In Evaluation of Multicomputers for Image Processing,
L. Uhr et al., eds, Academic Press, 1986, pp. 107-121
An Algorithmic Comparison of Meshes and Pyramids
Pyramid computers contain mesh computers, so for any problem pyramids should
never run slower than meshes,
and they have the potential of running much faster.
This would seem to indicate that pyramids should always be prefered, but it
ignores the fact that pyramids are more complicated and expensive to build.
It also ignores the facts that pyramids cannot always achieve their potential
for logarithmic time algorithms, and that in general it is much more difficult
to develop good pyramid algorithms.
These latter two facts are the focus of this paper, which begins an analysis of
the relative merits
of meshes versus pyramids by considering their optimal speed on a few tasks,
along with the difficulty of developing optimal algorithms.
Quentin F. Stout
Computer Science and Engineering, University of Michigan
Keywords: parallel computer, mesh-connected, pyramid,
image processing, convexity,
component labeling, divide-and-conquer, computer science
Other papers in parallel computing
Overview of work on parallel computing
Copyright © 2005-2016 Quentin F. Stout