Julia-for-HEP Discussion

Wednesday 9 Nov 2022, 16:00 → 18:00 Europe/Zurich

Philippe Gras (Université Paris-Saclay (FR))

Julia4HEP-event_1216692-chat.txt

16:00 → 16:10 Introduction

Speaker: Philippe Gras (Université Paris-Saclay (FR))

Slides

16:10 → 16:30 Experience from porting VecGeom to Julia

Pere evaluated Julia on a common HEP use case of detector geometry description. VecGeom C++ package has been partially reimplemented in Julia and performance were compared with the C++ version. Project Github repository: Geom4hep

Speaker: Pere Mato Vila (CERN)

Geom4hep-202211.pdf

16:35 → 17:05 Evaluation of the Julia language based on a fastjet algorithm

The Julia language has been evaluated for highly optimized fast jet algorithms from fastjet. The exercise has been lead in parallel by two persons. Github repositories: JetReconstruction.jl and Antikt.jl. See also Atell-Yehor Krasnopolski's presentaion at IRIS-hep

Speaker: Philippe Gras (Université Paris-Saclay (FR))

Slides

17:05 → 17:20 Experience from UnROOT development on type stability and performance optimization for HEP data

UnROOT is a package written fully in Julia to read ROOT files. It is very fast. The secret used to handle in inhomogeneous columnar data with high performance will be explained. Challenges in optimizing the code will be discussed.

Speaker: Jerry 🦑 Ling (Harvard University (US))

17:20 → 17:40 Code review (TBC)

The programming code from Geom4hep and Fastjet exercises has been reviewed. Hints for improving their performance will be suggested.

17:40 → 18:00 Discussion: conclusions to draw for the report and possible home work

Discussion about the conclusions to draw for being included in the report and possibly some next steps to understand differences observed between Julia and C++.

Draft of the Julia-for-HEP report