Description:

The SYCL & oneAPI Hackathon at CINECA, organized in collaboration with Intel, is a three-day, in-person event designed to help developer teams to port and/or optimize their codes for GPUs using the SYCL programming model and the oneAPI ecosystem.

During the event developers will have the chance to work alongside mentors and SYCL programming experts, which will guide them in porting and optimizing their code for GPUs.

Participants will also have the opportunity to run their code on Leonardo, one of the most powerful computers in the world (https://leonardo-supercomputer.cineca.eu/) and on the Intel Developer Cloud, to show performance portability among different high-performance GPUs.

Over the three days of the hackathon, participants will attend technical presentations focusing on SYCL & oneAPI development, optimization techniques and hints, as well as performance portability strategies, which will be validated on a project either proposed by the participants (BYOC: bring your own code) or provided by the instructors.

Master and PhD students are encouraged to apply to the hackathon and bring their own code. Participants can also bring code under non-permissive and copyleft licenses (e.g. proprietary licenses), the code will not be distributed to anyone other than the owner.

The project may focus on

• porting an existing codebase from CUDA to SYCL using a variety of tools, technologies, and techniques
• accelerate an existing sequential code written in Fortran /C by moving to C++ and SYCL, and running the final code on multiple architectures

Important Information:

The event is FREE of charge, but in order to guarantee the quality of the teaching and support offered during the hackathon, only a limited number of places are available.

The event will be held in ENGLISH, and only IN PERSON. The location for the event is CINECA’s site in Casalecchio di Reno, close to Bologna.

Full address for the venue: Via Magnanelli, 6/3, 40033, Casalecchio di Reno (BO).

Certificates attesting the participation in the event can be released for those that need it.

Target Audience:

Students and researchers interested in developing and optimizing parallel programs for GPUs using modern C++ and SYCL/oneAPI.

Some prior programming experience with C/C++ or Fortran and familiarity with Unix shell are required. Basic knowledge of GPU programming is not mandatory but would be helpful.