---
id: 20260730-T0-08
title: "Kernel Forge：用LLM自动生成和优化CUDA内核"
title_en: "Kernel Forge: LLM-Based Generation and Optimization of CUDA Kernels"
url: https://ai.daily.yangsir.net/daily/20260730-T0-08
issue_date: 2026-07-30
publish_date: 2026-07-29T04:00:00.000Z
category: tools
source_name: "arXiv cs.AI"
source_url: https://arxiv.org/abs/2607.24762
---

# Kernel Forge：用LLM自动生成和优化CUDA内核

arXiv 新论文介绍 Kernel Forge，一个基于LLM的智能体框架，用于自动生成和优化CUDA计算内核。当前主流模型运行时间集中在矩阵乘法、卷积等少数核心算子。Kernel Forge 能自动为这些算子编写和优化CUDA代码，减少人工调优成本，有望提升从硬件加速到模型部署的整个链条效率。

## English Version

**Kernel Forge: LLM-Based Generation and Optimization of CUDA Kernels**

A new arXiv paper introduces Kernel Forge, an LLM-based agent framework for automatic generation and optimization of CUDA kernels. Since most ML model runtime is concentrated on a few core operators like matrix multiplication and convolution, Kernel Forge automates the writing and optimization of CUDA code for these operators, reducing manual tuning costs and enhancing the efficiency of the entire hardware-to-deployment pipeline.

---

**来源**：[arXiv cs.AI](https://arxiv.org/abs/2607.24762)

**详情页**：https://ai.daily.yangsir.net/daily/20260730-T0-08

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*