High-Level GPU Programming

This training is a collaboration between CSC and Vienna Scientific Cluster / EuroCC Austria.

Learning outcomes:

After the course, participant will be able to write hardware-agnostic code to express parallelism using SYCL and Kokkos that can runon CPU, Nvidia GPU, or AMD GPU. Manage memory across devices, do basic performance analysis, and evaluate the drawbacks between different approches for programming GPUs.

Agenda:

Day 1, Wednesday 14.02, 9:00-16:00

Introduction to GPUs and GPU parallle programming model
C++ concepts
Introduction to SYCL, queues, command, command groups, kernels
Memory management in SYCL

Day 2, Thursday 15.02, 9:00-16:00

Expressing parallelism
Dependencies, synchronization

Day 3 Friday 16.02, 9:00-16:00

SYCL profiling
Interoperability with other libraries
Multi-GPU programming
Kokkos
Lecturers: Cristian-Vasile Achim (CSC), Jaro Hokkanen (CSC), Tuomas Rossi (CSC), Intel oneAPI certified instructor Soner Steiner from Vienna Scientific Cluster, NCC Austria

Language: English

Price

The fee covers the lunches, coffee and snacks during the event.

• For academic partipants: 180€ ex VAT.
• For industry: 840€ ex VAT.

This event is ON-PREMISE and the participation is limited to 28. The registration is at a first come, first served basis and it closes on February 7 at 23:59!

Registration and course details.