Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!cs.utexas.edu!rice!titan!preston
From: preston@titan.rice.edu (Preston Briggs)
Newsgroups: comp.lang.misc
Subject: Re: Common subexpression optimization
Message-ID: <4978@brazos.Rice.edu>
Date: 14 Feb 90 17:30:02 GMT
References: <4561@scolex.sco.COM> <14214@lambda.UUCP> <PCG.90Jan29131938@rupert.cs.aber.ac.uk> <2217@sunset.MATH.UCLA.EDU> <838.18:06:33@stealth.acf.nyu.edu> <2115@castle.ed.ac.uk> <5453:23:28:32@stealth.acf.nyu.edu> <1041@kunivv1.sci.kun.nl> <3668:04:34:52@stealth.a
Sender: root@rice.edu
Distribution: usa
Organization: Rice University, Houston
Lines: 26

In article <3668:04:34:52@stealth.acf.nyu.edu> brnstnd@stealth.acf.nyu.edu (Dan Bernstein) writes:
>The argument is whether hand optimizations are bad. My point is that
>there are optimizations that current compilers can't do at all in
>practice, even though there are no barriers in theory; if those must
>be done by hand, what's wrong with doing simpler ones by hand? You
>can't draw the line between trivial optimizations and so-called linear
>programming (inductive generalization, trading space for time, or
>whatever you want to call it).

Sure we can draw a line.  If the compiler can, in practise,
save us some work, it should.  Extending your argument will 
have us adding on our fingers again

	I don't need the machine to do this!
	After all, it's not my equal in number theory.

Better seems

	I don't need to do this!
	I've got machines to do it for me.


(Do you suppose we could modify an Eliza to carry on these arguments for us?)

Preston Briggs
preston@titan.rice.edu