There are a few other methods that can also be competitive in terms of the communication/computation ratio. One interesting one is the foam method, which takes a bunch of 3D regions out of the half shell (and makes the 3D representation look like a bunch of floating bricks). 1 particular study (KJ Bowers, 2005, Overview of Neutral Territory Methods) found that the foam method is as good as Shaw's neutral territory (shown in the next few slides) as the number of processors increases.
Another interesting one is the eighth shell method, which just takes an eighth of a shell. The same study found that for small numbers of processors, the eighth shell method is better than Shaw's neutral territory.
This comment was marked helpful 0 times.
dfarrow
Here's the PDF if anyone is interested. The different import regions are illustrated in Figure 2!
There are a few other methods that can also be competitive in terms of the communication/computation ratio. One interesting one is the foam method, which takes a bunch of 3D regions out of the half shell (and makes the 3D representation look like a bunch of floating bricks). 1 particular study (KJ Bowers, 2005, Overview of Neutral Territory Methods) found that the foam method is as good as Shaw's neutral territory (shown in the next few slides) as the number of processors increases.
Another interesting one is the eighth shell method, which just takes an eighth of a shell. The same study found that for small numbers of processors, the eighth shell method is better than Shaw's neutral territory.
This comment was marked helpful 0 times.
Here's the PDF if anyone is interested. The different import regions are illustrated in Figure 2!
This comment was marked helpful 0 times.