Cl.
- Prepared a version of getpack that works with both CVS and SVN. It is in LbScript, not yet released.
- Marco Ca.
-
- We want to add the possibility to LHCb nightly builds to wait that an LGC slot is completed, so we can use the build of the same day.
- The Win32 and SLC5 build machines are in ready. Hubert will check with Stefan how to build the nightlies on Win32. The SLC5 machine is lxbuild135.
- Hubert
- We should try the new version of CMT in the nightlies to validate it.
- Olivier
-
- Investigated the slowness of initialization of applications and identified a candidate for optimizations in the XML parsing. On one side we can simplify the detector description (for detector experts), on another side we can speed up the parsing by caching the grammar (DTD).
Marco Cl. will prepare an XML image of the CondDB, so that Olivier can investigate the DTD caching.
- Working with Markus to investigate the possibility of having DST in MDF format (for analysis in the Online farm).
--
MarcoClemencic - 03 Dec 2008