Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!elroy.jpl.nasa.gov!ncar!gatech!hubcap!fpst
From: sv001!news@UCSD.EDU (Kevin Sanders)
Newsgroups: comp.parallel
Subject: Re: Mapping DO Loops onto Hypercubes
Message-ID: <1991May22.122211.19994@hubcap.clemson.edu>
Date: 22 May 91 01:31:11 GMT
References: <1991May20.124350.27749@hubcap.clemson.edu>
Sender: fpst@hubcap.clemson.edu (Steve Stevenson)
Organization: NCR/Teradata, San Diego, CA
Lines: 19
Approved: parallel@hubcap.clemson.edu
Apparently-To: ucsd!comp-parallel

In article <1991May20.124350.27749@hubcap.clemson.edu> ouyang@TINMAN.CS.NYU.EDU (Pei OuYang) writes:
>
>I am interested in mapping Fortran nested DO loops onto hypercube computers,
>where a nested DO loop can be modeled by an iteration space and a set of
>dependence vectors.
....

A reasonable book on this subject is "Dependence Analysis for Supercomputing",
by Utpal Banerjee, available from Kluwer Academic publishers (1988).

The book demonstrates a method for determining dependence, limited to
DO loops and assignment statements.  It utilizes an integer programming
technique to determine dependencies; this technique is restricted to
DO loops whose indices are used in _linear_ functions only.

   | || |      Kevin Sanders
   |_||_|      NCR/Teradata JDO
  /\V/\V/\     (619) 597-3602
/_//_/\_\\_\   kevin.sanders@sandiego.ncr.com