Top positive review
9 people found this helpful
Quite unconventional, quite useful.
on 27 February 2003
This book is quite a strange case. For sure, it is the most widely used around the world for intro courses on Computer Architecturs (CAs). Could it be because Hennessy and Patterson are, at present and since a long time, two of the most prominent researchers in the field, Hennessy being now also President of Stanford University, Patterson a professor at Berkeley. But it would be too reductive to limit the view to this only. So we should move inside the book and try to understand the real (or other) reasons.
As an introductory text on CA, the approach is different than the somewhat classical one.
Those who'd expect a few introductory chapters on logic design (as, e.g., Mano & Kime's chapters or Murdocca's long appendixes) will find instead a short appendix that describes basic components (gates, registers, clocks and so on) at a high level (never mention digital abstracion & co.).
The path then is not a survey of general concepts & principles of CA with eventually some real examples as application. Instead, the process is a strictly step-by-step constructive one: they build from scratch a new system funding the design with plenty of considerations and tips, even with warnings on most common "fallacies and pitfalls". All this done through a very straightforward and clear language and with lots of figures, well paced and presented. As a result, coping with the topics is pretty an easy task, and the most likely result is a thourough understanding of what they present.
So what they present ? Substantially, the MIPS, a well known (thanks to this book and their authors too, of course) and widely sold (thanks to its true qualities) RISC processor. The authors have been leaders in the development of the RISC architecture, which admittedly is by now the only good choice for CPU designs since even Intel in its newest architectures reduce all down to the execution of RISC instructions. Anyway, the attention is not only on RISC (and MIPS) architectures: it's "mostly" on these, but there's space for short disgressions in the PowerPC, 80x86 and Pentium Pro (the book is dated 1997) field. This is done through a section named "Real stuff" in each chapter, where after they've extensively developed the subpart of the MIPS (be it the ISA, the ALU or Datapath & Control, the Pipeline and so on), they summarily look at how the same concepts have been developed by PowerPC and 80x86 or Pentium.
All in all, if the book has been assigned as a textbook for a course, little integration is needed to understand it and made it useful for the course; or if it is used a self first introduction to computer architectures and especially RISC architecture, the book will prove a very good choice. And this happens simply because the transfer of knowledge is effective as probably the authors have intended it to be.
If what is needed is a reference, then perhaps the step-by-step approach would suggest other choices (e.g. Tanenbaum, Murdocca, Stallings or Mano & Kime).