Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!ccu.umanitoba.ca!herald.usask.ca!alberta!aunro!ukma!psuvax1!wuarchive!zaphod.mps.ohio-state.edu!uakari.primate.wisc.edu!aplcen!visual1.jhuapl.edu!dave
From: dave@visual1.jhuapl.edu (Dave Weintraub)
Newsgroups: comp.lang.apl
Subject: ACORN, and other niceties
Message-ID: <1991Jun11.200234.6130@aplcen.apl.jhu.edu>
Date: 11 Jun 91 20:02:34 GMT
Sender: news@aplcen.apl.jhu.edu (USENET News System)
Reply-To: dave@visual1.jhuapl.edu (Dave Weintraub)
Organization: Johns Hopkins University
Lines: 40

This is being posted on INTERNET by
     dave@visual1.jhuapl.edu.
I can also be reached as
     dave@aplvm.bitnet
Dave Weintraub   The Applied Physics Laboratory  The Johns Hopkins University
   Johns Hopkins Road 3-145    Laurel MD 20723-6099   USA   (301)-953-5839
=========
I must relate an incident from last week...

I attended a seminar sponsored by CONVEX on issues and directions.  One speaker
was Ken Kennedy, of Rice University, speaking about language directions
for parallel processing.  He described intense ongoing work in FORTRAN
analysis and modifications for FORTRAN 90/2000 to allow specification
of machine-dependent parallel operations.

While I know that FORTRAN is the language of choice for most engineers
(I do not *agree*, I just accept this with a sigh),
I *had* to ask him about APL as an alternative.  I described
Bob Bernecke's ACORN work, and pointed out that Bob had done
in about a man week what Rice was spending man decades trying to do.
His replies (now remember, this is an industry expert speaking at a
public forum):

0.  "APL is not a good language for doing parallel algorithms."  His reasoning?
People take huge outer products and then reduce or slice along a subarray,
and thus the algorithms are inefficient (I tried to explain about
properly-formed arguments for inner product, and about the RANK operator).

1.  Bob's work, as described by me, was "an obvious fabrication".  Noone
could build an interlanguage translation scheme that efficient, that quickly.
And even if they could, it would be useless because of point 0.

2.  APL is difficult to learn (no interest in data indicating otherwise).


)FLAME OFF
as they (sort of) say in the discussion lists...
but I am still boiling!

Dave