[Pw_forum] OpenMPI problems with espresso 4.0cvs
Riccardo Sabatini
sabatini at sissa.it
Fri Apr 25 16:52:41 CEST 2008
Hi,
made some trials but nothing still works :(( I specified the
machine both with --host both with a machine file, nothing changes and
i tried to run it on just one core
mpirun -np 1 ../QE-MPI/pw.x < si.scf.cg.in
In this case the program doesn't even start, printing this output
----------------------------------------------------------------------
Program PWSCF v.4.0cvs starts ...
Today is 25Apr2008 at 14:37:21
Ultrasoft (Vanderbilt) Pseudopotentials and PAW
Current dimensions of program pwscf are:
Max number of different atomic species (ntypx) = 10
Max number of k-points (npk) = 40000
Max angular momentum in pseudopotentials (lmaxx) = 3
----------------------------------------------------------------------
I'll write down all the compilation iformation i have, if anyone
wonna try to see if it's a bug in the espresso 4 snapshot.
----------------from make.sys
DFLAGS = -D__GFORTRAN -D__FFTW -D__USE_INTERNAL_FFTW
FDFLAGS = $(DFLAGS)
IFLAGS = -I../include
MODFLAGS = -I./ -I../Modules -I../iotk/src \
-I../PW -I../PH
MPIF90 = mpif90
#F90 = gfortran
CC = cc
F77 = gfortran
CPP = cpp
CPPFLAGS = -P -traditional $(DFLAGS) $(IFLAGS)
CFLAGS = -O3 $(DFLAGS) $(IFLAGS)
F90FLAGS = $(FFLAGS) -x f95-cpp-input $(FDFLAGS) $(IFLAGS) $(MODFLAGS)
FFLAGS = -O3
-----------------------------------------------------------------------------
I use the following compilers,
GNU Fortran (GCC) 4.2.1 (Ubuntu 4.2.1-5ubuntu4)
cc (GCC) 4.1.3 20070929 (prerelease) (Ubuntu 4.1.2-16ubuntu2)
cpp (GCC) 4.1.3 20070929 (prerelease) (Ubuntu 4.1.2-16ubuntu2)
OpenMPI 1.1-2.5 ubuntu1
Will see if someone can reproduce the problem, in any case thanks
again for the help,
Riccardo
Quoting "O. Baris Malcioglu" <baris.malcioglu at gmail.com>:
> Hello again,
>
> Maybe you should try something like
>
> mpirun -np 4 --host localhost
>
> or a machine file explicitly stating you are working on the localhost,
>
> the reason why you are getting 3 errors but not 4 can be related to
> openmpi spawning jobs at other machines.
>
> You can try running the code you have compiled with mpi serially to
> see whether the compilation is broken or not.
>
> Hope this helps,
> Baris Malcioglu
> SISSA, Italy
>
----------------------------------------------------------------
SISSA Webmail https://webmail.sissa.it/
Powered by Horde http://www.horde.org/
More information about the users
mailing list