Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!gatech!hubcap!klaatu.rutgers.edu!josh
From: josh@klaatu.rutgers.edu (J Storrs Hall)
Newsgroups: comp.parallel
Subject: Re: any mesh algorithms use end-around connections?
Message-ID: <5545@hubcap.clemson.edu>
Date: 19 May 89 19:29:46 GMT
References: <5528@hubcap.clemson.edu>
Sender: fpst@hubcap.clemson.edu
Organization: Rutgers Univ., New Brunswick, N.J.
Lines: 21
Approved: parallel@hubcap.clemson.edu
To: comp-parallel@rutgers.edu



    Recently I discussed 2-D and 3-D mesh architectures with a professor of a
    class in advanced computer architecture that I am taking.  Prof says that
    end-around connections (outer edge to outer edge, face plane to face plane)
    aren't used...
    Dan'l Levy                 UNIX(R) mail:  att!ttbcad!levy, att!cbnewsc!danl

Assuming we're talking about the same thing, we use these connections 
on the DAP regularly.  To be specific:  consider one row or column of
the mesh as a shift register; the connections which allow you to do a 
recirculating shift.

If you have a mesh-shaped problem bigger than the machine, there are
two major ways to cut it up: 
(P[i,j]=processor, D[i,j]=datum, kxk processors, nxn data)
A: P[i%k,j%k] gets D[i,j]
B: P[i/k,j/k] gets D[i,j]

The DAP's hardware mapping to the frame buffer "encourages" you to use
A. In this case the end-around connections are extremely useful.

--JoSH