Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!iuvax!purdue!mentor.cc.purdue.edu!pur-ee!hankd
From: hankd@pur-ee.UUCP (Hank Dietz)
Newsgroups: comp.arch
Subject: Re: VLIW Architecture
Keywords: vliw
Message-ID: <13038@pur-ee.UUCP>
Date: 2 Oct 89 19:42:19 GMT
References: <251FCB3F.12366@maccs.dcss.mcmaster.ca> <1050@m3.mfci.UUCP> <1626@l.cc.purdue.edu>
Reply-To: hankd@pur-ee.UUCP (Hank Dietz)
Organization: Purdue University Engineering Computer Network
Lines: 41

In article <1626@l.cc.purdue.edu> cik@l.cc.purdue.edu (Herman Rubin) writes:
>In article <1050@m3.mfci.UUCP>, cutler@mfci.UUCP (Ben Cutler) writes:
>> In article <251FCB3F.12366@maccs.dcss.mcmaster.ca> cs4g6ad@maccs.dcss.mcmaster.ca (Custeau     RD) writes:
>| >
>| >   I am looking for references or information on VLIW architectures for
>| >a fourth-year architecture seminar.  Any help would be greatly appreciated.
>  
>> A good ``introductory'' text is ``Bulldog: A Compiler for VLIW Architectures'',
>> by John Ellis, which won the ACM Doctoral Dissertation Award in 1985.  It's
>> available from MIT Press.  For information on a commercial VLIW
>> implementation, contact Multiflow Computer at (203) 488-6090.
>
>As one who always finds ways to use the architecture that the compiler
>writers did not think about, I maintain that this book helps little.

VLIW machine compilers essentially search for optimal schedules of
instructions; I don't see how a full-width search could be ommitting things
that Dr. Rubin would want to do.  ;-)

>Why is it the case that people in the computing field think that someone
>can understand a computer in ignorance of its instruction set and the
>temporal operation of those instructions?

The primary reason humans have done so well is not that we are so much
better than other life forms, but rather that we build much better tools.
To accomplish any large task, a human must be able to be ignorant of at least
some details -- VLIW compiler technology is a prime example of a mechanism
keeping track of, and optimizing use of, a structure too complex for humans
to manage directly.

I suppose you'd rather not use any tools...  but then why do you want to use
the tools called "computers?"

						-hankd@ecn.purdue.edu

>-- 
>Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907
>Phone: (317)494-6054
>hrubin@l.cc.purdue.edu (Internet, bitnet, UUCP)

PS: Dr. Rubin's views do not represent those of Purdue University.  ;-)