Minutes (written by A. Ribon) of the LCG Generator Services monthly
meeting on 8 July 2009 dedicated to a first discussion on HepMC 2.06 .
---------------- -----------------
Attendees:
in room 32-1-A24:
- Andy Buckley (AB)
- Gloria Corti (GC)
- James Monk (JM)
- Alberto Ribon (AR)
- Lars Sonnenschein (LS)
remotely via EVO:
- Frank Gaede (FG)
- Lynn Garren (LG)
- David Grellscheid (DG)
- Leif Lonnblad (LL)
- Fabian Stoeckli (FS)
- Torbjorn Sjostrand (TS)
Excused:
- Judith Katzy (JK)
=====================================================================
LG presented first a brief overview of the current HepMC releases
and the recent changes that have been introduced, and then the
requests for HepMC 2.06 collected so far.
The aim of the meeting was to start the discussion on HepMC 2.06,
with the final discussion and decisions in January 2010.
Nobody requested to move up the release of HepMC 2.06 to this year.
Additional comments to the slides:
---------------------------------------------
o slide #5 (subjet: methods returning a const reference instead of a copy)
------------
Pere Mato had suggested to modify some methods so that they
return a const reference instead of a copy. This has been
implemented in HepMC 2.04.
LG said that this change is not completely backwards compatible,
so user code that was manipulating the copy returned by the
affected methods needs to be changed. Given that the experiments
are still using HepMC 2.03.09, there is no yet feedback on this
potential issue.
GC suggested that it was likely that the fixes should be in the
client code rather than in HepMC, as the intended semantics are
that the returned objects are not to be modified.
o slide #8 (subject: "proposal 1")
------------
No objections, so it will be proposed for final inclusion in
HepMC 2.06.
o slide #9 (subject: "proposal 2")
------------
No objections and approval by TS (via email) on a single Pythia
wrapper. So a single Pythia wrapper, based on Pythia 6.4, and
a single Herwig wrapper, based on Herwig 6.5, will be proposed
for final inclusion in HepMC 2.06.
o slides #10 & #11 (subject: "Proposal 3a & 3b")
-----------------------
AB explained the rationale of the request of being able to insert
random lines in the HepMC event header to be ignored: this would
allow that future extensions in the information content of the
HepMC header would not cause problems when read in by previous
versions of HepMC that are unable to interpret the meaning of
such extra new information.
LG commented that random lines should be accepted only in the
event header, not in the vertex and particle sections.
The proposals 3a & 3b will be kept for the final discussion
on HepMC 2.06 .
o slides #12-15 (subject: "Proposal 4")
--------------------
AB explained that the only reason for the request to introduce
container-like accessors (which can then be accessed via STL
iterators, or other means like Boost.Range's foreach) in HepMC
is convenience.
The implementation and its performance should be tried out.
For backward compatibility, the existing iterators of HepMC
should not be removed or changed.
Action: some prototyping and tests are needed in order to
******* arrive to a concrete proposal for the final discussion
on HepMC 2.06.
o slide #16 (subject: "Proposal 5")
--------------
Given that the size of the HepMC tar-ball is dominated by the
event dumps for the tests, AB suggested that the users could
download them separately.
LG prefers to keep the event dumps for the tests in the HepMC
distribution tar-ball, but she will investigate how to reduce
their sizes.
DG commented that this item should have lower priority with
respect to the other proposals. Everybody agreed.
o slide #17 (subject: "Proposal 6")
--------------
AB suggested to change the installation location of HepMC examples
to follow the common convention used by most of the packages
built with autotools.
An extra reason to use the standard $prefix/share location is
that the configure script --datadir flag will be respected.
DG agreed that it was better to use the standard locations.
This also affects the documentation, wich should be installed
into the location specified by the --docdir flag.
LG expects that most users do not care about the location
of the examples, as far as they are present somewhere.
No clear what to do. Perhaps we should wait for more feedback
on the HepMC 2.6 meeting in January.
o slide #18 (subject: "Proposal 7")
--------------
JM explained the rationale of the request: weights in HepMC
have been used recently for dealing with different PDFs, so
introducing named weights would be convenient and would make
their use more transparent.
DG asked whether an agreement on the string format and naming
is needed between the MC authors.
JM replied that the strings used for naming the weights should
be chosen freely, without need of agreed convention.
GC commented that we should keep in mind also the persistency
of HepMC events.
Action: some prototyping and tests are needed in order to
******* arrive to a concrete proposal for the final discussion
on HepMC 2.06.
o slide #19 (subject: "Final thoughts")
--------------
LG is worried about the growing gap between the HepMC version
used by the experiments (2.03.09), and the newest versions
used by the generator community.
AB commented that the current structure of GENSER makes it
impossible to use more than one version of HepMC, and an
extra directory level to "version" the GENSER area according
to the versions of common packages like HepMC would make testing
and version-switching easier.
The bootstrap script should make this feasible.
GC commented that the HepMC version used by the experiments
is agreed at the LCG Architect Forum: so this is the place
where to discuss the migration to a newer version.
GC suggests also that the LCG nightlies could be used for some
GENSER generators in order to build and test them against
the latest version of HepMC (the generator installed in the
GENSER AFS area have been built with HepMC 2.03.09).
Action : AR should investigate with Stefan Roiser about the
******* possibility of testing some GENSER generators in
the LCG nightlies.
Further discussion after Lynn's presentation:
-----------------------------------------------------------
o New status codes for fast simulation
-------------------------------------------------
JM has reported that some discussions have started at this
year's Les Houches workshop regarding the need of new status codes
in HepMC for fast simulation. Some generator authors, some
ATLAS members, and some authors of external fast simulation
packages have been so far involved in these discussions.
AR commented that ATLAS, CMS and LHCb collaborations should be
also involved in these discussions, and a concrete proposal
should be drafted before the end of the year in order to have
a chance to be considered for inclusion in HepMC 2.06.
o Monitor progress toward HepMC 2.06
----------------------------------------------------
AR should monitor that the prototypes and tests that are
needed for some items (see "Action" above) are actually
performed in time for the final meeting on HepMC 2.06
(expected around the end of January 2010).
The first check should be made by the end of September,
contacting by email the relevant people (LG, AB, JM).
DG suggested that at each LCG Generator Service monthly meeting
the status of the work toward HepMC 2.06 should be briefly
updated.
=====================================================================
There are minutes attached to this event.
Show them.