[gmx-users] SGI Altix - mdrun_mpi

[Nuno Loureiro Ferreira]

Hi guys

Just installed gmx on a SGI Altix system (64-way):

-- compiling mdrun with MPI support, single-precision
export CPPFLAGS=-I/home/nferreira/bin/fftw-3.0.1/include
export LDFLAGS=-L/home/nferreira/bin/fftw-3.0.1/lib
./configure --prefix=/home/nferreira/bin/gromacs-3.3.2 --enable-mpi 
--program-suffix=_mpi

make mdrun
make install-mdrun

-- compiling all gromacs package programs without MPI, single-precision
make distclean
./configure --prefix=/home/nferreira/bin/gromacs-3.3.2
make
make install



It runs fine on single nodes, but I can't put gmx running in parallel.
The jobs are submitted directly (see a sample script bellow), without 
any queuing handler, like PBS.




# This is my submission script
###
lamboot -v hostfile
lamnodes
# Testing mpi
mpirun -np 2 hello
# GMX run
grompp -np 2  -f equil -po equil_out -c after_em -p topology -o equil
mpirun -np 2 mdrun_mpi -np 2 -deffnm equil -c after_equil
lamhalt



# This is hostfile file
###
# My boot schema
localhost cpu=64


I tryed several stuff (full paths for mpirun, mdrun_mpi, etc), but I'm 
always getting the same error. I also tested a hello program (from 
LAM-MPI user guide) and it gives no problems. Bellow is the output of 
the submission script:


[EMAIL PROTECTED]:~/gmxbench ./script

LAM 7.1.3/MPI 2 C++/ROMIO - Indiana University

n-116608 ssi:boot:base:linear: booting n0 (localhost)
n-116608 ssi:boot:base:linear: finished
n0  localhost:64:origin,this_node
Hello, world! I am 0 of 2
Hello, world! I am 1 of 2
NNODES=2, MYRANK=1, HOSTNAME=behemoth
NNODES=2, MYRANK=0, HOSTNAME=behemoth
NODEID=0 argc=8
NODEID=1 argc=8
:-)  G  R  O  M  A  C  S  (-:

  GRoups of Organic Molecules in ACtion for Science

   :-)  VERSION 3.3.2  (-:
[ ... snipped ...]

Getting Loaded...
Reading file equil.tpr, VERSION 3.3.2 (single precision)
-
One of the processes started by mpirun has exited with a nonzero exit
code.  This typically indicates that the process finished in error.
If your process did not finish in error, be sure to include a return
0 or exit(0) in your C code before exiting the application.

PID 16623 failed on node n0 (127.0.0.1) due to signal 11.
-

LAM 7.1.3/MPI 2 C++/ROMIO - Indiana University



Any ideas? Searching the mailing list, seems that this is a recurrent  
issue, but I was not able to find and answer.

And, the machine admin is not proficient in MPI.

Cheers,
Nuno

P.S. Other programs are running fine on this machine using the same LAM-MPI.

___
gmx-users mailing listgmx-users@gromacs.org
http://www.gromacs.org/mailman/listinfo/gmx-users
Please search the archive at http://www.gromacs.org/search before posting!
Please don't post (un)subscribe requests to the list. Use the 
www interface or send it to [EMAIL PROTECTED]

Can't post? Read http://www.gromacs.org/mailing_lists/users.php

[gmx-users] Re: SGI Altix - mdrun_mpi

[Nuno Loureiro Ferreira]

Hi Chris, again ;-)

I'm also running gromacs on another Cluster intel based ( using PBS), 
and my script submission procedure is similar to yours!


Now, this new machine I'm trying to use is a SGI 64-way, thus I think 
that the lam nodes file is correct: 64 cpu's in a single node, or 
something like that. And, as you can see, the hello program does work 
on this machine when called from mpirun.


As I told in my previous e-mail, there are people using the LAM-MPI on 
this machine, and in fact, the hostfile and part of the script 
submission file were written by my colleagues. Thus, I think that the 
hostfile is correct, and that probably something is not ok with the 
submission file,  the gmx compilation (though the compilation was 
successful) or even with the system configuration.


I tried also the suggestions in your script (exporting PATH, 
LD_LIBRARY_PATH, LAMRSH), but getting exactly the same error.

Oh well ...

Thanks for the reply,
Nuno



___
gmx-users mailing listgmx-users@gromacs.org
http://www.gromacs.org/mailman/listinfo/gmx-users
Please search the archive at http://www.gromacs.org/search before posting!
Please don't post (un)subscribe requests to the list. Use the 
www interface or send it to [EMAIL PROTECTED]

Can't post? Read http://www.gromacs.org/mailing_lists/users.php