Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rochester!cornell!uw-beaver!mit-eddie!husc6!necntc!ima!johnl
From: johnl@ima.UUCP
Newsgroups: comp.compilers
Subject: machine independent optimizations
Message-ID: <593@ima.UUCP>
Date: Tue, 16-Jun-87 12:00:00 EDT
Article-I.D.: ima.593
Posted: Tue Jun 16 12:00:00 1987
Date-Received: Sun, 21-Jun-87 14:29:16 EDT
Sender: johnl@ima.UUCP
Lines: 30
Approved: compilers@ima.UUCP

In an earlier article, Bill Wulf is quoted as:
> Finally, a soap-box that I've been on for a long time:
> 
>    There is no such thing as a "machine-independent" optimization!
> 
> Not one! People who use the phrase don't understand the problem!
> There are lots of semantics-preserving transformations that improve
> code-size or speed for some machines under some circumstances. But
> those same transformations may be pessimizations for another machine!

I'm curious about "very high level" program transformations: for
example, operator properties such as associativity can be used to
transform certain non-tail-recursive functions to tail recursive ones
(whence getting an iterative program is straightforward).  Now it seems
to me that transforming a function that takes O(n) stack space -- not
to mention the recursive calls -- to one that uses O(1) space and no
function calls would be a win on pretty much all machines.
[High level transformations seem only to be considered in the context of
optimizing Lisp, probably because in conventional languages it's so hard
that people don't even try.  But I don't understand why the rest of the
compilers community never looks at Lisp. -John]
---
Saumya Debray		CS Department, University of Arizona, Tucson
     internet:   debray@arizona.edu
     uucp:       {allegra, cmcl2, ihnp4} !arizona!debray
--
Send compilers articles to ima!compilers or, in a pinch, to Levine@YALE.ARPA
Plausible paths are { ihnp4 | decvax | cbosgd | harvard | yale | cca}!ima
Please send responses to the originator of the message -- I cannot forward
mail accidentally sent back to compilers.  Meta-mail to ima!compilers-request