← all publishers

yuguo-jack

@yuguo-jack source repo

Report this profile Claim this profile

• 5 skills
• 0 followers
• 6 hours ago last updated

1. ▌
   Remote Ssh Docker Workflow · yuguo-jack

   Run remote compile, test, profiling, and debug tasks through SSH plus docker exec while keeping code edits local and synced to the remote node. Use when Codex must validate environment readiness, check ROCm/DTK/Hygon GPU card status, inspect Python packages, verify host-to-container workspace mounts, or execute project commands inside a remote container. Defaults for this cuda-optimized-skill workspace are hg@10.17.176.13:22, Docker container megamoe, host /home/hg/yuguo mapped to container /workspace, and repo path /home/hg/yuguo/cuda-optimized-skill mapped to /workspace/cuda-optimized-skill.
   0
   installs
2. ▌
   Cuda Kernel Optimizer · yuguo-jack bundle

   CUDA Kernel Iterative Optimizer (v2 — Roofline-Driven)
   0
   installs
3. ▌
   Hygon Hip Kernel Optimizer · yuguo-jack bundle

   Iteratively optimize Hygon DCU HIP / CK Tile kernels against a Python reference using hipprof, DTK tools, dccobjdump ISA verification, roofline-style budgeting, branch selection, ablation attribution, and gfx936/gfx938-aware optimization references. Use when the user asks to optimize HIP kernels, Hygon DCU kernels, gfx936/gfx938 kernels, CK Tile kernels, port CUDA kernel tuning workflows to DCU, validate DCU ISA patterns, or reason about DCU-specific inline assembly and source-backed HCU/AMDGPU builtins.
   0
   installs
4. ▌
   Hygon Hip Baseline Generator · yuguo-jack bundle

   Generate a Hygon DCU HIP/C++ baseline kernel and correctness harness from a Torch, Triton, TileLang, Python, or CUDA/C++ reference plus shape JSON, including evidence-backed CUDA-to-HIP/DCU conversion, then hand the validated baseline to the Hygon HIP kernel optimizer. Use when the user has no initial HIP/C++ kernel, asks to start from a ref implementation, provides only ref.py and shape/dims, needs CUDA source ported to HIP/DCU, or wants automatic baseline generation before iterative DCU optimization.
   0
   installs
5. ▌
   Hygon Triton Kernel Optimizer · yuguo-jack bundle

   Capture, triage, benchmark, and optimize TorchInductor or hand-written Triton kernels on Hygon DCU gfx936/gfx938. Use when investigating torch.compile or TorchInductor Triton performance on DCU, parsing Triton autotune logs, saving generated kernels and inputs, checking AMDGCN buffer/global load-store codegen, applying tl.assume or tl.multiple_of hints, diagnosing low-bandwidth pointwise/reduction/scatter kernels, or deciding whether to tune Triton, disable torch.compile for a region, or rewrite model code.
   0
   installs