mpirun problem

Report or discuss software problems and other woes

Moderators: arango, robertson

Post Reply
Message
Author
antoinemarie

mpirun problem

#1 Unread post by antoinemarie »

Hi,

I want to couple ROMS and SWAN, so I download openmpi and MCT. I compile build.bash with the follow command:


./build.bash -j 2


and I obtain my executable oceanM.

Then I type mpirun -np 2 ./oceanM coupling_test_head > out.log with:


Nthreads(ocean) = 1
Nthreads(waves) = 1
in coupling_test_head.in

and I have:


Coupled Input File name = coupling_test_head.in

MOD_COUPLER - Number assigned processors: 001
not equal to spawned MPI nodes: 001
Coupled Input File name = coupling_test_head.in

MOD_COUPLER - Number assigned processors: 001
not equal to spawned MPI nodes: 001


So I try with


Nthreads(ocean) = 1
Nthreads(waves) = 0


and I obtain: (attach file: out.log), but nothing happens.

I don't know why it doesn't work, so if someone can help me please?

Antoine
Attachments
out.log
(27.32 KiB) Downloaded 236 times

simion1232006
Posts: 60
Joined: Tue Sep 29, 2009 3:50 pm
Location: School of Environment System Engineering,UWA

Re: mpirun problem

#2 Unread post by simion1232006 »

I used to have the similar problem. You might not install the openmpi in the right way. I use mpich2, so not sure whether it is the same.

antoinemarie

Re: mpirun problem

#3 Unread post by antoinemarie »

Yes, here I use openmpi, but before, I have used mpich2, and I have exactly the same error.

I had tried to use mpich2, but when I had configured it, the command '--with-device=ch3:shm' didn't work.
So I had configured without this command. Then I had installed MCT.

With openmpi, I have configured MCT with the following command:

./configure FC=mpif90 CC=mpicc

Is it correct?

mathieu
Posts: 74
Joined: Fri Sep 17, 2004 2:22 pm
Location: Institut Rudjer Boskovic

Re: mpirun problem

#4 Unread post by mathieu »

One possible explanation, that you need to check, is that the mpif90 and mpirun/mpiexec do not come from the same origin.

antoinemarie

Re: mpirun problem

#5 Unread post by antoinemarie »

I have checked, and mpirun and mpif90 come from the same origin. I try to use mpich2, so I configure MCT with:


./configure F90=gfortran FC=mpif90.mpich2 CC=mpicc.mpich2


and now I have the following error:


antoine@antoine-VirtualBox:~/roms/Projects/swan_test_head$ mpirun.mpich2 -np 2 ./oceanM coupling_test_head.in

....

WEST_FSCLAMPED Western edge, free-surface, Clamped condition.
WEST_M2REDUCED Western edge, 2D momentum, reduced-physics condition.
WEST_M3RADIATION Western edge, 3D momentum, radiation condition.
WEST_TGRADIENT Western edge, tracers, gradient condition.
WET_DRY Wetting and drying activated.

Process Information:

Node # 0 (pid= 26691) is active.



SWAN is preparing computation ...

application called MPI_Abort(comm=0x84000002, 4) - process 0
APPLICATION TERMINATED WITH THE EXIT STRING: Hangup (signal 1)


:?: :?: :?:

I don't know why, I think my build.bash (attach file) is correct?
Attachments
build.bash
(16.74 KiB) Downloaded 244 times

simion1232006
Posts: 60
Joined: Tue Sep 29, 2009 3:50 pm
Location: School of Environment System Engineering,UWA

Re: mpirun problem

#6 Unread post by simion1232006 »

Try mpiexec command see whether it works. As if you use shared memory machine and use the gforker manager, the mpirun command might not work.

mathieu
Posts: 74
Joined: Fri Sep 17, 2004 2:22 pm
Location: Institut Rudjer Boskovic

Re: mpirun problem

#7 Unread post by mathieu »

I think the problem is now probably in SWAN. Look after the Errfile* for the problem.

antoinemarie

Re: mpirun problem

#8 Unread post by antoinemarie »

It works!! :D

For the last error, it was just the path of grid file which were wrong.

Thanks a lot Simion and Mathieu.

Antoine

Post Reply