Individual skill

cpu-kernel-authoring

Use when writing, optimizing, or benchmarking a C++ CPU kernel with AVX2 or AVX512 intrinsics for the Hugging Face kernels ecosystem. Not for CUDA kernels: use cuda.

.devin/skills/cpu-kernel-authoring/SKILL.md

Read pinned source

Source commit b0e8ce89a19fac880251dc3ea1babfeb4503a4fe

What was checked

Matches the observed repository HEAD: Yes
Scanner rules are current: Yes
Repository scan complete: No
Skill scan complete: No
Observed behavior: Local scripts
Last checked: Nov 4, 2026

These are observations at the recorded commit, not a guarantee of safety. Review the source before running it.

Install this skill

Choose an agent in your Stack or copy a command below. Commands fetch the pinned commit and copy only this skill directory. Existing destinations are refused. Copying a command does not install anything.

An install command is unavailable: current rules, complete repository and skill scans, an eligible behavior level and a permitted license are required.

Repository context

Repository stars (not skill ratings): 55
Repository last updated: Sep 28, 2026
License: Apache-2.0

Observed agent integration: No specific integration evidence. Standard SKILL.md installation does not imply verified compatibility.

odd Repository topic
odin Repository topic
outline Repository topic
outlines Repository topic
coding-paradigm Repository topic
outline-driven-development Repository topic
claude-plugins Repository topic
codex-cli Repository topic
gemini-cli-extensions Repository topic
outline-driven Repository topic
writing Derived category

Files and excerpt

File metadata and permitted, redacted excerpts come from the recorded scan. Source links are pinned to that commit.

Browse files

SKILL.mdcore · 10,944 bytes
SKILL.mdcore · 10,944 bytes
SKILL.mdcore · 10,944 bytes
agents/openai.yamlasset · 250 bytes
references/brgemm_patterns.yamlasset · 14,484 bytes
references/build_system.mddoc · 7,931 bytes
references/correctness.yamlasset · 6,037 bytes
references/dtype_optimizations.yamlasset · 4,214 bytes
references/huggingface-kernels-integration.mddoc · 4,550 bytes
references/implementation_reference.mddoc · 12,527 bytes
references/memory_patterns.yamlasset · 3,559 bytes
references/optimization_levels.yamlasset · 5,946 bytes
references/optimization_strategies.mddoc · 4,919 bytes
references/quantized_gemm_patterns.yamlasset · 19,032 bytes
references/runtime_dispatch.yamlasset · 8,742 bytes
references/simd_optimization_patterns.yamlasset · 7,154 bytes
references/threading_patterns.yamlasset · 3,072 bytes
references/workflow_details.mddoc · 8,921 bytes
scripts/analyze_op.pyruns · 11,282 bytes
scripts/benchmark_cpu.pyruns · 17,051 bytes
scripts/config.pyruns · 1,014 bytes
scripts/config.yamlasset · 641 bytes
scripts/cpu_profiler.pyruns · 11,735 bytes
scripts/trial_manager.pyruns · 21,533 bytes
scripts/validate_cpu_kernel.pyruns · 13,403 bytes

Read excerpt


# CPU kernel authoring

## Contract

| Field | Bound contract |
|---|---|
| Trigger | A C++ CPU kernel for the Hugging Face kernels ecosystem must be written, optimized, or benchmarked with AVX2 or AVX512 intrinsics against a PyTorch baseline. |
| Authority | Reversible local. Writes C++ kernel sources, `build.toml`, and `torch_binding.cpp` under the kernel directory, a wheel under `dist/`, the installed kernel package in the active Python environment, and trial state under `trials/<kernel_name>/` and `output/`. Rollback is version control for the sources, `pip uninstall <package>` for the package, and removal of `dist/`, `trials/<kernel_name>/`, and `output/`. No remote mutation. |
| Side effect | Kernel sources and build files change; a wheel is built and installed; trial directories and result records accumulate. |
| Done | The kernel passes the correctness check in `scripts/benchmark_cpu.py`, every trial up to `max_trials` has run or the speedup exceeded `early_stop_speedup`, and the best trial is finalized into `output/` with its final measurement; or a failure class from the table below is reported with the recovery step taken. |

## Inputs

- Kernel name (required): the trial-tree label, for example `my_rmsnorm`. Used only by `trial_manager.py`, which accepts it as a single directory name under `trials/`, never a path.