[MITgcm-devel] [MITgcm-cvs] MITgcm/eesupp/inc CVS Commit

[Menemenlis, Dimitris (3248)]

Jean-Michel, thanks for forwarding message to devel list.
This is not high-priority but I wanted to at least mention it for possible discussion.
The background is that yesterday we attempted to run MITgcm on 70K
cores and failed twice because of compiled-in limits.
The first limit was MAX_NO_PROCS, easy to fix, but took a little time
to diagnose as model crashed before error message was written out.
The second, and the reason for model crash, appears to be
in some obscure recess of SGI's mpt routines.

Dimitris Menemenlis

On Feb 10, 2014, at 7:55 PM, Jean-Michel Campin <jmc at ocean.mit.edu> wrote:

> Hi Dimitris,
> 
> Do you expect an answer to this question:
>> Shall we remove this check?
> If so, probably better to post it separatly to mitgcm-devel,
> so that every one can see it (not many developers check
> all the content of all cvs-commit message).
> 
> Cheers,
> Jean-Michel
> 
> On Mon, Feb 10, 2014 at 05:06:32PM -0500, Dimitris Menemenlis wrote:
>> Update of /u/gcmpack/MITgcm/eesupp/inc
>> In directory forge:/tmp/cvs-serv10868
>> 
>> Modified Files:
>> 	EEPARAMS.h 
>> Log Message:
>> Attempting to run on 70K processes on pleiades.
>> Shall we remove this check?
>> It's not doing anything useful right now ...
>> except that we have to remember to increase
>> every time we run on a bigger config.