Choose timezone
Your profile timezone:
ROOT Review planning
ROOT Meeting
Present: the ROOT team, Peter, Marco Cl., Andreas, John Harvey, Pere, Bill (R)
Axel: Progress by the ROOT team
o Proposed to not adopt the pcms for ROOT6 Beta but the headers
- The pcms are part of a performance optimisation
- Headers do work, they are just slower
o A change in the rootmap files is proposed when templates are involved: the rootmap
files will contain a forward declaration of all templates for which instances are part of the rootmap file.
o Marco: how the proposed change in the rootmap file will affect the current usage of the infrastructure?
Completely new feature: nothing
o Pere: Don't we need something for the needed header file for the template? No, we don't.
Fons: Tasks for ROOT 6.00.beta (November Release)
o Important classification of the tasks for the 6.00.beta (November) release using jira
filters:
- Open Tasks: https://sft.its.cern.ch/jira/issues/?filter=10871
- Open Bugs: https://sft.its.cern.ch/jira/issues/?filter=10873
o Marco: makes the point about the fact that Reflex was abandoned by the experiments because it
was declared by ROOT that it would not be supported in version 6.
The team recognises the big and timely work by the experiments to migrate away from Reflex.
o Fons: The jira items are updated daily, all interested people can have a look at the filters to get a "snapshot"
of the current status
o Philippe: When https://sft.its.cern.ch/jira/browse/ROOT-5465 is closed we can make a tag to checkpoint the progress
Peter:
o In ALICE HLT working, except error report with line number, which was provided thanks to cint: this was expected and no action is to be taken by the ROOT team.
o Analysis is to be tried: nothing fancy there, the feeling is that no blocker will show up. This is the last missing piece of Alice software which has not been ported to ROOT6
o Material Caluclation: investigation ongoing. Bruteforce: script that updates the code commit by commit, compiles and runs.
We will know what commit is and which is the culprit of the degradation in physics performance (as mentioned in all previous meetings, this is not a feature of ROOT6 but is present also in ROOT5)
o AliRoot compiled against ROOT6 will be easy to build, just a Cmake call. The recipe to set it up will be provided for the
people intersted in testing.
Valeri:
o Mail circulated about progress:
"""
1) I managed (with Atlas Nightlies team support) to configure the Atlas nightlies correctly to test the latest corrections.
2) Atlas nightlies revealed that the package AtlasPoolCnvSvc has the Reflex dependeny and it was not on my "to do" list. It was missed.
3) One needs to find the reflex-less solution for that package yet. That is 1 FTE at least.
4) Right now I spend 10% on ROOT-6 Atlas.
"""
o Valeri intervened also at the meeting in person. The most important point is the Reflex dependency in the AtlasPoolCnvSvc which was not
migrated to TClass (the interested reader could go through the minutes of July the 10th:
https://indico.cern.ch/getFile.py/access?resId=0&materialId=minutes&confId=260890)
Marco:
o Promising situation, but PyROOT is needed for everything: it should be fixed asap (added a posteriori: it is fixed https://sft.its.cern.ch/jira/browse/ROOT-5465)
o Marco asks why we need the compiled dictionaries if headers are there: for IO customisations, to percolate in the TClass meta system the
properties in selection files, to have functions at disposal like the streamer w/o JITting them
o Pere: header files cannot be properly distributed. We cannot run on Grid and we cannot have validation.
- Alice can run also on grid sites: the headers are distributed
- LHCb: nothing was tested, but the headers are distributed as well (no distincion between "dev" and "bin" LHCb distributions)
- CMS accepted this solution
Bill:
o Will crosscheck with the CMS management if it is possible to have an integration build running of CMSSW against ROOT6.
Present: the ROOT team, Peter, Marco Cl., Andreas, John Harvey, Pere, Bill (R)
Axel: Progress by the ROOT team
o Proposed to not adopt the pcms for ROOT6 Beta but the headers
- The pcms are part of a performance optimisation
- Headers do work, they are just slower
o A change in the rootmap files is proposed when templates are involved: the rootmap
files will contain a forward declaration of all templates for which instances are part of the rootmap file.
o Marco: how the proposed change in the rootmap file will affect the current usage of the infrastructure?
Completely new feature: nothing
o Pere: Don't we need something for the needed header file for the template? No, we don't.
Fons: Tasks for ROOT 6.00.beta (November Release)
o Important classification of the tasks for the 6.00.beta (November) release using jira
filters:
- Open Tasks: https://sft.its.cern.ch/jira/issues/?filter=10871
- Open Bugs: https://sft.its.cern.ch/jira/issues/?filter=10873
o Marco: makes the point about the fact that Reflex was abandoned by the experiments because it
was declared by ROOT that it would not be supported in version 6.
The team recognises the big and timely work by the experiments to migrate away from Reflex.
o Fons: The jira items are updated daily, all interested people can have a look at the filters to get a "snapshot"
of the current status
o Philippe: When https://sft.its.cern.ch/jira/browse/ROOT-5465 is closed we can make a tag to checkpoint the progress
Peter:
o In ALICE HLT working, except error report with line number, which was provided thanks to cint: this was expected and no action is to be taken by the ROOT team.
o Analysis is to be tried: nothing fancy there, the feeling is that no blocker will show up. This is the last missing piece of Alice software which has not been ported to ROOT6
o Material Caluclation: investigation ongoing. Bruteforce: script that updates the code commit by commit, compiles and runs.
We will know what commit is and which is the culprit of the degradation in physics performance (as mentioned in all previous meetings, this is not a feature of ROOT6 but is present also in ROOT5)
o AliRoot compiled against ROOT6 will be easy to build, just a Cmake call. The recipe to set it up will be provided for the
people intersted in testing.
Valeri:
o Mail circulated about progress:
"""
1) I managed (with Atlas Nightlies team support) to configure the Atlas nightlies correctly to test the latest corrections.
2) Atlas nightlies revealed that the package AtlasPoolCnvSvc has the Reflex dependeny and it was not on my "to do" list. It was missed.
3) One needs to find the reflex-less solution for that package yet. That is 1 FTE at least.
4) Right now I spend 10% on ROOT-6 Atlas.
"""
o Valeri intervened also at the meeting in person. The most important point is the Reflex dependency in the AtlasPoolCnvSvc which was not
migrated to TClass (the interested reader could go through the minutes of July the 10th:
https://indico.cern.ch/getFile.py/access?resId=0&materialId=minutes&confId=260890)
Marco:
o Promising situation, but PyROOT is needed for everything: it should be fixed asap (added a posteriori: it is fixed https://sft.its.cern.ch/jira/browse/ROOT-5465)
o Marco asks why we need the compiled dictionaries if headers are there: for IO customisations, to percolate in the TClass meta system the
properties in selection files, to have functions at disposal like the streamer w/o JITting them
o Pere: header files cannot be properly distributed. We cannot run on Grid and we cannot have validation.
- Alice can run also on grid sites: the headers are distributed
- LHCb: nothing was tested, but the headers are distributed as well (no distincion between "dev" and "bin" LHCb distributions)
- CMS accepted this solution
Bill:
o Will crosscheck with the CMS management if it is possible to have an integration build running of CMSSW against ROOT6.
There are minutes attached to this event.
Show them.