Power and Performance: Software Analysis and Optimization

Why Performance and Power Belong in the Same Conversation 💻

Modern software does not run in a vacuum. Every instruction executes on real hardware with real power and thermal limits. Power and Performance: Software Analysis and Optimization shows software engineers how to measure what their code actually does on Intel processors and Linux systems—and then make it faster, leaner, and more power-efficient.

What the Book Covers

Jim Kukunas builds the material in four parts, starting with the architecture and operating-system background you need before tuning anything.

• Intel architecture and the Linux software stack: how hardware and OS interact, from early Intel architecture to modern power-management features.
• Performance analysis methodologies: how to profile systematically instead of guessing where the bottlenecks are.
• Profiling tools in practice: using Intel VTune Amplifier XE and perf to collect meaningful data.
• Common performance problems and fixes: practical starting points for correcting issues the tools reveal.

From Measurement to Optimization ⚙️

The book treats optimization as a data-driven craft. It explains why the famous warning about premature optimization is often misread, and why Big-O analysis alone is not enough when caches, prefetching, and out-of-order execution sit between your source code and the silicon. You will see how frequency scaling and deep sleep states shape power use, and why the “race to idle” principle usually favors doing work quickly so the CPU can return to a low-power state.

AT&T; assembly syntax is used throughout the assembly examples, keeping the low-level discussion consistent with common Linux tooling.

Who This Book Is For

Software engineers, systems developers, and performance analysts who work with Intel hardware and Linux will find the material directly applicable. It is also useful for developers who suspect their code is slower or more power-hungry than it needs to be but need a structured way to find out. The book does not promise magic fixes; it teaches a repeatable process of measurement, analysis, and targeted improvement.

Get the Digital Edition

This digital edition gives you searchable, portable access to the full text. If you care about how your software performs on real machines—not just whether it compiles—this is a grounded, technical guide from Morgan Kaufmann.