The International Conference for High Performance Computing, Networking, Storage, and Analysis
The International Conference for High Performance Computing, Networking, Storage, and Analysis
Authors: Tom Deakin (University of Bristol), Nevin Liber (Argonne National Laboratory (ANL)), Biagio Cosenza (University of Salerno), Aurora Perego (CERN)
Abstract: For the past decade, the open-standard SYCL programming model has provided a portable way to program heterogeneous systems across application domains such as fusion energy, molecular dynamics, aerospace, and AI, running on some of the largest GPU-accelerated machines, such as Aurora.
In this BoF, we will bring together the community of everyone using and developing SYCL applications and implementations. We will showcase new features due for release at SC25, and discuss feedback and priorities for SYCL Next. A panel of SYCL experts, runtime/compiler implementers, and application specialists will lead an audience discussion and Q&A.
Long Description: Since its release ten years ago, SYCL has emerged a standard programming model enabling modern workloads to be dispatched to heterogeneous and accelerated devices. SYCL is a Khronos Standard maintained by multiple members, with representatives from industry, academia, and hardware vendors, with the mission of leading the delivery of modern code for heterogeneous compute, moving ahead of slower language standards. The abstractions in SYCL 2020 provide a spectrum of control crucial for achieving high performance on the latest Exascale systems. SYCL provides a way to program heterogeneous systems in a portable, productive and performant way, while providing additional lower-level explicit controls and interoperability for device and platform specific optimisations. This unique blend of convenience with a route to control stands SYCL aside from other programming models, abstraction layers, and C++ itself.
Since the major SYCL 2020 release, which added abstractions and features for HPC, SYCL has seen increased use in application domains needing large Exascale-class machines. SYCL is performing real science and machine learning/artificial intelligence workloads on large-scale systems, including Aurora, Frontier, LUMI, Perlmutter, Polaris and Dawn. We are seeing it being used across many scientific domains, including molecular dynamics, aerospace, fusion energy, machine learning, automotive, as well as creating innovative abstractions for distributed memory programming. These applications have driven the design of new features, which will be presented for feedback in this BoF.
Building on the success of Birds of a Feather sessions at Supercomputing since 2019, we plan a community-led BOF at SC25. The BOF will set the topic of conversation through a series of short presentations outlining the latest standard extensions added to SYCL over the past year, highlights from the SYCL ecosystem of applications and libraries, and open the discussion with the community to shape the future priorities of SYCL. This will be led by the SYCL Working Group Chair, Tom Deakin, along with other members of the Working Group representing users, application and library writers, and SYCL implementors. This expert panel will lead an audience-led discussion and field questions with attendees on the presentation topics, along with their own topics of interest. This BOF will provide a strong voice from the HPC community to help bring feedback around both the SYCL ecosystem and the standard itself.
Website: https://www.khronos.org/sycl/