> ## Documentation Index
> Fetch the complete documentation index at: https://se7en.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# 书籍概览

<img src="https://mintcdn.com/se7en/5ANsthJq5rhHDKJi/images/ai-systems-performance-engineering.jpg?fit=max&auto=format&n=5ANsthJq5rhHDKJi&q=85&s=b947408c6d7138c5d3df488b6db4dd57" alt="AI Systems Performance Engineering 封面" style={{ float: 'right', width: '180px', marginLeft: '20px', marginBottom: '10px' }} width="1152" height="1371" data-path="images/ai-systems-performance-engineering.jpg" />

## 关于本书

**AI Systems Performance Engineering: Optimizing Model Training and Inference Workloads with GPUs, CUDA, and PyTorch** — Chris Fregly 著，O'Reilly 出版。

本书探讨如何从系统层面优化 AI 工作负载的性能，涵盖硬件架构、软件栈、训练与推理优化等关键主题。

## 阅读目标

* 理解 AI 系统性能的核心瓶颈与优化方向
* 掌握从硬件到软件的全栈性能分析方法
* 能够在实际项目中应用性能工程的思维方式

## 当前进度

<Info>
  首轮阅读计划已确定前五次安排，后续章节将根据实际阅读节奏持续更新。请查看[阅读计划](/books/ai-systems-performance-engineering/schedule)了解详情。
</Info>

## 章节目录

* [第一章：介绍与 AI 系统概览](/books/ai-systems-performance-engineering/chapters/ch01)
* [第二章：AI 系统硬件概览](/books/ai-systems-performance-engineering/chapters/ch02)
* [第三章：GPU 环境下的 OS、Docker 与 Kubernetes 调优](/books/ai-systems-performance-engineering/chapters/ch03)
* [第四章：分布式网络通信调优](/books/ai-systems-performance-engineering/chapters/ch04)
* [第五章：基于 GPU 的存储 I/O 优化](/books/ai-systems-performance-engineering/chapters/ch05)
* [第六章：GPU 架构、CUDA 编程与最大化占用率](/books/ai-systems-performance-engineering/chapters/ch06)
* [第七章：GPU 内存访问模式的分析与调优](/books/ai-systems-performance-engineering/chapters/ch07)
* [第八章：占用率调优、Warp 效率与指令级并行](/books/ai-systems-performance-engineering/chapters/ch08)
* [第九章：提升 CUDA Kernel 效率与算术强度](/books/ai-systems-performance-engineering/chapters/ch09)
* [第十章：Kernel 内流水线、Warp 特化与协作线程块集群](/books/ai-systems-performance-engineering/chapters/ch10)
* [第十一章：Kernel 间流水线、同步与 CUDA 流有序内存分配](/books/ai-systems-performance-engineering/chapters/ch11)
* [第十二章：动态调度、CUDA Graphs 与设备发起的 Kernel 编排](/books/ai-systems-performance-engineering/chapters/ch12)
* [第十三章：PyTorch 的分析、调优与扩展](/books/ai-systems-performance-engineering/chapters/ch13)
* [第十四章：PyTorch 编译器、OpenAI Triton 与 XLA 后端](/books/ai-systems-performance-engineering/chapters/ch14)
* [第十五章：多节点推理、并行、解码与路由优化](/books/ai-systems-performance-engineering/chapters/ch15)
* [第十六章：大规模推理的分析、调试与调优](/books/ai-systems-performance-engineering/chapters/ch16)
* [第十七章：推理中分离式 Prefill 与 Decode 的扩展](/books/ai-systems-performance-engineering/chapters/ch17)
* [第十八章：高级 Prefill-Decode 与 KV Cache 调优](/books/ai-systems-performance-engineering/chapters/ch18)
* [第十九章：动态与自适应推理引擎优化](/books/ai-systems-performance-engineering/chapters/ch19)
* [第二十章：AI 辅助性能优化与百万级 GPU 集群扩展](/books/ai-systems-performance-engineering/chapters/ch20)