Port cuTile GPU kernels to tested Triton code.
What it does
Converts cuTile GPU kernels into Triton implementations checked for correctness and performance.
When to use it
Use when porting cuTile kernels to Triton or debugging existing Triton translations.
How to use it
Give it a cuTile kernel in your project; it creates a Triton implementation, tests it, and benchmarks it against cuTile.
What you provide
- An existing project
Latest review
What you need · 6
Requires Python to run the prescribed correctness and performance tests; no version is specified.
The Triton implementations import triton and triton.language.
The cuTile source implementations used as conversion sources and benchmark baselines import cuda.tile.
The supplied kernel implementations use torch for tensors and reference results.
Access · 0
No additional actions listed.