Intel Proposes Adding Full SYCL Programming Model Support To Upstream LLVM

Intel engineers are proposing that full support for the SYCL programming model be added to upstream LLVM. This is part of their broader oneAPI effort and embracing Khronos' SYCL standard for single-source C++ heterogeneous programming from CPUs to GPUs, FPGAs, and other accelerators.

Intel makes extensive use of SYCL as their heterogeneous programming model for targeting their widerange of hardware with oneAPI as well as helping the hardware/software ecosystem see broad SYCL support. Intel makes use of LLVM downstream in their DPC++ compiler stack and numerous other components.

Over the past several years Intel's LLVM-based SYCL compiler has matured well and over time their engineers have made numerous request for comments (RFC) over getting more of the code upstreamed. On Friday an RFC was submitted over adding full support for the SYCL programming model to LLVM. Since their prior upstream efforts their design and SYCL support have evolved, they've received a much broader range of feedback, there is more real-world use, and thanks to Codeplay there is more support for SYCL on non-hardware platforms.

Those interested in learning more can see their RFC message over their hopes of SYCL support in upstream LLVM. No responses have been posted yet but in any case it's still likely to be a long endeavor before all of the SYCL support would be upstream in the LLVM compiler codebase. In any event it's great to see Intel committed to getting this upstream.