> This is basic fortran programming issue, you may want to consult
> some fortran programming book.
>
> A.Chan
It is more an issue with understanding the usual implementations
of the MPI Fortran bindings, namely, having to include mpif.h in
ALL procedures that wish to make use of the MPI variable
On Jun 22, 2010, at 6:46 PM, Mihaly Mezei wrote:
> I just tried adding the include statement, but this is what I got:
>
> lapis.mssm.edu % mpif90 -o mpi0 mpi0.f
> /share/apps/openmpi-1.4.1/include/mpif-common.h(402): error #6406:
> Conflicting attributes or multiple declaration of name. [MPI_
I don't think you can declare a function with a fortran parameter:
subroutine testsubr(MPI_COMM_WORLD,ireadok)
1) If you've already included mpif.h with in testsubr(),
you don't need the 1st argument above.
2) If you don't have mpif.h in testsubr(), the 1st argument
could be MPI_comm. In
> I think the problem is that you didn't include mpif.h in testsubr().
> Hence, the value of MPI_INTEGER was undefined -- I don't think it's a
> problem with the value of MPI_Comm.
That's correct.
You also don't then need to pass MPI_Comm_World around, it is a
parameter defined in mpif-common.h w
HEUREEKA!
The solution that works is to keep using MPI_WORK_COMM in the MPI_ subroutine
calls, include mpif.h in the subroutine, but don't try to pass it as an
argument or save it in an other integer (as I did under mpich).
Thanks, Jeff, for your help and patience with my stumblings.
Mihaly Me
- Original Message -
From: Jeff Squyres
List-Post: users@lists.open-mpi.org
Date: Tuesday, June 22, 2010 6:36 pm
Subject: Re: [OMPI users] Fortran - MPI_WORLD_COMM - correction
> I think the problem is that you didn't include mpif.h in
> testsubr(). Hence, the value of MPI_INTEGER wa
Hi,
The previous example I posted shows that using a different integer does not
work, Attached is a case where I tired to pass the MPI_COMM_WORLD to the
subroutine as one of it arguments, resulting in an invalid datatype error.
Mihaly Mezei
Department of Structural and Chemical Biology, Mount
I think the problem is that you didn't include mpif.h in testsubr(). Hence,
the value of MPI_INTEGER was undefined -- I don't think it's a problem with the
value of MPI_Comm.
On Jun 22, 2010, at 6:21 PM, Mihaly Mezei wrote:
> Hi,
>
> attached is mpi0.f (23 lines) and the screen dump of the c
MPI_COMM_WORLD is defined by a parameter statement, so it
is a fortran contant. The following f77 program fails to compile.
> cat ts_param.f
Program test
integer mm
parameter (mm = 9)
common /cmblk/ mm
end
> gfortran ts_param.f
ts_param.f:4.23:
common /cmblk/
Hi,
attached is mpi0.f (23 lines) and the screen dump of the compilation with the
error message.
Thanks for the lightening-speed reaction!
Best,
Mihaly Mezei
Department of Structural and Chemical Biology, Mount Sinai School of Medicine
Voice: (212) 659-5475 Fax: (212) 849-2456
WWW (MSSM hom
On Jun 22, 2010, at 6:04 PM, Mihaly Mezei wrote:
> In mpich, I could assign MPI_WORLD_COMM (i.e., mpich did not accept
> MPI_WORLD_COMM in a common block either) to an integer and that was what I
> put into the common block,. Openmpi does not accept that new integer as a
> communicator.
It sho
On Jun 22, 2010, at 5:55 PM, Mihaly Mezei wrote:
> mpi0.f(2): error #6414: This PARAMETER constant name is invalid in this
> context. [MPI_COMM_WORLD]
> common /MPI_DM/ MPI_COMM_WORLD,NUMNOD,MYRANK
> --^
> mpi0.f(21): error #6412: A dummy argument name is invalid in th
Greetings, and apologies:
In mpich, I could assign MPI_WORLD_COMM (i.e., mpich did not accept
MPI_WORLD_COMM in a common block either) to an integer and that was what I put
into the common block,. Openmpi does not accept that new integer as a
communicator.
Mihaly Mezei
Department of Structura
Greetings,
I used mpich earlier, where I could get the value of MPI_WORLD_COMM in the main
program and pass it to subroutines via a common block (as the MPI_WORLD_COMM is
assumed to be an integer and I even was able to confirm that it is declared to
be an integer in mpif-common.h). However, ope
I have compute-nodes with twin eth interfaces 1GigE and 10GigE. In the
OpenMPI docs I found an instruction:
" It is therefore very important that if active ports on the same host
are on physically separate fabrics, they must have different subnet
IDs."
Is this the same "subnet" that is set via an
Sorry, just found out about the "--debug-daemons" option, which
allowed me to google a meaningful error message and find the solution
in the archives of this list.
For the record, the problem was that the "orted" being launched on the
remote node is the one from the system-wide MPI install, not th
Hello,
On Tue, Jun 22, 2010 at 8:05 AM, Ralph Castain wrote:
> Sorry for the problem - the issue is a bug in the handling of the
>pernode option in 1.4.2. This has been fixed and awaits release in
>1.4.3.
>
Thank you for pointing this out. Unfortunately, I still am not able
to start remote proc
On Jun 22, 2010, at 10:57 AM, Michele De Stefano wrote:
> Anyway, I've discovered that configure searches for
> and does not find it.
>
> Installing valgrind-devel solves the problem ... but I didn't need it
> before (I mean I didn't need it before Open MPI 1.4.2).
Yes, you did. :-)
OMPI alwa
I tried to send the config.log ... but it seems to be too big, and the
message bouces back.
Anyway, I've discovered that configure searches for
and does not find it.
Installing valgrind-devel solves the problem ... but I didn't need it
before (I mean I didn't need it before Open MPI 1.4.2).
Bes
Did you mean to send this to the openfoam mailing list?
We don't know much (anything) about the internals and usage of openfoam --
you'll need to contact them for help. This list is for Open MPI and general
MPI support only.
Sorry!
On Jun 22, 2010, at 9:43 AM, wrote:
> hello, i'm new with
hello, i'm new with openfoam and i have a problem with a mpirun, when a type in the terminal : mpirun -np 30 -machinefile machinefile turbFoam -parallel | tee 2>&1 log/turb.logit runs, after a while it stops with this message : Create time
What is the output of ompi_info | grep memchecker?
Can you also send the stdout from OMPI's configure where it is checking for
valgrind compatibility? It should be fairly obvious, probably deep in the
middle of the output of configure. Here's what mine looks like:
+++ Configuring MCA framewor
Is gdb installed too ? I might be wrong, but I encountered a similar issue when
building 1.4.2 and this was related to valgrind and gdb (and their respective
libraries) not being available on the building and running host.
E.
On Tuesday 22 June 2010 11:14:02 Michele De Stefano wrote:
> valgrind
On Jun 21, 2010, at 3:01 AM,
wrote:
> i have done a "mpirun" on 30 processeur then i have done a "reconstructPar",
> now i have modified my controlDict file with a startime from the latestTime
> and the endTime 1000 intervalls plus than the last , i want to know if i need
> to redone a decom
valgrind is installed, and worked with Open MPI 1.4.1.
2010/6/22 Eloi Gaudry :
> Hi Michele,
>
> You may actually need to have gdb/valgrind installed before configuring and
> building OpenMPI with the --enable-memchecker option.
>
> Regards,
> Eloi
>
>
> On Tuesday 22 June 2010 10:44:16 Michele D
Hi Michele,
You may actually need to have gdb/valgrind installed before configuring and
building OpenMPI with the --enable-memchecker option.
Regards,
Eloi
On Tuesday 22 June 2010 10:44:16 Michele De Stefano wrote:
> Good morning.
>
> I'm experiencing some problems with Open MPI 1.4.2, built
Good morning.
I'm experiencing some problems with Open MPI 1.4.2, built on a CentOS
5.5, on a 32 bit machine.
I've configured it with
./configure --prefix=/usr/local/openmpi --enable-debug
--enable-memchecker --enable-heterogeneous
When I run "/usr/local/openmpi/bin/mpiexec -np 3 ls", Open MPI
Had to go back into the deep, dark archives and look at the source code to
address this. :-)
Your understanding is correct - ompi should see those envars and automatically
use xgrid.
I don't see anything in the source code beyond those variables described in the
reference you provided. You cou
It is in the latest 1.4.3 nightly tarball now.
On Jun 18, 2010, at 12:33 PM, Jeff Squyres wrote:
> On Jun 18, 2010, at 9:57 AM, David Turner wrote:
>
>> I downloaded openmpi-1.4.3a1r23261.tar.gz, dated June 9. It behaves the
>> same as 1.4.2. Is there a newer version available for testing?
>
Sorry for the problem - the issue is a bug in the handling of the pernode
option in 1.4.2. This has been fixed and awaits release in 1.4.3.
On Jun 21, 2010, at 5:27 PM, Riccardo Murri wrote:
> Hello,
>
> I'm using OpenMPI 1.4.2 on a Rocks 5.2 cluster. I compiled it on my
> own to have a threa
30 matches
Mail list logo
