gpu-kernel-ako4all
GPU Kernel AKO4ALL
Use this skill to run a disciplined GPU-kernel optimization loop with AKO4ALL as the outer framework and stack-specific bundled references for Triton, CUDA C++/PTX, CUTLASS/CuTe C++, and CuTe DSL.
This is a derivative synthesis, not original material. Every upstream skill or document referenced is copied into this skill under references/ and templates/. Do not go to .claude/skills, .copilot/skills, temporary clones, or source repositories to read the upstream skills. Read the bundled materials. Preserve the attribution in references/source-attribution.md when copying, publishing, or adapting this skill.
Use This Skill When
- optimizing an existing kernel after a real hotspot has been proven
- writing a new Triton, CUDA C++/PTX, CUTLASS/CuTe C++, or CuTe DSL kernel for an AI-infra path
- creating an AKO4ALL microbench harness for a kernel family
- interpreting
nsysorncuresults before changing tiling, memory movement, pipeline, or epilogue structure - porting a kernel between implementation styles while preserving correctness and performance evidence
Do not start here from a vague "make it faster" request. First establish the target kernel, shape family, dtype/layout contract, hardware, and baseline runtime.
Mandatory Reference Gate
Read the AKO loop reference before any implementation:
More from bbuf/sglang-auto-driven-skills
h100
SSH into host `h100_sglang`, enter Docker container `sglang_bbuf`, work in `/sgl-workspace/sglang`, and use the ready H100 remote environment for SGLang development and validation. Use when a task needs remote CUDA work, GPU-backed smoke tests, diffusion checks, or a safe remote copy instead of local-only execution.
34h100-sglang-diffusion
SSH into host `h100_sglang`, enter Docker container `sglang_bbuf`, work in `/data/bbuf/repos/sglang`, and use the ready H100 remote environment for SGLang **diffusion** development and validation. Use when a task needs diffusion model smoke tests, Triton/CUDA kernel validation, torch.compile diffusion checks, or a safe remote copy for diffusion-specific SGLang changes.
33sglang-prod-incident-triage
Replay-first debug flow for SGLang serving problems. Use when a live or recent server shows health-check failures, latency or throughput regressions, queue growth, timeouts, distributed stalls, crash dumps, wrong outputs after deploys, or PD/EP/HiCache issues, and the job is to turn the problem into a replay plus the right next debug tool.
33llm-serving-auto-benchmark
Framework-independent LLM serving benchmark skill for comparing SGLang, vLLM, TensorRT-LLM, or another serving framework. Use when a user wants to find the best deployment command for one model across multiple serving frameworks under the same workload, GPU budget, and latency SLA.
19llm-torch-profiler-analysis
Unified LLM torch-profiler triage skill for `sglang`, `vllm`, and `TensorRT-LLM`. Use it to inspect an existing `trace.json(.gz)` or profile directory, or to drive live profiling against a running server and return one three-table report with kernel, overlap-opportunity, and fuse-pattern tables.
18sglang-sota-performance
End-to-end SGLang SOTA performance workflow. Use when a user names an LLM model and wants SGLang to match or beat the best observed vLLM and TensorRT-LLM serving performance by searching each framework's best deployment command, benchmarking them fairly, profiling SGLang if it is slower, identifying kernel/overlap/fusion bottlenecks, patching SGLang code, and revalidating with real model runs.
15