Beyond the Bottleneck: From Classic Systems to Modern AI and HPC
A comprehensive bilingual technical book on software performance measurement, analysis, and optimization. We bridge the gap between theoretical complexity (Big-O) and real-world hardware behavior.
Core Philosophy: "Measure, Don't Assume."
Available in Traditional Chinese (繁中) and English.
- System software engineers (embedded, firmware, OS)
- Performance-conscious application developers
- Anyone who has wondered why their O(1) algorithm is slower than O(n)
| Part | Topics |
|---|---|
| I. Foundations | Benchmarking methodology, measurement techniques, statistical analysis |
| II. Benchmark Suites | CPU benchmarks (CoreMark, SPEC), memory (STREAM), profiling (perf) |
| III. Analysis | Roofline model, algorithm complexity, cache & branch prediction |
| IV. Data Structures | Array vs linked list, hash vs tree, sorting algorithms |
| V. Parallelism | SIMD vectorization, multi-core, memory allocators |
| VI. Embedded | Binary footprint, compiler optimization, stack analysis, RTOS |
| VII. AI/HPC | ML benchmarks, GPU, LLM inference, Edge AI, quantization |
| VIII. Case Studies | Web server, database query, ML inference optimization |
| IX. Synthesis | How to benchmark, how to optimize, CI/CD for performance |
- 35 chapters across 9 parts
- 8 appendices (automation, embedded, I/O, power, exercises, reading paths, environment, performance models)
performance-and-benchmarking-public/
├── README.md
├── LICENSE
├── manuscript/ # English version
│ ├── chapters/ # 35 chapters
│ ├── appendices/ # 8 appendices
│ ├── front_matter/ # Preface, TOC
│ └── back_matter/ # Bibliography
└── manuscript-zh-TW/ # 繁體中文版
├── chapters/ # 35 章
├── appendices/ # 8 附錄
├── front_matter/ # 前言、目錄
└── back_matter/ # 參考文獻
-
Clone the repository
git clone https://github.com/djiangtw/performance-and-benchmarking-public.git cd performance-and-benchmarking-public -
Read the book - Start with
manuscript/chapters/chapter01.md(English) ormanuscript-zh-TW/chapters/chapter01.md(繁中)
This work is licensed under CC BY 4.0.
You are free to share and adapt this material for any purpose, even commercially, as long as you give appropriate credit.
Danny Jiang - GitHub | djiang.tw@gmail.com
This is a read-only public repository. The book is developed in a private repository.
Feedback Welcome:
- Open an issue for typos, errors, or suggestions
- Discussions and questions are encouraged
Note: Pull requests cannot be accepted as this is a one-way sync from the private development repository.
This book builds upon decades of work by the systems performance community. Special thanks to the creators of CoreMark, STREAM, perf, and the countless engineers who have shared their benchmarking knowledge.
January 2026