Intel's oneAPI Is Coming To AMD Radeon GPUs
Intel and the Heidelberg University Computing Center are announcing today they are establishing the "oneAPI Academic Center of Excellence." Great for academia, but what's more interesting to the masses that as part of that Intel and the University of Heidelberg are working to add oneAPI support for AMD Radeon GPUs.
As we have previously covered, Codeplay has already been working on oneAPI/DPC++ support for NVIDIA GPUs but to date there hadn't been any announcements concerning AMD Radeon support for this open-standard API. Codeplay's approach has been working to get SYCL running atop NVIDIA CUDA.
It's quite interesting to see now that thanks to Germany's URZ and Intel that there is now AMD Radeon support being addressed. I did seek further information on their plans but prior to the embargo lift didn't receive any further details. Presumably this oneAPI Radeon support is coming by building off the existing AMDGPU LLVM back-end given the SYCL/DPC++ support being LLVM-based and thus just extending the AMDGPU LLVM back-end. Thanks to AMD's open-source driver stack and full-featured support it should be more straight-forward for adding this oneAPI support than the currently pursued NVIDIA approach of going through CUDA. But once I have any more details on their technical plans or expected timeline will pass them along. In any case, great to see that oneAPI will see Radeon GPU acceleration support.
how come that @phoronix is faster than we are oO. Anyway, yes we are working on adding SYCL 2020 features to hipSYCL to enable increased portability of oneAPI code via hipSYCL.
— Aksel Alpay (@illuhad) September 29, 2020
Update: One of the Heidelberg developers involved has confirmed they are working on adding SYCL 2020 functionality to hipSYCL as their oneAPI approach. The hipSYCL code in current form can be found via GitHub. and previously covered on Phoronix as a SYCL implementation for CPUs and ROCm and CUDA.
13 Comments
