Check what MOAT needs from a person -- ports approved and ready to submit upstream, maintainers waiting on a reply, upstream PRs that merged or closed, and landed work going stale. Use at the start of a session, after approving a port, when a maintainer…
Port a CUDA project to ROCm/HIP for AMD GPUs. Strategy selection, the AMD-strict-where-CUDA-is-lenient fault classes, wavefront-size portability, library substitutions, and the validation bar. Use when planning, implementing, reviewing or validating a…
Port a CUDA project to ROCm/HIP for AMD GPUs. Strategy selection, the AMD-strict-where-CUDA-is-lenient fault classes, wavefront-size portability, library substitutions, and the validation bar. Use when planning, implementing, reviewing or validating a…
Publish the intake queue and carry a person's decision back into state. Use after a batch of intake screens finishes, when someone replies to the queue issue, or when asked what is waiting on a fork-or-decline decision.
Check what MOAT needs from a person -- ports approved and ready to submit upstream, maintainers waiting on a reply, upstream PRs that merged or closed, and landed work going stale. Use at the start of a session, after approving a port, when a maintainer…
Resume MOAT. Detect this host's AMD platform, find the next actionable project, and dispatch the right role. Use when starting work in the MOAT repo or when asked to continue porting or "port next".
Review a CUDA-to-ROCm/HIP port for correctness, minimal footprint, the AMD fault classes, build-system correctness, test adequacy, and not breaking upstream CUDA/CPU behavior. Use when reviewing a port branch, when asked to review code changes, or when the…
Resume MOAT. Detect this host's AMD platform, find the next actionable project, and dispatch the right role. Use when starting work in the MOAT repo or when asked to continue porting or "port next".