Path: utzoo!utgpu!jarvis.csri.toronto.edu!clyde.concordia.ca!uunet!aplcen!samsung!cs.utexas.edu!rutgers!att!drutx!druhi!lu
From: lu@druhi.ATT.COM (david lu)
Newsgroups: comp.arch
Subject: Re: The Art of Computer Architecture
Summary: Hennessy & Patterson book
Message-ID: <4575@druhi.ATT.COM>
Date: 20 Dec 89 17:05:45 GMT
References: <797@mindlink.UUCP> <960020@hpcllla.HP.COM> <1989Dec19.204601.8509@cec1.wustl.edu>
Organization: AT&T, Denver, CO
Lines: 46

In article <1989Dec19.204601.8509@cec1.wustl.edu>, jps@wucs1.wustl.edu (James Sterbenz) writes:
> In article <960020@hpcllla.HP.COM> daryl@hpcllla.HP.COM (Daryl Odnert) writes:
> >There was sufficient interest in the Hennessy and Patterson book that
> >I've decided to post the information here:
> 
> Given the RISC background of the authors, how balanced a book is this,
> i.e. is it a unbiased architecture text or is it a textbook
> on RISC architecture?

I took Dr. Hennessy's class on Computer Architecture at Stanford last
year and we used a pre-beta release version of the said book (Computer
Architecture: A Quantitative Approach).  In other words, the version
of the book was done by Copymat and bound by plastic spirals.  I bring
this up as a disclaimer that most of my comments will be based on this
version, not the bound (most recent) version.

I found the book to be quite insightful and very informative.  Since the
class was being taught by Hennessy, my thinking may have become skewed
in favor of RISC, but looking at the book a second time, I've come to
realize that it is not an overly biased book on RISC.  There are sections
that favor RISC architecture and attempt to show what makes RISC a better
alternative (e.g. pipelining), but as a whole, issues are covered that
can be applied to all systems (e.g. memory hierarchies, performance and
cost evaluation, I/O, what makes a good benchmark).

Though it is not a truely unbaised book on computer architecture (I don't 
think there is such a thing), I found it to be quite good.  Especially
the sections on memory systems and pipelining (However, it does not 
explore the usage of pipelines on CISC machines.  That came in EE382 with
Dr. Flynn).

Since taking the class, I've come across a description of the book and I
see that the authors have included a chapter on vector processors (a 
topic for all types of machines), a chapter parallel and special purpose
processors, and an appendix on computer arithmetic.

Essentially, I found the book to be quite good and not extremely biased
on RISC.  I'm not sure if it makes a good first computer architecture
book, but definitely an excellent second book.

------------------------
These comments are strictly my own.  Any thoughts or discussion are 
welcomed (but no CISC/RISC wars please).
-- 
-David Lu
lu@druhi.att.com	..att!druhi!lu