[Bioclusters] mpi library problem - more info

[landman]

Ahh....

On Wed, 25 Jun 2003 14:25:14 +0100, andy law (RI) wrote
> All,
> 
> apologies - more info for my previous posting...
> 
> systems are x86.
> 
> mpich and pgapack were compiled using gcc 2.96. The fortran test 
> code for pgapack was compiled with g77 (0.5.26).
> 
> The problem code is Fortran 90 and is being compiled with the Intel 
> compiler (version 7.1).

There are often many problems with inter-compiler linking.  The intel fortran
has options to use gcc/ld for linking, but they usually have restrictions on
which versions to use.

It is often recommended that you get the matching C compiler (from Intel), and
compile mpich and pgapack with it and the fortran compiler (from Intel).  You
might be able to get this to work, though you would need to dig into the intel
compiler documentation about linking to gcc.  The issue with pgapack (and
anything from netlib for that matter) compiled with the g77 and linked with
ifc may not be solvable.  You will likely need to re-compile the fortran with
a single compiler (Intel).  Luckily, f90 is a superset of f77, so this
"should" not be a problem (knowing full well that it is, as I had spent a
number of weeks at SGI figuring out such problems for customers a few years ago).


--
Joseph Landman, Ph.D
Scalable Informatics LLC,
email: landman@scalableinformatics.com
web  : http://scalableinformatics.com
phone: +1 734 612 4615