Heterogeneous Programming with SYCL - AdaptiveCpp Edition
Heterogeneous Programming with SYCL - AdaptiveCpp Edition
We cordially invite you to the hands-on workshop for “Heterogeneous Programming with SYCL - AdaptiveCpp Edition” organized by Saarland University. This hands-on workshop will take place in-person at Johannes Gutenberg University Mainz on December 11th, 9:00 - 18:00.
Register here: https://forms.office.com/e/gXYAGke1sW
Registration closes December 5th - the number of slots is limited. You will receive a manual confirmation via E-Mail that your registration is confirmed.
These days, most high-performance computing facilities are sporting numerous heterogeneous nodes, i.e. they provide some sort of accelerator, most often GPUs. To make the most out of the available hardware, research codes increasingly explicitly program GPUs. Often, vendor-specific programming models, such as CUDA or HIP, or lower-level APIs, such as OpenCL, are used for this.
In this workshop, we will give a hands-on introduction to SYCL: a vendor-independent, single-source programming model for heterogeneous programming that is embedded into C++. One of the main contributors of AdaptiveCpp (one of the two major SYCL implementations, grown at Heidelberg University) will take you through the basics of using SYCL with AdaptiveCpp in practice. After this workshop you will be able to write code that runs fast on any CPU and GPUs from all major vendors (Nvidia, AMD, Intel).
You will learn in particular:
- What is SYCL?
- Setting up AdaptiveCpp.
- Writing cross-platform compute kernels that run on GPUs but also on CPUs.
- Getting your data where it is needed.
- And more..
We will go through the material together, so bring your laptop. We are planning to provide you with guest access to GPU-nodes for the workshop (but are still finalizing the details). However, when working locally, we require Linux for the workshop (MacOS works with CPUs only). More details (e.g. link to the workshop material) will be provided in the days before the workshop.
Organizer: Joachim Meyer
Participation is free of charge for attendees from German universities, academic computing centers, and research institutions. Please only register for the course if you are really going to attend. Registration closes December 5th.