schung-amd

Closing as I can't reproduce this with Ubuntu 22.04 + ROCm 6.2.2 on a 7900XTX, configuring `spack` with `+rocm amdgpu_target=gfx1100` and specifying ROCm-packaged clang as the compiler. I did, however,...

Hi @unclemusclez, when is this issue occurring? I can see the place in the source code where this error is emitted, and it looks like it should be picking up...

This should be addressed in ROCm 6.2 with lazy loading (https://github.com/ROCm/hipBLASLt/commit/28eb8258d967f3ccaab5aed891bf40d62cdd099d), so hopefully once WSL for 6.2 is released this is fixed. > ROCm 6.1.2 is nice, but really we...

I've seen other requests for cooperative groups support on Windows and am reaching out internally to push for support if feasible. That being said, I am unaware of the reason...

Hi @Mhyho, thanks for pointing this out. The documentation is incorrect here, WSL does not use `/dev/kfd`. Try `--device=/dev/dxg` instead; an example docker run on WSL can be found in...

Hi @Mhyho, did this work for you?

Thanks for confirming! I'll see if the docs need to be revised on this.

Hi @ZJLi2013, can you provide a reproducer for this? I'm not seeing this in a quick test, `setuptools` seems to process `#include `, `#include `, and `#include ` fine when...

I was able to reproduce the mangling with your code and found the root cause. We do a hipify pass in CUDAExtension [here](https://github.com/ROCm/pytorch/blob/267f82b860ccdf32df3dcb92e2435b64ba0f117a/torch/utils/cpp_extension.py#L1131 ) which maps CUDA header names to...

Closing for now as it's gotten a bit stale; had some internal discussions regarding how we're hipifying these includes but didn't gain any traction on changes. Let me know if...