What implementation of MPI is required?

[wl@eimb_ru]

Dear colleagues,

Could you, please, inform me, which exact version and implementation of MPI is required for the CCSM3.0?

I could not find any details in the documentation and in the forum in the CCSM site.

I have unsuccessfully tried latest versions of LAM/MPI, Open-MPI and MPICH.

All of them were supplied with auto-configuration scripts, those scripts and compilations run without errors.

However, the model didn't compile with LAM/MPI. Unfortunately I cannot send you exact details, because I don't remember exact name of component, which was required for the model and was absent in LAM/MPI.

The compilation of the model was successful with Open MPI and MPICH.
However their mpirun doesn't recognize -pg switch, given by the model run scripts.

I use the 'generic_linux' configuration.
I run the model on the Debian Linux with kernel 2.6.15

uname -a gives:

$ uname -a
Linux chena 2.6.15-1-amd64-k8-smp #2 SMP Thu Jan 19 07:46:58 UTC 2006 x86_64 GNU/Linux

I also have edited model script, as suggested in
http://bb.cgd.ucar.edu/showthread.php?t=116&highlight=generic_linux

If any MPI implementation suits, please, tell me how to configure it properly.

Thank you.
---
Best regards,
Wladimir

 

[wl@eimb_ru]

configure switches were
--with-device=ch_p4 --with-comm=shared

It is also possible to use mpiexec (http://www.osc.edu/~pw/mpiexec/index.php) to run components.