tschuett added a comment. In D71016#3063706 <https://reviews.llvm.org/D71016#3063706>, @bader wrote:
> In D71016#3063457 <https://reviews.llvm.org/D71016#3063457>, @tschuett wrote: > >> It feels like you are doing codegen(OpenCL kernel) in Sema. Are OpenCL >> kernels the only approach. > > We need to update the description of the patch to clarify that it applies to > other GPU programming models as well. When the patch was uploaded SYCL > targeted OpenCL kernels only and today the downstream implementation can > target CUDA, HIP and Intel oneAPI Level Zero kernels as well. > SYCL kernel is defined as C++ callable type, but typical GPU kernel interface > is a C-like function. In addition to that there might be additional > restrictions on passing data from the host system (e.g. image types can be > passed as a member of C++ class, etc.). The solution here is emit a wrapper > function, which initializes and invokes SYCL callable object. > > Does it answer your question or you would like to see changes in addition to > the description update? Would a codegenSYCL directory help you to separate Sema from code generation? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D71016/new/ https://reviews.llvm.org/D71016 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
