[MPICH] Confused about compiling/using mpich2

[Gaetano Bellanca]

Hi,
sorry  for the delay in answering. I made some tests.

I'm using 'module' to set the environment. Therefore, I'm rather sure 
that when I'm using mpich, everything of mpich2 is not used and 
vice-versa (I have incompatibilities declared in the module loading 
procedures).

But it is true what you said! When I run the code compiled with 
mpich2 with mpirun of mpich (mpirun -n 2 ), the code run but not with 
2 processes. The only way I have to run 2 processes without any error 
is compiling with mpich1.2.7p1 and running it with mpirun of this 
version. In this case, I have 2 processes running!

Both mpirun and mpiexec of mpich2 causes the code to die when trying 
to do a simple MPI_SEND and MPI_RECV.
Then, I changed my way of using mpd. Before I used a ring started by 
root at boot. Now, using mpd from my user, I don't have a crash of 
the code, but it waits undefinitely during the MPI_SEND and MPI_RECV 
procedures.

Maybe I'm wrong in something relevant to the use of mpich2 on a 
single machine with dual core. There is something that I have not understood.

Unfortunately, I don't have my cluster (10 single core single 
processors) available at this moment. I'm running on a single machine 
with dual core athlon (ubuntu linux 7.10).
I cannot say anything about a simulation on a real cluster, for the moment.

Regards.

Gaetano



At 22.08 08/01/2008, you wrote:
>If your code is compiled with MPICH2, it shouldn't run with 
>MPICH-1's mpirun. Make sure that all your code is compiled with the 
>compile scripts from MPICH2 and there are no mpif.h files from 
>MPICH-1 in any of your source directories.
>
>Rajeev
>
>
>----------
>From: owner-mpich-discuss at mcs.anl.gov 
>[mailto:owner-mpich-discuss at mcs.anl.gov] On Behalf Of Gaetano Bellanca
>Sent: Monday, January 07, 2008 12:16 PM
>To: mpich-discuss at mcs.anl.gov
>Subject: Re: [MPICH] Confused about compiling/using mpich2
>
>Dear Darius,
>
>thank you for your answer. But I'm not sure this is the problem. In 
>fact, MPI_Finalize is far from the point where the code is stopped 
>by this error, and there are a lot of other routines before the end 
>of the program. Moreover my code works works with MPICH1.2.7p1 
>(without this problem) and also with MPICH2 (if i run it with mpirun 
>of the mpich1.2.7p1 instead of mpiexec or mpirun of MPICH2).
>
>I'm rather confused about that .... maybe I'm doing something wrong 
>in using mpiexec.
>
>Regards.
>
>Gaetano
>
>
>----------
>Gaetano Bellanca - Department of Engineering - University of Ferrara
>Via Saragat, 1 - 44100 - Ferrara - ITALY
>Voice (VoIP):  +39 0532 974809     Fax:  +39 0532 974870
>mailto:gaetano.bellanca at unife.it
>
>----------


----------
Gaetano Bellanca - Department of Engineering - University of Ferrara
Via Saragat, 1 - 44100 - Ferrara - ITALY
Voice (VoIP):  +39 0532 974809     Fax:  +39 0532 974870
mailto:gaetano.bellanca at unife.it

----------
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20080111/c782d045/attachment.htm>