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

[Chris Hill]

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.

Chris

Ed Hill wrote:

>On Fri, 2004-12-10 at 18:47 -0500, Matthew Robert Mazloff wrote:
>  
>
>>MNC_MAX_ID = 1000, is hard coded into the model.  However, when I break the
>>model into 24 tiles, this value is too small.  Is there any reason why I
>>shouldn't change this value to 1500 so I can run the model with 24 tiles?  Why
>>is there a max number of allowed indices? 
>>    
>>
>
>
>Hi Matt,
>
>Short answer:
>
>Please go ahead and set MNC_MAX_ID to as big a value as you need.  I'd
>increase it by a factor of 10 every time you get an "out of memory"-type
>error.  Unfortunately, there is no easy way of knowing exactly how much
>memory MNC will need since it depends (to a great extent) upon run-time
>variables.
>
>
>Long answer:
>
>While its very portable, optimizes readily, and can be understood by
>essentially all scientists, Fortran77 is *archaic*.  It doesn't (easily)
>allow for dynamic memory allocation.  So the sizes of nearly all the
>important arrays in MITgcm are set at compile-time.  And despite my
>desire to write MNC in C (which supports dynamic memory allocation), the
>group decided to use F77.  So, for now at least, thats what we have.
>
>I'm hoping that MITgcm will evolve so that the core routines no longer
>rely upon Fortran COMMON blocks.  Perhaps some important new feature or
>combination of features will precipitate this change.
>
>Ed
>
>  
>