NVIDIA 用 AI Agent 将 CUDA Tile 内核从 Python 翻译为 Rust
原标题：Translating CUDA Tile Operations from Python to Rust Using Agentic AI
AI 摘要
NVIDIA 团队开发了一个 AI agent 技能，将 TileGym 中的 CUDA Tile Python 和 Triton-TileIR 内核翻译为 cuTile Rust。他们用该技能移植了全部 24 个公开 TileGym 算子（约 40 个 GPU 内核），平均达到 cuTile Python 性能的 99.5%。该多智能体流水线包含分析、设备内核、主机与 FFI 代码、基准测试等阶段，并通过 IR diff 和验证脚本进行机器可检查的判定。
正文节选
cuTile Rust (cutile-rs) is a tile-based system for safe, idiomatic GPU kernel authoring in the Rust programming language. Extending the Rust ownership model to tile-based GPU kernels, it splits mutable outputs into disjoint pieces and preserves the host-side ownership contract across kernel launches. It also allows programmers to opt out locally when they need lower-level control, enabling direct execution of Tile IR operations. The TileGym CUDA tile kernel library has accumulated a large librar