Bright and Dark Sides of Heterogeneous Computing in High Energy Physics

by Dr David Rohr (CERN)

Friday 31 Jul 2026, 09:00 → 12:30 Europe/Zurich

593/R-010 - Salle 11 (CERN)

Abstract:

The course will give an overview of heterogeneous computing for event reconstruction in high energy physics.
The topics will range from multi-threading v.s. multi-processing, over pipelines to hide latencies both on host and on GPU computing to dynamic workload distributions in heterogeneous environments.
In particular, it will show that different problems will require different computing solutions and that there is no magic bullet.
And it will point out the differences between classical high performance computing and computing in high energy physics.
The course will be complemented with some simple exercises.
The purpose of the exercise will not be lots of C++ developments, but rather to modify a provided reference application and perform some benchmarks.

Lecturer Bio:

David Rohr was born in Mannheim Germany, and is a physicist and developer of fast algorithms for event reconstruction at CERN. He works for the ALICE experiment and currently has the positions as lead developer for the TPC detector, leader of the working group for software for hardware accelerators, and computing coordinator of the ALICE 3 upgrade of the experiment.
David studied physics in Heidelberg and finished his PhD at the University of Frankfurt in 2015. He has always been working interdisciplinary in scientific computing. Since 2020, David holds a staff position at CERN.
Before coming to CERN, David was working for designing, benchmarking, and commissioning several computing centers in Germany and also the High Level Trigger farm of the ALICE experiment at CERN.
One of his focus is the low-level optimization of algorithms and applications, for both CPUs and GPUs, ranging from assembler code, explicit vectorization, and GPU computing to modern C++ code.