From: "Rob Slade, Social Convener to the Net" <roberts@mukluk.decus.ca>
Newsgroups: rec.arts.books.reviews
Subject: Review of High Performance Computing by Dowd
Followup-To: alt.books.technical,misc.books.technical
Date: 30 Jan 1995 19:42:48 MET
Organization: University of Colorado at Boulder
Originator: brock@ucsub.Colorado.EDU

BKHGHPRF.RVW  941202
 
%A   Kevin Dowd dowd@maddie.atlantic.com
%C   103 Morris Street, Suite A, Sebastopol, CA   95472
%D   1993
%G   1-56592-032-5, GSA #GS-02F-6095A
%I   O'Reilly & Associates, Inc.
%O   U$25.95 800-998-9938 707-829-0515 fax: 707-829-0104 nuts@ora.com
%P   371
%T   "High Performance Computing", 
"High Performance Computing", Dowd, 1993, 1-56592-032-5, U$25.95
We were setting up the DECUS Symposium in, for a wonder, my home town. The security guard who had been spotted at the door of the demo area was quite keen on the assignment, as he was a bit of a computer freak who was inordinately proud of just having purchased a 66 MHz 486DX2, then the fastest available clock speed in the "personal" computer lines. He was very interested in what we were setting up. Unfortunately, when he got around to asking what the clock speed on "this baby" was, he picked an Alpha. I was somewhat apologetic in telling him that this line of CPUs started at 150 MHz. "So," he finally stammered, "that meant it was about twice as fast as his new machine?" Well, no, it had a different architecture. Depending upon what you were doing, it was probably six to eight times as fast.

You have to hand it to him. He was much more interested in the new vistas of computing that this opened up, than disappointed to find his machine's limits.

"Home" computers are growing in power, workstations are dropping in price, microprocessors are powering small mainframes. We are poised on the cusp of the choice of a new CPU platform. Pipelines, caches, superscalars and RISC (most *especially* RISC) abound. What does it mean? How do you choose?

The promotional material for this book says that it tells programmers how to tweak code for the new processor architectures. So it does. What they *don't* say is that, once again, they have reached into the depths of the Sebastopol Institute for the Insanely Entertaining and come up with one Kevin Dowd, who explains memory with vacuum cleaners, benchmarks with pie eating contests, and pipelining with kids in the sewer. This is a book on computer architecture that is fun to read. Do you realize how bizarre that is?

Dowd does, however, know his stuff. Part one covers RISC, memory and optimising compilers. Part two, dealing with porting and tuning code, addresses clarity, identifying porting problems, timing and profiling parallelism, clutter, loops, memory reference, and language issues. Part three looks at benchmarks and performance, while part four gives an introduction to parallel computing.

Listen, an *executive* could read this. And there's a good chance he'd even *understand* it!

 
copyright Robert M. Slade, 1994   BKHGHPRF.RVW  941202
 
============= 
Vancouver      ROBERTS@decus.ca         | Life is
Institute for  Robert_Slade@sfu.ca      | unpredictable:
Research into  rslade@cue.bc.ca         | eat dessert
User           p1@CyberStore.ca         | first.
Security       Canada V7K 2G6           | 

This page was generated on January 31, 1995 by text2html.Bug reports and comments to : mfx@pobox.com.