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), James Brodman (Intel Corporation), Michael Wong (Codeplay Software Ltd, UK; Khronos Group Inc), Yasaman Ghadar (Argonne National Laboratory), Rod Burns (Codeplay Software Ltd, UK)
Abstract: The SYCL programming model provides an open standard way to program heterogeneous systems in modern C++. Since the major SYCL2020 release, which added abstractions and features for HPC, SYCL has seen increased use in application domains needing large exascale-class machines, including fusion energy, molecular dynamics, and aerospace.
In this Birds of a Feather session, we will bring together the community of everyone using and developing SYCL applications and implementations. We will discuss future directions and seek feedback on priorities for SYCLNext. A panel of SYCL experts, runtime/compiler implementers, and application specialists will lead an audience discussion and Q&A.
Long Description: SYCL 2020 has emerged as a standard programming model enabling modern C++17 workloads to be dispatched to heterogeneous devices. SYCL is a Khronos Standard language maintained by multiple members, with representatives from industry, academia, and hardware vendors, with the mission of leading the delivery of modern C++ code for heterogeneous compute, aligning, informing and meshing with the ISO C++ standard language. 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, whilst 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.
At SC19, through Intel’s oneAPI and DPC++, SYCL was adopted to be the programming model for AURORA, the first CORAL exascale computer using Intel processors and accelerators. Subsequently SYCL’s use in HPC has expanded to Frontier, Perlmutter, Summit and LUMI in Europe. 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.
Building on the success of Birds of a Feather sessions at Supercomputing since, 2019 (and forming part of the C++ BOF on prior occasions), we have planned a community-led BOF at SC’23. Following a short, 5-minute, presentation outlining plans for a future direction for SYCL and the latest ecosystem update by current SYCL Working Group Chair, Michael Wong, a panel of invited SYCL experts will help lead an audience discussion and field questions taken live from the audience.
We plan for the discussion to focus on audience questions, however we expect to discuss the outlined strategy for SYCLNext, alignment with new ISO C++ features for heterogeneous compute, implementation and application successes and failures. This BOF will provide a strong voice from the HPC community to help feedback to the development of SYCL going forward.
The panel will be a mixture of implementers of the standard along with application developers. We are actively seeking to host a balanced group of panelists that represents all diversity. We are acutely aware of the lack of diversity within this community, and are continually working to discover those working on SYCL in their role; during the BoF we will ask for contributions to the SYCL.tech project page to help promote work from underrepresented demographics.
Our panel will include the following members of the SYCL community:
Michael Wong, Codeplay (Chair SYCL Working Group); James Brodman, Intel (oneAPI implementor); Yasaman Ghadar, ANL (expert on teaching and using SYCL at Exascale); a representative from the domain of SYCL in Machine Learning; and a number of representatives from scientific names to be confirmed. The panel will be chaired by Tom Deakin, University of Bristol.
Website: https://sycl.tech