skills.rest
sayalinvidia
sayalinvidiaCOMMUNITY
Explore all skills by this author →
GitHub

tilegym-adding-cutile-kernel

Add a cuTile kernel operator to TileGym with dispatch registration and benchmarking.

Updated Apr 16, 2026
Search Tags:#testing#backend#benchmark#kernel#cuda#cutile#tilegym
One-click install
npx skills add https://github.com/sayalinvidia/sayali-skills-test --skill tilegym-adding-cutile-kernel
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: tilegym-adding-cutile-kernel
Source: https://github.com/sayalinvidia/sayali-skills-test/tree/main/skills/tilegym-adding-cutile-kernel
Command: npx skills add https://github.com/sayalinvidia/sayali-skills-test --skill tilegym-adding-cutile-kernel

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill provides a complete, end-to-end workflow for adding a new cuTile kernel operator to TileGym, including dispatch registration, backend implementation, exports, tests, and performance benchmarking.

Core Features & Use Cases

  • End-to-end operator development: from dispatch to backend and tests, enabling seamless integration of new cuTile kernels.
  • Automated validation and benchmarking: ensures correctness, safety, and performance before publication.
  • Use Case: When introducing a new cuTile op (e.g., my_op) to TileGym, this skill guides the developer through the full lifecycle, including tests and benchmarks.

Quick Start

Follow the six-step workflow to register and implement a new cuTile kernel operator in TileGym, and run the accompanying benchmarks.

Frequently Asked Questions about tilegym-adding-cutile-kernel

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I add a cuTile kernel operator to TileGym?▼

To add a cuTile kernel to TileGym, follow the six-step workflow covering dispatch registration, backend implementation, exports, and pytest validation before publication.

What is the process for benchmarking a new cuTile kernel in TileGym?▼

Benchmarking a cuTile kernel in TileGym involves running automated performance validation after backend implementation and pytest checks to ensure safety and correctness before publication.

Do I need to run pytest validation before publishing a cuTile op to TileGym?▼

Yes, pytest validation and linting are enforced repository conventions that must pass before a new cuTile operator can be published to TileGym.

What does dispatch registration involve when integrating a cuTile kernel?▼

Dispatch registration is the first stage of cuTile kernel integration, linking the new operator into TileGym before proceeding to backend implementation and testing.

Can I update an existing cuTile op in TileGym end-to-end?▼

Yes, the TileGym workflow applies when creating, updating, or integrating a cuTile op, guiding you through dispatch, backend implementation, tests, and benchmarks.

What are the limitations of adding a backend implementation for a cuTile kernel?▼

The workflow requires strict adherence to repository conventions, passing linting and pytest validation, meaning any backend implementation failing these checks cannot be published.

Related Skills

Explore More Software Engineering Skills→
NVIDIANVIDIA
796

monkey-patch-kernels-to-transformers

Monkey-patch TileGym CUDA kernels into HuggingFace Transformers models.

Official
Advanced
yo-stevenyo-steven

adding-cutile-kernel

Adds a cuTile GPU kernel operator to TileGm with full integration.

Community
Advanced
NVIDIANVIDIA
796

adding-cutile-kernel

Automate adding cuTile kernel operators to TileGym with registration, backend, tests, and benchmarks.

Official
Advanced
jarmak-personaljarmak-personal

precision-compliance

Verify and wire dual-precision dispatch plans for CUDA spatial kernels.

Community
Advanced