Second course on porting code and algorithms to GPUs

30 Jun 2025, 14:00 → 4 Jul 2025, 12:00 Europe/Rome

Dipartimento di Fisica e Astronomia

Alexis Pompili (Istituto Nazionale di Fisica Nucleare), Carlo Battilana (Istituto Nazionale di Fisica Nucleare), Simone Gennai (MIB), Tommaso Diotalevi (Università e INFN, Bologna)

---

 

This training focuses on parallel programming on GPUs. No fee is foreseen.

This event builds upon the first course on porting algorithms to heterogeneous architectures, which was organised in 2023. It expands its scope and covers selected topics in greater detail than the previous training.

The course is primarily intended for personnel hired through the ICSC project to work on projects and research activities within the framework of Spoke 2 (“Fundamental Research and Space Economy”); however, it is also open to anyone interested within the fundamental physics community, including, in particular, Ph.D. students and young post-docs.

Main lecturers: Andrea Bocci (CERN), Felice Pantaleo (CERN)

Facilitators for the hands-on sessions:  Simone Balducci and Aurora Perego.

Technical support: Daniele Spiga.

Additional activities are foreseen during the week. A seminar with hands-on, about the optimisation of the XGBoost Machine Learning algorithm with Optuna, by U.Sozbilir of the University of Bari, is planned on thursday morning.

Details on how to connect to the computing resources used for the training will be sent by email to the registered participants.

The meeting will be held in Bologna at the historic building housing the Physics and Astronomy Department, located at via Irnerio 46, 40126 Bologna (BO), Italy.

---

Monday 30 June

14:00 → 14:30 Welcome (info & logistics)

Speakers: Alexis Pompili (Istituto Nazionale di Fisica Nucleare), Carlo Battilana (Istituto Nazionale di Fisica Nucleare), Simone Gennai (MIB)

Introduction.pdf

Useful information and FAQs

14:30 → 16:15 Lecture GPU (part 1)

Speaker: Felice Pantaleo (CERN)

slides and hands-on instructions

16:15 → 16:35 Coffe Break

16:35 → 17:50 Lecture GPU (part 2)

Speaker: Felice Pantaleo (CERN)

18:00 → 19:00 Welcome drink

Tuesday 1 July

09:00 → 10:45 Hands-on GPU

Speaker: Felice Pantaleo (CERN)

10:45 → 11:15 Coffee Break

11:15 → 13:00 Lecture GPU (part 3)

Speaker: Felice Pantaleo (CERN)

13:00 → 14:30 Lunch break

14:30 → 16:15 Introduction to performance portability with Alpaka (part 1)

Speaker: Dr Andrea Bocci (CERN)

Performance portability with alpaka.pdf

16:15 → 16:45 Coffee break

16:45 → 18:30 Introduction to performance portability with Alpaka (part 2)

Speaker: Dr Andrea Bocci (CERN)

Wednesday 2 July

09:00 → 10:45 Hands-on Alpaka

Speaker: Dr Andrea Bocci (CERN)

Hands-on.pdf

10:45 → 11:15 Coffee Break

11:15 → 13:00 Hands-on Alpaka

Speaker: Dr Andrea Bocci (CERN)

13:00 → 14:30 Lunch break

14:30 → 16:15 Consolidation (more exercises or specific use cases)

Speaker: Dr Andrea Bocci (CERN)

16:15 → 16:45 Coffee break

16:45 → 18:30 Consolidation (more exercises or specific use cases)

Speaker: Felice Pantaleo (CERN)

20:30 → 22:30 Social dinner

(On each own funds but all together in a pre-booked good restaurant)

social.pdf

Thursday 3 July

09:00 → 10:00 Lecture: Hyperparameter optimization with OPTUNA

Speaker: Dr Ümit Sözbilir

Optuna Optimization

10:00 → 10:15 Mini break

10:15 → 11:00 Lecture: Hyperparameter optimization with OPTUNA

Speaker: Dr Ümit Sözbilir

11:00 → 11:30 Coffee break

11:30 → 13:00 Hands-on: optimization in OPTUNA

Speaker: Dr Ümit Sözbilir

Hands-on Slide

Optuna Exercise

XGBoost Example

13:00 → 14:30 Lunch break

14:30 → 16:30 Consolidation (more exercises or specific use cases)

Speaker: Dr Andrea Bocci (CERN)

16:30 → 17:00 Coffee break

17:00 → 18:00 Consolidation (more exercises or specific use cases)

Friday 4 July

09:00 → 11:00 More consolidation (on specific use cases)

11:00 → 11:20 Break

11:20 → 12:00 Wrap-up and closeout