[MPICH] running an MPI prog

[Rajeev Thakur]

Yes, you need to use the exact same MPI implementation for compiling and
running. 

Your cpi example also shows the same problem. Each process thinks it is in
own world, not a world of 3 processes.

Rajeev 

> -----Original Message-----
> From: Mark.Donohue at avtechgroup.com 
> [mailto:Mark.Donohue at avtechgroup.com] 
> Sent: Friday, July 27, 2007 11:07 AM
> To: thakur at mcs.anl.gov
> Cc: mpich-discuss at mcs.anl.gov
> Subject: RE: [MPICH] running an MPI prog
> 
> Hello Rajeev.  Good question as I didn't compile the MPI 
> program myself.
> But you're probably onto something as I am using MPICH2-1.0.5 
> and I have
> a strong hunch the code was compiled with MPICH1.  That's a 
> show-stopper
> huh?  Would your suggestion be to install the same version of MPICH1
> that compiled the code?  Thanks again!
> -mark
> 
> -----Original Message-----
> From: Rajeev Thakur [mailto:thakur at mcs.anl.gov] 
> Sent: Thursday, July 26, 2007 9:55 PM
> To: Mark Donohue
> Cc: mpich-discuss at mcs.anl.gov
> Subject: RE: [MPICH] running an MPI prog
> 
> Make sure you are compiling and running your program with mpif90 and
> mpiexec from the same MPI implementation. Give the full path 
> to mpiexec
> if needed.
> 
> Rajeev 
> 
> > -----Original Message-----
> > From: owner-mpich-discuss at mcs.anl.gov 
> > [mailto:owner-mpich-discuss at mcs.anl.gov] On Behalf Of Anthony Chan
> > Sent: Thursday, July 26, 2007 7:29 PM
> > To: Mark.Donohue at avtechgroup.com
> > Cc: mpich-discuss at mcs.anl.gov
> > Subject: Re: [MPICH] running an MPI prog
> > 
> > 
> > Can you run a simple MPI program like cpi in 
> > <mpich2-build-dir>/examples ?
> > 
> > cd <mpich2-build-dir>/examples
> > <mpich2-build-dir>/bin/mpiexec -n 4 cpi
> > 
> > A.Chan
> > 
> > On Thu, 26 Jul 2007 Mark.Donohue at avtechgroup.com wrote:
> > 
> > > Hi yall.  Please help a simpleton with some basic MPICH
> > knowledge.  I am
> > > trying to run an MPI version of a CFD program that I use
> > regularly.  I
> > > installed MPICH2 on our Linux cluster, but when I try and
> > run, I get the
> > > error at the bottom.  I split my computational domain into two 
> > > partitions, and the program complains that the number of
> > partitions (2)
> > > doesn't equal the number of processors.  Huh?  Is there a 
> way to fix
> 
> > > this?  I realize that this error is coming from my MPI code
> > and not from
> > > MPICH itself, but I was hoping to gain some insight into 
> why this is
> 
> > > happening.  For some reason I thought specifying "-n 2"
> > meant that it
> > > will begin running with two processors.  All the 
> documentation I've 
> > > found hasn't been able to answer this for me.  Thanks a bunch!
> > > -Mark
> > >
> > >
> > >
> > > Aero1% mpiexec -l -machinefile myhosts -n 2 usm3d_52p.mpip4 
> > > Baseline_Elev
> > > 0: ** WARNING: fp stack is not empty
> > > 0: ** after call to etime in routine usm52p
> > > 0:
> > > 0: ** popping the fp stack and continuing...
> > > 1: ** WARNING: fp stack is not empty
> > > 1: ** after call to etime in routine usm52p
> > > 1:
> > > 1: ** popping the fp stack and continuing...
> > > 1: FORTRAN STOP
> > > 1:  the MPI has been initialized on            1  processors
> > > 1:   no. of partitions != no. of processors
> > > 0: FORTRAN STOP
> > > 0:  the MPI has been initialized on            1  processors
> > > 0:   no. of partitions != no. of processors
> > >
> > >
> > 
> > 
> 
>