On Thursday, October 16, 2003, at 10:09 AM, Robert LeBlanc wrote:
> I was under the impression that LAM and MPICH were source compatible,
> but not binary compatiable. I tried to compile a program (MrBayes) and
> got an error that said that it could not locate mpidefs.h and
> mpi_errno.h. In looking around on the Internet, I found that these
> files are part of the MPICH implementation. Is there something that I
> am missing for these files to not be included with LAM?
LAM and MPICH are (for the most part) source compatible. They have a
complete F90 module and we have many pieces of MPI-2 that they do not,
so we aren't source compatible for pieces that the other doesn't use.
But that isn't your problem.
Your problem is that the developers of MrBayes did something incredibly
stupid and included the mpi.h header file from MPICH in the MrBayes
source directory. I'm not sure I can conceive of a reason they did
that, but there may be one. So the compiler is picking that mpi.h up
instead of the LAM mpi.h and life gets miserable from there.
I would see what happens if you remove the mpi.h from the distribution
- the preprocessor should then use the LAM mpi.h and life should be
better. If not, I would recommend talking to the MrBayes developers.
Hope this helps,
Brian
--
Brian Barrett
LAM/MPI developer and all around nice guy
Have a LAM/MPI day: http://www.lam-mpi.org/
|
