Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uflorida!mephisto!mcnc!uvaarpa!murdoch!astsun7.astro.Virginia.EDU!gl8f
From: gl8f@astsun7.astro.Virginia.EDU (Greg Lindahl)
Newsgroups: comp.lang.misc
Subject: Re: The Forbiden
Message-ID: <1990Aug16.204409.4744@murdoch.acc.Virginia.EDU>
Date: 16 Aug 90 20:44:09 GMT
References: <24279@megaron.cs.arizona.edu>
Sender: news@murdoch.acc.Virginia.EDU
Organization: Department of Astronomy, University of Virginia
Lines: 32

In article <24279@megaron.cs.arizona.edu> gudeman@cs.arizona.edu (David Gudeman) writes:
>In article  <6465@helios.ee.lbl.gov> austern@ux5.lbl.gov (Matt Austern) writes:

>>The more assumptions a compiler can make about a program, the better
>>it can do the optimization, and in many cases, the only way a compiler
>>can safely assume that a programmer doesn't do something is to make it
>>illegal.
>
>This is the assumption that I find obviously false.

Try this for an example: FORTRAN doesn't allow aliased arguments in
subroutine calls. This allows FORTRAN compilers to make all sorts of
nice optimizations. Although it is possible to insert run-time checks
to make sure aliasing doesn't happen, IF the programmer declared the
array to the proper size in the subroutine, I have yet to meet a
compiler which does so.

I might also note that the Convex vectorizing C compiler has a switch
which allows the compiler to assume that arguments are not aliased.
Again, no run-time check is available, but it is theoretically
possible.

I'd write valid programs and like it when they run fast. I like
compilers that can warn me of invalid programs at compile or run-time,
but I hate compilers which skip optimizations because it might trip an
invalid program. I'll debug where I have good compilers, and run where
the machine is fast. If I'm using C, I'm willing to write with the
restriction of no aliasing.


--
"In fact you should not be involved in IRC." -- Phil Howard