Problems with MPI and inlet_test

Discussion of how to use ROMS on different regional and basin scale applications.

Moderators: arango, robertson

Post Reply
Message
Author
mistral
Posts: 23
Joined: Wed Jan 16, 2008 11:17 pm
Location: Privat

Problems with MPI and inlet_test

#1 Unread post by mistral »

Hello everybody

I can't use ROMS very well , but till now I have run in a correct way several Test-case bot idealized and realistic case .
Now I to work whit ROMS and SWAN coupled but for personal need I must set the work on my laptop first (intel dual core processor / Ubuntu 9.04 ).
At present I have compiled MCT and MPI (gfortran compiler) .
I have modified Linux-gfortran.mk editing this code lines :

ifdef USE_MCT

# MCT_INCDIR = /home/my/MCT/mct
# MCT_LIBDIR = /home/my/MCT/mct
# FFLAGS += -I$(MCT_INCDIR)
# LIBS += -L$(MCT_LIBDIR) -lmct -lmpeu

FFLAGS += -I/home/my/MCT/mct -I/home/my/MCT/mct/mpeu
MCT_LIBDIR ?= /home/my/MCT/mct
MPEU_LIBDIR ?= /home/my/MCT/mpeu
LIBS += -L$(MCT_LIBDIR) -lmct -L$(MPEU_LIBDIR) -lmpeu

and the build.sh in this way :

setenv USE_MPI on
setenv USE_MPIF90 o
setenv FORT gfortran

# setenv USE_DEBUG
# setenv USE_ARPACK
setenv USE_LARGE on

Then I have compiled the test case LAKE_SIGNELL , (including # define MPI) .
I have run it using : mpirun -np 2 oceanM ocean_lake_signell.in

in this way run work correctly on 1 core and not on both (2) of then as expected .
What is the problem???
What should I do to start “run” on both cores???
I have then tried to use the TEST_HEAD whit :

mpirun -np 2 oceanM coupling_test_head.in

but it didn't work and I have received this error :

mpirun -np 2 oceanM coupling_test_head.in

Coupled Input File name = coupling_test_head.in

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


would you please help me to solve my problems???
Best Regards

User avatar
kate
Posts: 4088
Joined: Wed Jul 02, 2003 5:29 pm
Location: CFOS/UAF, USA

Re: Problems with MPI and inlet_test

#2 Unread post by kate »

Can you run ROMS alone with MPI on two cores?

mistral
Posts: 23
Joined: Wed Jan 16, 2008 11:17 pm
Location: Privat

Re: Problems with MPI and inlet_test

#3 Unread post by mistral »

i can do it with one core (in lake_signell test case , and in test_head received error write on),
but i can't with two core...is it possibile run it???



There are other ways that do not know??


thanks

debcox
Posts: 15
Joined: Tue Mar 09, 2004 3:21 pm
Location: University of New South Wales, Australia

Re: Problems with MPI and inlet_test

#4 Unread post by debcox »

Perhaps this is explained elsewhere on the ROMS forum, but I'm not sure that it is possible to run a program on 1 core of a dual core processor. How do you know that you are not already using both cores?

Could someone more knowledgeable about these kinds of issues clarify this?

jcwarner
Posts: 1182
Joined: Wed Dec 31, 2003 6:16 pm
Location: USGS, USA

Re: Problems with MPI and inlet_test

#5 Unread post by jcwarner »

I am not sure the exact state of this discussion, but here is some hopefully helpful information.
IF you compile with MPI, you can run the program with 1, 2, 3, or however many processors that you have. But you need to set the number of tiles correctly.
For a simulation using ROMS only, the Number I * Number J tiles needs to = the total number of mpi processors requested.
For a coupled simulation of roms + swan, the number of swan + roms processors needs to = the total number of mpi processors requested. These are set in the coupling.in file.

Post Reply