23 tagged gpu computing, measured the same way as everything else here.
Browse within: llm-inference 19Quantization 11amd-gpu 11hip 11machine-learning 11compilers 8inference 8llmops 8
Skill Claude CodeCodex
NVIDIA-specific TornadoVM workflows on the CUDA backend — profile with Nsight Systems (nsys) and read NVTX ranges, use the hybrid library-task API (cuBLAS/cuBLASLt/cuDNN/cuFFT/cuSPARSE/cuTENSOR/CUTLASS/NCCL native calls from Java), print the generated CUDA-C kernel, and fuzz-test the CUDA-C backend with tornado-fuzz.…
Skill Claude CodeCodex
Run a measured performance/correctness campaign on the TornadoVM runtime — profile first (JFR for host, Nsight Systems for driver), falsify the hypothesis with a standalone probe, A/B against a same-build baseline, then ship one PR per measured claim. Use when asked to make TornadoVM faster, to find where time goes in…
Skill Claude CodeCodex
Build, test, and contribute to TornadoVM's OpenCL/CUDA/Metal backends — make BACKEND targets, tornado-test (quickPass/single-class), checkstyle-before-PR, one-line commits, the develop-branch PR flow, and the recurring graalJars/setvars build gotchas. Use when building the project, running or adding unit tests, or…
Skill Claude CodeCodex
Use for hipfire quant calibration, imatrix-driven experiments, KLD/PPL quality evaluation, k-map/format selection, MQ/HFQ/HFP/MFP tradeoff work, ParoQuant-style weight transform planning, and KV policy planning. Use when deciding whether a calibrated model candidate should be promoted, rejected, packaged, or sent…
Skill Claude CodeCodex
Use Kernel Atlas to collect phase-aware hipfire measurements and render ISA Fit View visualizations for AMD GPU kernels, quant formats, and architectures. Use when a user asks how MQ/HFQ/HFP/Q8 quants occupy hardware, asks for an ASCII ISA visualization, wants to compare gfx1010/gfx1030/gfx11/gfx12 kernel fit, or…
Skill Claude CodeCodex
Use when porting a hipfire feature/fix branch authored against pre-0.1.20 master onto post-modular master. Walks through the engine→hipfire-runtime + per-arch-crate split mechanically, then surfaces semantic conflicts that need human judgment.
Skill Claude CodeCodex
Onboard or periodically reverify and benchmark a Hugging Face model on an exact target GPU platform. Use when asked to add a model recipe, refresh a maintained recipe on a supplied GPU server, benchmark serving, create reproducible experiments and a durable results report, fully qualify and tune the model's Emmy…
Skill Claude CodeCodex
Use this skill when the user asks to re-run an article's benchmarks, reproduce blog post numbers, validate that an article URL still holds, check whether the latest code still performs like a published post, or otherwise compare re-measured Emmy results with published results. It fetches the article, finds its…
Skill Claude CodeCodex
Tune Emmy kernels for a Hugging Face model, traced IR, or golden YAML. Use when asked to tune a model or golden set, seed MCTS with model-proposed knob configurations, compare hybrid proposals against MCTS-only search, diagnose slow or failing kernels, refresh per-GPU goldens, or produce a per-kernel tuning findings…
Skill Claude CodeCodex
Operate local OpenHCS microscopy workflows through the bundled MCP server. Use when Codex needs to inspect plate data, discover processing functions, author or validate pipelines, compile and execute jobs, control viewers, or interact with a running OpenHCS GUI.