From jwagener@amoco.com Mon Aug 16 10:11:33 1993
Received: from interlock.amoco.com by dkuug.dk with SMTP id AA13706
  (5.65c8/IDA-1.4.4j for <SC22WG5@dkuug.dk>); Mon, 16 Aug 1993 22:10:57 +0200
Received: by interlock.amoco.com id AA11085
  (InterLock SMTP Gateway 1.1 for SC22WG5@dkuug.dk);
  Mon, 16 Aug 1993 15:10:27 -0500
Received: by interlock.amoco.com (Internal Mail Agent-2);
  Mon, 16 Aug 1993 15:10:27 -0500
Received: by interlock.amoco.com (Internal Mail Agent-1);
  Mon, 16 Aug 1993 15:10:27 -0500
Date: Mon, 16 Aug 93 15:11:33 CDT
Message-Id: <9308162011.AA23701@trc.amoco.com>
From: Jerrold L. Wagener <jwagener@amoco.com>
To: SC22WG5@dkuug.dk
Cc: dvasicek@amoco.com
Subject: FORALL
X-Charset: ASCII
X-Char-Esc: 29

To X3J3:

The August issue of SigPlan Notices just came across my desk and it contains 
an article entitled "Compiling Machine-Independent Parallel Programs".  This 
article provides, I think, some interesting background for our upcoming work 
on FORALL.

The setting is Modula-2, extended with a FORALL construct - actually, two 
flavors of FORALL, one of which is (more or less) semantically equivalent to 
the HPF FORALL and one that is even "more parallel" (imagine that!).  
Actually, this FORALL provides general process parallelism, and "data 
parallelism" is cast as merely a special form of process parallelism.

Thus, as I understand it, this FORALL can be used not only with array 
operations, pretty much (if not exactly) as in HPF, but also for the X3H5 
parallel sections functionality.  I'm not advocating this FORALL, of course 
(I haven't had time to completely digest it), but I thought it might be 
useful background for developing the requirements for the next standard.

If you don't have access to SigPlan Notices, the author is Michael Philippsen 
at Universitat Karlsruhe, email address:  phlipp@ira.uka.de
He might have an electronic version he can send you.

Jerry