Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: Notesfiles $Revision: 1.7.0.10 $; site ccvaxa
Path: utzoo!watmath!clyde!cbosgd!ihnp4!pur-ee!uiucdcs!ccvaxa!preece
From: preece@ccvaxa.UUCP
Newsgroups: net.lang
Subject: Re: Efficiency of Languages (and co
Message-ID: <800003@ccvaxa>
Date: Wed, 6-Nov-85 12:31:00 EST
Article-I.D.: ccvaxa.800003
Posted: Wed Nov  6 12:31:00 1985
Date-Received: Fri, 8-Nov-85 07:25:36 EST
References: <189@opus.UUCP>
Lines: 38
Nf-ID: #R:opus.UUCP:189:ccvaxa:800003:000:1785
Nf-From: ccvaxa.UUCP!preece    Nov  6 11:31:00 1985


[Quoted from a posting responding to a complaint that algorithm
analysis tended to consider architectural rather than computational
issues.  I don't have the name of the original author.]
> On the contrary, analysis of complexity of algorithms is generally done
> with considerable care to separate these issues and identify the
> assumptions about architecture implied by the algorithm.  If you're not
> accustomed to reading in the area, you may find yourself a little
> befuddled because you don't understand some of the assumptions commonly
> left implicit.
> Dick Dunn	{hao,ucbvax,allegra}!nbires!rcd		(303)444-5710
> x3086
----------
Well, you both missed the major architectural assumption in the
examples you were arguing about, which is that the number of
compares is a reasonable measure of cost.  This assumption is
usually mentioned in texts on the subject.  Certainly in many
cases there are other, more appropriate measures (in sorting,
for instance, the cost of moving data may sometimes be a much
larger factor than the cost of the compares).

You also both missed the point that throwing N processors at
a problem does not change its complexity -- all those compares
are still being done SOMEWHERE -- even though it does change
the elapsed time to completion (again raising the question
of what is a useful measure of complexity).

Overall I'd have to say that the first author was more correct:
the common conception of algorithm analysis does tend to include
major architectural assumptions that may not be valid for
unconventional and innovative architectures.  Those assumptions
ARE usually considered by those who write in the discipline,
but are often ignored by those who have merely studied it.

-- 
scott preece
gould/csd - urbana
ihnp4!uiucdcs!ccvaxa!preece