home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.parallel
- Path: sparky!uunet!gatech!hubcap!fpst
- From: maniattb@cs.rpi.edu (Bill Maniatty)
- Subject: Re: Torus vs. Hypercube
- Message-ID: <1992Dec30.124438.28257@hubcap.clemson.edu>
- Apparently-To: comp-parallel@cis.ohio-state.edu
- Sender: fpst@hubcap.clemson.edu (Steve Stevenson)
- Nntp-Posting-Host: electra.cs.rpi.edu
- Organization: Clemson University
- References: <1992Dec27.201910.28352@ee.eng.ohio-state.edu> <1992Dec29.210648.18508@hubcap.clemson.edu>
- Date: Tue, 29 Dec 1992 22:08:12 GMT
- Approved: parallel@hubcap.clemson.edu
- Lines: 59
-
- In article <1992Dec29.210648.18508@hubcap.clemson.edu>, heller@Think.COM (Steve Heller) writes:
- |>
- |> In article <1992Dec29.142815.27527@hubcap.clemson.edu> David Karr <karr@cs.cornell.EDU> writes:
- |>
- |> In article <1992Dec27.201910.28352@ee.eng.ohio-state.edu> bai@rainier.eng.ohio-state.edu (Baback A. Izadi) writes:
- |>
- |> >Couple of weeks ago, I think heard a speaker claiming that
- |> >2-dimensional Torus and hypercube have basically the same
- |> >topology. How could this be true?
- |>
- [Stuff Deleted]
-
- The topology of the connections imbedded in the torus might make a 2
- dimensional torus more connected than the corresponding hypercube.
-
- Consider an 8 nearest neighbor connectivity with toroidal wrap, with O being
- the central processor, the others labelled after the points on a compass:
-
- NW N NE
- \|/
- W-O-E
- /|\
- SW S SE
-
- (a real world example of this is found in a MasPar MP-1),
- which is more connected than the corresponding ``Manhattan'' style connection:
-
- N
- |
- W-O-E
- |
- S
-
- If we look at a simple 2 dimensional hypercube (square)
- versus the corresponding 8 nearest neighbor topology with toroidal wrap
- the hypercube requires 2 hops to go to opposing corners, while every node
- in the 8 nearest neighbor torus can be reached in just one hop.
-
- [Stuff Deleted]
-
- |> The fact that the 4^2 torus is isomorphic to the binary 4-cube led to an
- |> interesting question here at TMC about six months ago.
- |>
- |> Is the 4^k torus isomorphic to the binary 2k-cube?
- |>
- |> The answer and the proof (which I found quite fun) is left as an excercise
- |> to the reader.
- |>
- |> -steve
-
- Is this a closely guarded corporate secret :-) ?
-
- Bill
-
- --
- |
- | maniattb@cs.rpi.edu - in real life Bill Maniatty
- |
-
-