kernel-tileir-optimization

Installation
SKILL.md

Triton TileIR Optimization

Optimize EXISTING Triton kernels for NVIDIA's TileIR backend on Blackwell GPUs. This skill does NOT write kernels from scratch -- that is the Triton Specialist's job.

Principles

TileIR vs PTX Backend

TileIR is NVIDIA's compiler backend for Triton that generates optimized CUDA code using CGA-level (Cooperative Grid Array) tile representations. Critical differences:

Installs
4
GitHub Stars
14.3K
First Seen
Jul 3, 2026
kernel-tileir-optimization — nvidia/tensorrt-llm