ROOT Parallelisation, Performance and Programming Model Meeting

Thursday 28 Sept 2017, 16:00 → 18:25 Europe/Zurich

40/R-A10 (CERN)

Danilo Piparo (CERN)

CERN number: 71400 Extension: 109284483#

Present: Danilo, Xavi, Guilherme, Enrico, Gerri, Zhe, Oksana, Pere

Enrico:
ROOT-9018
New feature: the callbacks.
- Run something every 10 entries
- Run something every 10 entries on a slot (needs to be thread safe)
- Run something once, before starting the event loop, useful for example to register in a browser

Gerri and automatic bins:
- Prototyped on this
- A situation where histos are mergeable and with ounder/over-flows
- O/U-flows: you take as the min of the final histo the max of the mins and as max the min of the max

Round table:
- Option 3 or 4 for the activation of MT
- Activate by default is by too risky with the LOC not thread-safe
- Problems are not logical nor technical:
  o THashList: linear search - we need to improve there
  o We need THashList with a RWLock in the right places
Work around: put a lock around LOC - this will not be sufficient even if it helps
If we want to work this around we need to protect the dtor of TObject

We'll wait for the bulkIO to be available and working before diving into uproot. The performance is going to be probably already very good.

16:00 → 16:45 Sprint Status

We'll review the items of the sprints in JIRA (https://sft.its.cern.ch/jira/browse/ROOT-8855 and linked items.) to summarise the status as well as the progress in the different corners and redistribute work if necessary.

16:45 → 17:05 Round Table

Topics to be tackled:
- List of cleanups in 6.12
- Report of planning meeting about strategy for activating IMT by default for 6.12
- uproot in C++ to power-up the TRootDS