Next: Parallelization strategy (distributed memory)
Up: Introduction: mesh computations
Previous: Introduction: mesh computations
In a mesh computation:
- Data is based on an N-dimensional grid
(N = 1, 2, or 3), with
one or more variables per cell (grid point).
- Computation consists of some sequence of the
following operations:
- computing, for each cell, new values for
one or more variables, based on old
values of variables in that cell and
nearby cells (neighbors, next-to-neighbors,
etc.)
- (optionally) reading in values for variables.
- (optionally) writing out values of variables.
- (optionally) computing global reductions over
the whole grid -- e.g., global maximum or
sum of one of the variables.
(Frequently the compute-new-values and reduction operations
are performed repeatedly in a time-step loop.)
Figure 1 illustrates the basic operation -- computing
new values in terms of old values -- in a 2-dimensional grid.
Figure 1: Basic mesh computation.
Berna L Massingill
Mon Jun 8 19:35:58 PDT 1998