Nicolas Miller
Nicolas Miller
There's a SPIR-V call for the max sub-group size, so essentially the wavefront size: * https://github.com/intel/llvm/blob/sycl/libclc/amdgcn-amdhsa/libspirv/workitem/get_max_sub_group_size.cl You may also be able to use the `__AMDGCN_WAVEFRONT_SIZE` macro.
It looks like a missing symbol from compiler-rt, I think this is something we likely should support. But in the meantime it may be possible to fix it by using...
> Can you also elaborate more on a possible missing symbol from compiler-rt? Is that something I can correct? Not really, to clarify on this we actually don't currently support...
Hello, thanks for the report! I've been looking into this, there was some issues in the clang driver for the `Windows-Clang.cmake` module to work properly: * https://github.com/intel/llvm/pull/6699 And a similar...
Quick update to say all the commits are now merged, so it should work with the latest
@ThomasRobertFR maybe you could comment on that.
Could you try inlining again? https://github.com/intel/llvm/pull/7046 may have helped with that
@romanovvlad could you have a quick look at the `allocateMem` changes in this patch? The ABI tests seem to suggest they're ABI breaking changes but since your patch in https://github.com/intel/llvm/pull/6600...
I'm picking this patch up, and I wanted to summarize again the changes happening here. So the main goal is to support `sycl::context` with multiple devices for the CUDA plugin....