[MITgcm-devel] mnc and "global" files

[Ed Hill]

On Wed, 2005-09-07 at 22:21 -0400, Baylor Fox-Kemper wrote:
> Wait,
>    The model will be outputting per-processor pickups.  How will those 
> be labeled?
>    -Baylor

Hi Baylor,

No, please, there is no "per-processor" anything.  Everything is handled
on either a per-tile or per-face basis.  And, at least initially, the
model will:

 - gracefully (automatically) READ either per-face or per-tile
   files using the following approach:

   for each tile:
   - first look to see if a per-face file exists for the current 
     tile and, if it does, read the info needed within that face
   - if the per-face file does not exist then look for a per-tile 
     file and use it or, if it doesn't, exit with a no-data-
     available error

 - only WRITE per-tile files.

And I'll write a script (already have one!) that takes all the per-tile
files and assembles them into per-face files so that one can run the
same domain (the same physical grid) using two different tile
decompositions.  So you'll need to do a tiles-to-faces assembly ONLY if
you are changing the tile decomposition.

And if you change the number of processors while keeping the tile
decomposition constant, then *nothing* needs to be changed.  Thats why
we never make any of the processor stuff explicit.

Ed

-- 
Edward H. Hill III, PhD
office:  MIT Dept. of EAPS;  Rm 54-1424;  77 Massachusetts Ave.
             Cambridge, MA 02139-4307
emails:  eh3 at mit.edu                ed at eh3.com
URLs:    http://web.mit.edu/eh3/    http://eh3.com/
phone:   617-253-0098
fax:     617-253-4464