Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!samsung!emory!hubcap!gene
From: gene@CS.BU.EDU (Gene Itkis)
Newsgroups: comp.parallel
Subject: Re: scalability of n-cubes, meshes (was: IPSC Communications)
Message-ID: <7377@hubcap.clemson.edu>
Date: 8 Dec 89 14:25:44 GMT
Sender: fpst@hubcap.clemson.edu
Lines: 18
Approved: parallel@hubcap.clemson.edu

In article <7373@hubcap.clemson.edu> bjornl@tds.kth.se (Bj|rn Lisper) writes:
>In article <7274@hubcap.clemson.edu> gene@cs.bu.edu (Gene Itkis) writes:

>... So, if you care about
>>scalability, the 3D meshes are no good - they'll overheat, cook themselves.
>>Thus the best you can hope for in our physics is 2D mesh.
>
>You implicitly assume that the 3D mesh is a solid body. This is not
>necessarily true. It could for instance consist of a stack of cards, with
>air flowing between them, or have channels with coolant flowing through. In
>this way, you get the same cooling area as for the 2D mesh.

Yes, provided that the distance between the "cards" is not constant, ie growing
with the number of automata.  This architecture would still remain effectively
2D. If you are to pack automata densely (leaving a constant volume around every
automaton) you eventually run into the overheating problem even if you run
coolant between automata. Think of water being "generated" instead of heat. Do
not forget, we want scalability.