Newsgroups: comp.lang.apl
Path: watmath!watserv1!torn!utcsri!rpi!usc!wupost!uunet!haven.umd.edu!socrates!socrates!rockwell
From: rockwell@socrates.umd.edu (Raul Deluth Miller-Rockwell)
Subject: Re: SIGNUM of teaching numerical methods
In-Reply-To: andrew@rentec.com's message of 20 Jul 92 00:15:40 GMT
Message-ID: <ROCKWELL.92Jul20011749@socrates.umd.edu>
Sender: rockwell@socrates.umd.edu (Raul Deluth Miller-Rockwell)
Organization: Traveller
References: <1992Jul17.171448.13672@chpc.utexas.edu> <1079@kepler1.rentec.com>
	<ROCKWELL.92Jul19155653@socrates.umd.edu> <1089@kepler1.rentec.com>
Date: Mon, 20 Jul 1992 06:17:49 GMT
Lines: 40

Andrew Mullhaupt:
   APL doesn't cut it in scientific computing despite the fact that
   something not unlike APL would be quite nice. J goes in the
   opposite direction of what is wanted.

You're not looking far enough ahead, imho.

Andrew:
   >   Oh by the way - I should point out that Splus isn't perfect
   >   either - it has the stupidest function calling ever implemented
   >   in an interpreter - it _copies_ every argument so that
   >   functions can't modify their arguments, _no matter how large
   >   they are_.

Raul:
   >This does allow fast implementation of the language. 

Andrew:
   But it prevents an implementation of the language from being fast.

It does no such thing.

   No. Having talked to the people who implement it, they maintain
   that they are constrained by the language spec. (And why shouldn't
   they be?)

I don't know about SPlus, but Tim Budd's work indicates that global
optimization is a viable technique with APL.  [The reason it might not
work for SPlus is that it's O(n^3) in code length -- though there are
some optimizations you can do if the code has nice properties.]

With a globally optimizing compiler you can get the kinds of
serial-machine optimizations you're looking for, even with
call-by-value semantics.

And, J seems to be doing a lot of the right things for this kind of an
implementation to be viable.

-- 
Raul Deluth Miller-Rockwell                   <rockwell@socrates.umd.edu>
