Ezra's Bookshelf

Computer Architecture

by John L. Hennessy and David A. Patterson · 617 pages · ~11 hrs

Science & TechnologyEducation

Recommended on

buy on bookshop.org →
Computer Architecture: A Quantitative Approach is the standard graduate-level textbook on how processors and computer systems are designed, and its founding conviction is right there in the subtitle: architecture should be driven by measurement, not intuition or marketing. John Hennessy and David Patterson argue that every design choice, from instruction sets to memory hierarchies, must be justified by its measurable effect on performance, cost, and energy, and they teach readers to reason with tools such as Amdahl's Law and careful benchmarking. The book walks through the core building blocks of modern machines: instruction-level parallelism and pipelining, caches and memory hierarchy design, vector processors and GPUs, multicore and thread-level parallelism, and warehouse-scale computers that power cloud services. The sixth edition, published in 2017, uses the open RISC-V instruction set for its examples and adds a chapter on domain-specific architectures, including Google's Tensor Processing Unit, arguing that with the end of Moore's Law and Dennard scaling, specialized hardware tuned for particular workloads such as machine learning is the main remaining path to large gains in performance and efficiency. The authors are among the most influential figures in computing. Hennessy led the Stanford MIPS project and later became Stanford's president, and Patterson led the Berkeley RISC project; together they shaped the reduced instruction set approach that underlies most of today's processors, work recognized with the Turing Award. Each chapter ends with real-world case studies. Dense and technical, it is the book to read for anyone who wants to understand, from first principles, why the chips driving the AI boom look the way they do.

For fans of