© 1986 by British Computer Society
| ||||||||||||||||||||||||||||||||||||||||||||||||||||
Computing Random Fields
1 Computer Laboratory, University of Cambridge, Corn Exchange Street, Cambridge CB2 3QG, UK, 2 Hewlett Packard Laboratories, Filton Road, Stoke Gifford, Bristol BS12 6QZ, UK
A program that moves over a 2- or 3-dimensional mesh performing refinements by moving vertices or introducing new ones is a common enough matter. If the data structure for the mesh persists across the entire calculation its management is simple. However, it may be that the mesh data are needed only transiently, and in this case the maintenance of a (possibly enormous) data structure across the whole calculation seems very space-inefficient. This point takes on extra strength when the mesh is supposed to model some random field, such as a mountain range, cloud formation or meadow plain (generated no doubt by a method related to Mandelbrot's fractal curves) when the only use of the mesh is to make a picture for display. Existing methods for dispensing with the need for the whole mesh across the whole calculation complicate the program considerably. This paper proposes the use of co-routines to control the independent update of all edges depending on a changing vertex. In this way facets are passed out as soon as they are complete and can then be displayed and disposed of when convenient. Only visibility issues dictate how long a piece of data survives.
Received May 1985.
* Now at Hewlett Packard Laboratories, Filton Road, Stoke Gifford, Bristol BS12 6QZ
Computer Laboratory, University of Cambridge, Corn Exchange Street, Cambridge CB2 3QG