[MITgcm-support] Problem for the MITgcm parallelized compilation

Yuan Lian lian at ashimaresearch.com
Wed Apr 10 06:55:14 EDT 2013

Hi Francois,

A quick answer to your questions is that these runtime errors are not compilation errors.

Both errors mean you didn't set up the grid correctly. Simply changing npx and npy won't work. You need to make sure the total number of grid points is not changed, assuming you didn't change data file or topog.slope. It will be helpful to provide your SIZE.h file and data file so people can answer your questions better.


François Lefeuvre <francois.lefeuvre at gmail.com> wrote:

Dear all,

I am a M2 student which is doing an internship. I need to use the MITgcm code in order to solve my problem. Actually, for now, I am testing the verification cases which are given in the .tar file that I have downloaded. So, I chose two examples and I managed to run both. But now, I want to parallelize the computation because it will be needed for my problem. Thus, I changed the nPx/nPy number in the SIZE.h file. Then, I checked the option file that I have to put when I compile the code with genmake2 (I read the HOW TO manual and changed a little bit one of the files which has been made for the most similar machine that is to say I changed directories of MPIrun libraries and includes), and followed the instruction of the Building with MPI part. Moreover, I run these two codes locally, so I don't need any script. The result is that I manage to compile the code but when I want to execute it, it doesn't work.

The last command I launch and the message error I get for the intenal_waves example is :
    lefeuvre at sheena:~/MITgcm_c64f/verification/internal_wave/build> mpirun -np 2 ./mitgcmuv > output.txt
    At line 2128 of file mdsio_read_field.f (unit = 9, file = 'topog.slope')
    Fortran runtime error: Non-existing record number

The last command I launch and the message error I get for the tutorial_baroclinic_gyre example is :
    lefeuvre at sheena:~/MITgcm_c64f/verification/tutorial_baroclinic_gyre/build> mpirun -np 2 ./mitgcmuv > output.txt

I don't know in which way I need to look for errors. According to you, Where does the problem may come from?
- The parallelization
- Some compiler problem
- The input/data file
- something else
Are these two errors linked?

Sorry for my english and thank you for help, sincerely François Lefeuvre.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20130410/81ae18d5/attachment.htm>

More information about the MITgcm-support mailing list