tilegym-converting-cutile-to-triton

Maintained by nvidia

Converts cuTile GPU kernels (@ct.kernel) to Triton (@triton.jit). Handles standard in-repo conversion, debugging (cudaErrorIllegalAddress, shape mismatch, numerical mismatch), and mapping cuTile idioms (ct.load/ct.store, ct.Constant, ct.launch) to Triton equivalents. Covers dual-kernel layout flags (e.g. transpose=True

Current version: 1.0.0
License: CC-BY-4.0 AND Apache-2.0
Network access: Unknown / not assessed
Review status: Not verified

Problem it solves

This catalog entry helps users find and evaluate tilegym-converting-cutile-to-triton for the task described by its available catalog summary. Confirm the exact scope in the linked original source when one is available.

When to use it

Consider tilegym-converting-cutile-to-triton when its available catalog summary matches the task at hand. When available, review the linked original source before use for precise instructions, requirements, and limitations.

tilegym-converting-cutile-to-triton is listed as an agent skill in RefHub. The listed maintainer is nvidia. The available catalog summary is: Converts cuTile GPU kernels (@ct.kernel) to Triton (@triton.jit). Handles standard in-repo conversion, debugging (cudaErrorIllegalAddress, shape mismatch, numerical mismatch), and mapping cuTile idioms (ct.load/ct.store, ct.Constant, ct.launch) to Triton equivalents. Covers dual-kernel layout flags (e.g. transpose=True When available, review the linked original source for exact usage instructions, required tools, and limitations.

Installation and updates

These commands are displayed for copying only and are never executed on RefHub servers. Review the linked upstream source before running them.

Install command

npx skills add nvidia/skills --skill tilegym-converting-cutile-to-triton -y

Agent compatibility

No compatibility test has been recorded

Do not assume agent compatibility until documented test evidence is available.