Get cuTile GPU kernels checked for correctness.
What it does
Creates, debugs, and optimizes cuTile GPU kernels with numerical validation.
When to use it
Use for writing, converting, debugging, optimizing, or validating cuTile kernels, or answering cuTile API questions.
How to use it
Give it a kernel implementation request; it writes a Python file and runs validation, asking for clarification when the implementation is ambiguous.
Uses
TI
CU
Good to know
- starts one helper agent per item, so the count depends on the work
Latest review
What you need · 4
Requires Python to execute generated kernels for the mandatory validation loop.
Requires cuTile, imported as cuda.tile in the generated kernel structure.
Access · 3
This skill
TileGym
Read
Reads TileGym kernel examples
https://docs.nvidia.com/cuda/cutile-python
Read
Consults the cuTile language specification