[MITgcm-devel] Re: [MITgcm-support] MNC_MAX_ID too small for 24 tiles

[Ed Hill]

On Sat, 2004-12-11 at 12:25 -0500, Chris Hill wrote: 
> Hi Ed,
> 
>  Does MNC memory/id range etc... scale  in proporation to number of 
> tiles, it probably should.
>  We have 2000+ tiles in some configurations these days.


Hi Chris,

Currently, MNC memory use (the lookup tables) scales approximately as:

  MEM ~ (# of NetCDF vars)*(# of concurrently open files)
  MEM ~ (# of NetCDF vars)*(# of tiles)*(files/tile)

So with MPI, there shouldn't be an excessive number of files per process
and thus the scaling should be good.

If, however, someone runs a job using many tiles within a single process
(whether they are using just one process or they are doing MPI and
choosing to have many tiles on one or more of those processes) then its
a good idea to increase MNC_MAX_ID.

I just added a few paragraphs and a table describing this in the manual.

Ed

PS - We could do something like:

  MNC_MAX_ID = factor*(nSx*nSy)

but then we'd still have to adjust the "factor" based on some run-time
things.


-- 
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