Minutes of LCG Generator Services project planning meeting
20-Oct-2006
-----------
People attending: WP (Witek Pokorski)
PM (Pere Mato)
GC (Gabriele Cosmo)
TS (Torbjorn Sjostrand)
MM (Michelangelo Mangano)
MK (Mikhail Kirsanov)
AM (Andreas Morsch)
Patrick Robbe
AR (Alberto Ribon)
People attending remotely: PR (Peter Richardson)
FK (Frank Krauss)
LL (Leif Lonnblad)
BK (Borut Kersevan)
LD (Lev Dudko)
--------------------------------------------------------------------
Organizational matters:
All the partipants agreed on the following:
The purpose of the meeting is to plan and monitor the progress of the
Generator Services project. Representatives of all stakeholders should
be present. Meeting will take place every 6 months. Minutes will be
first distributed internally and then made public to serve as the main
project planning document.
No further suggestions concerning other potential participants of the
meeting.
--------------------------------------------------------------------
Review of the project mandate :
All the participants agreed that the project should continue under the
original mandate but the priorities should be re-discussed.
FK and PR mentioned that in particular MCDB and HepPDT should be low
priority items and more effort should be invested in testing. These
points were discussed later on.
--------------------------------------------------------------------
Roles and expectations:
All participants agreed on the general content of the presented table
(Slide 11).
MM and FK commented the user support item by saying that they prefer
to know all the problems and bugs, to have a complete picture, even if
some of them will be dealt by the Generator Service. Currently the
list of problems is reported in the GENSER Release Notes, but probably
it would be more convenient in the future to collect them in a Web
page, accessible to anyone (in particular the authors). WP pointed
out that a dedicated Savannah portal has been created. GC suggested to
have a kind of "hypernews" forum, distinct from the bug tracker, to
discuss or collect all the problems, but no decision was taken on that
matter. It was agreed, however, that all the problems relevant to the
authors will be forwarded and eventually submitted to their bug
tracking systems (if existing).
MM commented that authors are getting a number of useful feedback
outside CERN, e.g. from Fermilab experiments. It would be useful that
the Generator Service could setup a web page, or any other useful
tool, that allows each MC authors to collect and update all these
extra information which can be useful also for LHC people. WP
suggested that the created Savannah portal can be used for that
purpose.
It was agreed that this issue will certainly be followed.
---------------------------------------------------------------------
Work-packages:
1) GENSER
All the participants agreed on the proposed changes, i.e. to move away
from SCRAM and to use native build methods/systems. The generators
will be built using the approach that is used now for the external
software, in the dedicated directory 'external/MCGenerators'.
The Generator Services team is starting the migration to the new
structure right now in order to be able to provide to the experiments
all the required generators before the end of year 2006.
WP suggested to review the list of used generators and prioritize it
(high/normal priority).
ACTION: experiments were asked to provide prioritized lists of the
generators that they need (by 27 October 2006).
TS suggested that it would be good to have a list of all availble
generators version with labels like 'recommended', 'deprecated', 'not
tested', etc. It was agreed that such a list will be created and
accessible from the web page.
The issue of EvtGen was raised by PR. It was agreed that Generator
Services should take care of converging to a single version.
Comment from BK about the release cycle. Experiments would realy like
to have new versions (especially with bug fixes) of generators
available with a delay of max few days. The 'continous' release cycle,
therefore, was considered as a good idea.
Comment from MK concerning the new GENSER CVS. In the cases like
Pythia 6, where the code comes in one big file, Generator Services
team split it in a few files and build several libraries. It was
agreed that in such cases the 'split' code will be put in the new
GENSER CVS to keep track of the applied procedures.
2) Generators testing and validation
All the participants agreed that validation is vital and that it
should be considered as a high priority issue.
It was agreed that the authors either provide or at least suggest the
tests that should be carried out regularly by the Generator Service.
There was a discussion about possible inclusion of specific tests
performed by the experiments. It was agreed that moving all the
experiments' test to Generator Services validation would create a
maintenance problem and therefore should not be done. On the other
hand, as TS suggested, a number of representative tests (particularly
relevant for specific experiments) could be added to the list of the
tests performed by Generator Services. Patrick said that he would
inform us about the tests that LHCb was using.
The general opinion was that tests expected from the Generator Service
should be mainly regression tests, i.e. comparisons of numbers, or
histograms, between different versions. Physics Validation,
i.e. validation on data, should be left mainly to the experiments and
authors.
ACTION: authors and experiments were asked to revise the list of tests
performed now in GENSER and suggests additional ones that
should be done
It was agreed that the test will be documented on the web page with
possible links to some files containing the output.
3) Event record and particle properties table ATLAS, CMS and LHCb
acknowledged that they were using HepMC as the OO event record and
that they were expecting LCG to continue to support it. It was agreed
that no changes in the project were required concerning HepMC. In the
same time it was pointed out by BK and FM that experiments (ATLAS and
CMS) are finding bugs the generator-to-HepMC 'translators' and that
those should be tested by someone. It was agreed that the Generator
Services team will do that work by extending the tests done in GENSER
to cover also the testing of HepMC 'translators'.
There was a discussion on HepPDT which indicated yet again that a new
solution needed to be found. FK, PR and TS stated that
Herwig/Sherpa/Pythia will not
link to any external packages in order to get particle
data and therefore does not plan to support HepPDT. Moreover, it would be very difficult to converge to a unique
solution as far as the decay modes are concerned. This point of view
was shared by several participants. On the other it was clear that
the experiments needed a mechanism to set the masses and the lifetimes
of particles to be the same in different generators they use as well
as in Geant4 and in other parts of the event processing software. As
TS suggested a possible solution would be to have a list of masses and
lifetimes (not decay modes), in a simple text format, which should be
updated regularly and regarded as the reference for all the simulation
software.
WP suggested that a technical meeting devoted to that subject should
be organized.
4) First level support and communication
(see comments in 'Roles and expectations')
TS pointed out that the definition of Level-1 support was not clear.
WP answered that the role of Level-1 support would be to act as the
contact point for all the requests and reports from the LHC
experiments related to MC generators. Generator Services would then
deal with those issues that can be solved without the involvement of
the authors. As agreed during the earlier discussion, the authors
would be kept informed about all the reports and requests also those
handled by the Generator Services.
WP said that the web page will be further improved. In particular
there will be a list of generators and their version numbers with
labels like 'recommended', 'deprecated', 'not tested' etc (following
the suggestion by TS). There will also be a link to a page with all
the tests and validation performed.
General comment about the proposal of tutorials (done by the authors
addressed to individual experiments): the proposal is good and
interesting for both the authors and the experiments, but the
involvement of Generator Services is most likely not needed.
5) Event database (MCDB)
FM: CMS wants to use MCBD
WP: MCDB seems to be ready to use. Remaining key issues should be
solved by the end of 2006.
BK: MCDB is not high in the list of priorities for ATLAS. ATLAS will
not use MCDB in production. ATLAS would prefer to see more
testing/validation on the generators.
Patrick: No interest in MCDB for LHCb.
FK and PR again expressed opinion that the effort in the first
priority should be put in the validation of generators and not MCDB.
Comment from LD (sent by email after the meeting): numbers on slide 32
concerning the effort invested in MCDB were misleading due to the fact
that they were not taking into account the planning over the entire
year. Because of some administrative problems, the effort invested in
MCDB has been concentrated in the last quarter of the current
year. The overall effort invested in MCDB has been in fact at the same
level as the effort invested in GENSER.
Comment from PM: question is whether Generator Services should invest
effort in any further development of MCDB.
WP: Taking into account the feedback so far the answer is 'no'.
ACTION: resource providers are asked next year to reduce the available
effort dedicated to MCDB and move it to GENSER and to the
validation of generators
---------------------------------------------------------------------
Milestones:
The proposed milestones were accepted by all the participants.
---------------------------------------------------------------------
Manpower:
It was stressed by FK that the planning for the next year should take
into account the low priority assigned to MCDB and high priority
assigned to GENSER and validation.
WP agreed and said that he would move effort from MCDB to the
validation of generators.
---------------------------------------------------------------------
Meetings:
The proposed meeting schedule was accepted by the participants.
---------------------------------------------------------------------
Final remarks:
No further comments from the participants concerning the project planning.
Generator Services team expects:
1) feedback from the experiments regarding the list of generators, and
their versions, that are currently needed, and the distinction between
"high priority" and "normal priority" between the various generators.
2) feedback from the MC authors and experiments about the list of tests.
There are minutes attached to this event.
Show them.
