[wip] sparse marlin rocm compilation

[jcaip]

cc @petrex I tried to get this working to run some benchmarks and was running into some issues.

Does this require a specific ROCm version? I'm using 6.3.3 and I'm running into some stuff like

error: use of undeclared identifier '__builtin_amdgcn_global_load_lds'

Aside from that, I ran into a lot of issues with namespace issues with __half2, which I think I fixed by using ::__half2, and one other bug with the hip version of cudaSetAttribute needing a function pointer.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/1847

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 11 New Failures

As of commit f580c78 with merge base ffb4350 ():

NEW FAILURES - The following jobs have failed:

• Build Linux Wheels / build / upload / manywheel-py3_9-cuda11_8 (gh)
  Unable to download artifact(s): Artifact not found for name: pytorch_ao__3.9_cu118_x86_64
• Build Linux Wheels / build / upload / manywheel-py3_9-cuda12_6 (gh)
  Unable to download artifact(s): Artifact not found for name: pytorch_ao__3.9_cu126_x86_64
• Build Linux Wheels / build / upload / manywheel-py3_9-cuda12_8 (gh)
  Unable to download artifact(s): Artifact not found for name: pytorch_ao__3.9_cu128_x86_64
• Code Analysis with Ruff / build (3.9) (gh)
  Process completed with exit code 1.
• PR Label Check / Check PR Labels (gh)
  Process completed with exit code 1.
• Run Float8 Tests / test (H100, linux.aws.h100, --pre torch torchvision torchaudio --index-url https://download.pytor... / linux-job (gh)
  RuntimeError: Command docker exec -t 929d78a288e275e3052631e3754186dc05f9e2d308c50116df8f32dd2d0e4275 /exec failed with exit code 1
• Run Float8 Tests / test (SM-89, linux.g6.4xlarge.experimental.nvidia.gpu, --pre torch==2.7.0.dev20250122 --index-url... / linux-job (gh)
  RuntimeError: Command docker exec -t a566c748de485ea2d6064de48a157248276a769786e628096c872dcb6931ba7a /exec failed with exit code 1
• Run Regression Tests / test (CUDA 2.3, linux.g5.12xlarge.nvidia.gpu, torch==2.3.0, cuda, 12.1) / linux-job (gh)
  RuntimeError: Command docker exec -t 5dcc4b0b0b769d5c4fe5e7164612c3cd0a57db11f3e5f600bce57bc7bc9e724e /exec failed with exit code 1
• Run Regression Tests / test (CUDA 2.4, linux.g5.12xlarge.nvidia.gpu, torch==2.4.0, cuda, 12.1) / linux-job (gh)
  RuntimeError: Command docker exec -t 82a171eb0b93b8b2577a600eff1315299311b7a298cfdfc78d186d0e5f143917 /exec failed with exit code 1
• Run Regression Tests / test (CUDA 2.5.1, linux.g5.12xlarge.nvidia.gpu, torch==2.5.1 --index-url https://download.pytorch... / linux-job (gh)
  RuntimeError: Command docker exec -t 0d334d48008c4520c6638d5edfc0ff15e9de88f7c504b74ff1b71f63fbcdee49 /exec failed with exit code 1
• Run Regression Tests / test-nightly (CUDA Nightly, linux.g5.12xlarge.nvidia.gpu, --pre torch==2.7.0.dev20250122 --index-... / linux-job (gh)
  RuntimeError: Command docker exec -t 8e8ed9c1c924c6e498f56c4851d942d6981fe9bccfa32497b4abf0db8073c4b9 /exec failed with exit code 1

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[petrex]

@jcaip let me take a look and maybe add documentation. thanks

[jcaip]

Thanks @petrex, happy to share more debug info / hop on a call and walk you through the issues if that's helpful.