Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!yale!cmcl2!kramden.acf.nyu.edu!brnstnd
From: brnstnd@kramden.acf.nyu.edu (Dan Bernstein)
Newsgroups: comp.lang.misc
Subject: Re: Array references cannot be made optimal
Message-ID: <27139:Nov1920:13:2690@kramden.acf.nyu.edu>
Date: 19 Nov 90 20:13:26 GMT
References: <11839@life.ai.mit.edu> <3289:Nov1323:05:4790@kramden.acf.nyu.edu> <1990Nov18.024745.2424@bushido.uucp>
Organization: IR
Lines: 18

In article <1990Nov18.024745.2424@bushido.uucp> dbc@bushido.uucp (Dave Caswell) writes:
> .There does not exist an algorithm that will convert an (optimal) array
> .program into its optimal pointer form. That summarizes what I've said so
> .far in this thread.
> .What does it mean for real languages? It means that the programmer
> .*must* be allowed to look at an array x from the vantage point of x[a],
> .rather than x[0]. This array shifting doesn't have to look like C's
> Clearly that isn't a proof.  The word "*must*" is inappropriate in that
> context.

You're right. What I should have said was ``If the language doesn't let
the programmer look at [etc.] ... then it will be restricting his
freedom to express certain optimizations.''

To a language designer, that means that the language *must* provide some
equivalent feature; but most people don't care to design languages.

---Dan