54th ROOT Parallelism, Performance and Programming Model Meeting

Thursday 14 Feb 2019, 16:00 → 17:30 Europe/Zurich

4/S-030 (CERN)

Danilo Piparo (CERN)

tentative scaling

Present: Guilherme, Stephan, Stefan, Enrico, Enric, Javi, Danilo, Philippe

Actions

- Understand the different behaviour of VTUNE in 6.16 and master

- Understand the time needed to run the python part histo writing part. 15s is far too much to flush the histos on disk

- Understand the time needed to start: 8s seems a lot

- Communicate to Elisabetta that:

  - 2 loops are taking place

  - Start in batch mode to avoid the additional graphics thread

Performance Measurement of WMass Analysis

- VTune unstable: will move to Perf

- It would be nice to have jitting debugging symbols

- We should think if we can help users too automatically bookkeep the histograms

- Ideas for the debug report of RDF: how many times the evt loop runs, how many columns were read

16:00 → 16:20 Preliminary Performance Analysis of WMass CMS Analysis

Speaker: Guilherme Amadio (CERN)

CMS W-mass Performance Analysis.pdf