ROOT Team Meeting

Europe/Zurich
32/1-A24 (CERN)

32/1-A24

CERN

40
Show room on map
Axel Naumann (CERN)
Description
CERN number: 71400 Extension: 109284483#

Attendees: Brian, Guilherme, Jakob, Philippe, Danilo, Oksana.

Jakob: resurrecting ACAT benchmark to address TTreeReader overhead.

Guilherme: Will address Dan’s request to augment interface of TFileMerger after the release.

Danilo: Bugs fixes in various parts of the I/O.   Bug with IMT and adding branches will be put in the v6.16 branch (+2 other in cmake and interpreter).  Working with Stefan on reading old Roofit workspace, it seems to be failing since v6.12.  

Philippe: This is now on my plate.

Danilo: Working with Javier to reduce the overhead reading from TTreeCache.

Philippe; The challenge is to set the training phase correctly and/or do training by hand.

Danilo: What about the factor 2 between SetBranchAddress and TTreeReader; this is becoming urgent-ish as we need to write a paper including benchmark.

Philippe: This is now on my plate.

Danilo: Brian did you have time to run the CMS open data of ROOT with Bulk DataSource?

Brian: No because it requires addition to the code in the DataSource (I am working on it now).

Axel: Jenkins is not yet ready for running properly PRs.

Brian: Where can I find the meeting note for the PPP meeting?

Danilo: On the indigo agenda page.

Philippe: I fixed schema evolution code for the case of unique_ptr, the old code was marking all unique_prt has being never nullptr (aka comment //->) and resolve several issue in TTreeCache when using SetEntryRange.

Oksana: I am writing ACAT abstract and helping Brian with Bulk I/O.  Plan is to enable for 6.18 LZ4 as the default, aiming to do so as early as possible to gather more feedback.

Axel: Enum constant for compression, why not a class enum?  

Oksana: Yes, this is the end plan for v6.18

Axel: No, if we release v6.16 as is, we will have to keep the enum ‘forever’.

Brian: Eventually I would like to have a CompressionSetting class

Axel: I like the idea but would also like to introduce new global scope

Philippe: Issue might be to have to introduce overloads ….

Conclusion of the discussion:

temporary solution: put the enum (non-class so they decay to int) into either a class or namespace to have better name and still have the old function as is.

Longer solution: those enum names are re-used for constexpr CompressionSettings object.

Oksana: I will work on this in the coming days.

 

There are minutes attached to this event. Show them.
    • 16:00 16:05
      News 5m
    • 16:05 16:15
      Shift handover 10m
    • 16:15 16:25
      Meeting Summaries and Plans 10m
      • I/O
      • PPP
      • TMVA
      • Planning / Godparents /...
    • 16:25 16:40
      Topics 15m
      • Delayed build tool / package manager (2/2)
      • Release status
      • CI infra changes
    • 16:40 16:55
      Round table 15m
    • 16:55 17:00
      A.O.B. 5m