[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [ESPResSo] MPI issue -- *RESOLVED*
From: |
Tristan Bereau |
Subject: |
Re: [ESPResSo] MPI issue -- *RESOLVED* |
Date: |
Wed, 03 Oct 2007 14:51:34 -0400 |
User-agent: |
Thunderbird 2.0.0.6 (X11/20070801) |
Hello,
First of all, I apologize for sending many emails.
For future reference, I was able to solve my problem by simply renaming
my folder /opt/mpich, and reconfiguring ESPResSo.
That way, ESPResSo wasn't able to find it, and used openmpi instead.
I still wonder, though, why ESPResSo kept trying on using mpich instead
of anything else.
Thanks a lot,
Tristan
Ulf Schiller wrote:
> Hi Tristan,
>
> Tristan Bereau wrote:
> > Thanks for your answer.
> > I looked at the config.log (attached). First, it looks like mpirun,
> > mpicc, and mpich are recognized pretty much everywhere in the file.
> > But looking carefully towards the end of the file (~line 1200),
> there is:
> > MPI_FAKE_FALSE=''
> > MPI_FAKE_TRUE='#'
> > which doesn't look right to me at all (unless a '#' sign means a
> 'no'...?)!
> > It's strange, because it did find all the relevant mpi files on my
> computer.
> > I tried reinstalling OpenMPI (1.2.3), but it still does the same
> problem.
>
> Thanks for providing the config.log. It seems indeed that ESPResSo's
> build system does not recognize the MPI environment correctly. As far as
> I know, not many of us have used ESPResSo with OpenMPI so far. On the
> one hand, this means that it may take some time until your problem can
> be fixed, on the other hand your reports are of great help to improve
> ESPResSo.
>
> > Is there some trick to solve that?
>
> I'm not an expert in the build system so I can only make a guess: My
> suspicion is that the dependency mode set by configure is not correct:
> CC='mpicc'
> CCDEPMODE='depmode=gcc3'
> Can you try to configure ESPResSo again with
> ./configure CC=mpicc
> and check whether the values of these variables change?
> I had a similar problem recently which could be solved this way. I'm not
> sure whether it'll solve your problem as well, but it's worth a try...
>
> Many regards,
> Ulf
>