Data declarations

The archetype defines a PROCESS COMMON block (data common to the process -- each process's block is distinct) for data to be used by its subroutines. These variables are available to the application programmer; useful variables include:

• Process type -- host or grid -- and process indices (for grid processes). (The process type variable is replaced in the no-host-process implementations with a logical variable indicating whether this process is the ``designated I/O process''.)
• Actual size of local data -- may be different from the local-dimension PARAMETER if the number of processes does not evenly divide the size of the grid.

See comments in INCLUDE file mesh_common.h for more information.

The application programmer defines all application data, including grid-based data. Dimensions of grid-based arrays should be declared using the archetype-supplied size parameters mentioned above.

Example

Continuing the example of the preceding section, grid data can be declared thus:

• In the host process, the whole array is declared:

          REAL MYDATA(NX, NY, NZ)

  Observe that this declaration is only needed if array MYDATA is to be read/written from the host process.
• In a grid process, a local section (including ghost boundaries) is declared:

          REAL MYDATA(IXLO:IXHI, IYLO:IYHI, IZLO:IZHI)

  (PARAMETERs IXLO, IXHI, etc. are computed by the archetype implementation from the user-specified grid dimensions, ghost boundary size, and process grid dimensions.)

  Indices for the local section exclusive of ghost boundaries are
  (1:NXlsize, 1:NYlsize, 1:NZlsize).