Path: utzoo!attcan!uunet!europa.asd.contel.com!sura.net!haven!adm!cmcl2!kramden.acf.nyu.edu!brnstnd
From: brnstnd@kramden.acf.nyu.edu (Dan Bernstein)
Newsgroups: comp.arch
Subject: Re: What the compiler won't do you for you
Message-ID: <24280:Feb2802:55:4991@kramden.acf.nyu.edu>
Date: 28 Feb 91 02:55:49 GMT
References: <8728@exodus.Eng.Sun.COM> <13578:Feb2705:46:4391@kramden.acf.nyu.edu> <8787@exodus.Eng.Sun.COM>
Organization: IR
Lines: 40

In article <8787@exodus.Eng.Sun.COM> chased@rbbb.Eng.Sun.COM (David Chase) writes:
> Simpler method:
> compiler writer writes library of inlineable code.
> compiler writer documents names of routines that are in this library.
> programmer uses those names.

I give up. The man just refuses to see that this does NOT produce
portable code. Sun writes one library, and DEC writes another one, and
Convex writes another one, and a programmer who uses these libraries
will NOT be able to run his programs on more than one machine.

> No need to promise that
> in the future, a special chunk of code will result in the use of a
> special instruction.

Nor does he see that the compiler writer NEVER has to make such a
promise. Again it is obvious that his viewpoint is twisted by his
Sun-specific environment. Too bad for him that the rest of the world
uses hundreds of different architectures.

If David opens his eyes he will see that what I am proposing will ALWAYS
improve the portability of efficient code. But because this requires a
bit of work on the compiler-writer's part, his eyes will remain closed.

> In addition, as I said before, there is ample evidence that time is
> better spent doing a good job of optimizing code in a general way.  If
> it paid to do the things that you propose, it would have showed up in
> compiler tuning.  People do look at the code that is generated, and
> try to come up with ways to make it better.  Somehow, nothing like
> your idea even made it onto the list, let alone near the top.

Actually, David, I found out from e-mail that essentially the same ideas
have been published before, in many separate papers.

I still find it amazing that David can draw a line between optimizing
*(p+a) into a single instruction on a Convex, optimizing c = a/b and
d = a%b into a single instruction on most CISC chips, and optimizing a
short Hamming weight function into a single instruction on a Cyber.

---Dan