On Jun 4, 2009, at 1:34 PM, Brent Kraczek wrote:
> I am attempting to compile libmesh on a cluster that uses mvapich and
> icc , but I keep getting a compiler error. I don't know if this is
> something you have seen before or if there is a work-around, short of
> asking Intel.
>
> Specifically, I am using
> /share/apps/mvapich/intel/bin/mpicxx ,
> and the cluster set-up currently only uses static libraries, so I have
> used the --disable-shared flag when running configure for libmesh:
> ./configure --prefix=${HOME}/gmxLocal --enable-threads
> --enable-float=no --enable-sse2 --enable-shared
>
> I changed the configure.in file to choose icc as the C++ compiler,
> so I
> believe everything is consistent.
> autoconf configure.in > configure2
> ./configure2 --enable-everything --enable-second --disable-perflog
> --disable-bzip2 --disable-vsmoother --disable-shared
This is not the right way to do this. Just set your environment up
correctly before running configure. For instance... here's what we
use with intel compilers (that we use with mvapich on our cluster as
well... although we use the 11.0 version of the intel compilers):
export CC=mpicc
export CXX=mpicxx
export FC=mpif90
export F90=mpif90
export F77=mpif77
export libmesh_CFLAGS="-DMPICH_SKIP_MPICXX"
export libmesh_CXXFLAGS="-DMPICH_SKIP_MPICXX
Then make sure that doing a "which mpicxx" points to /share/apps/
mvapich/intel/bin (and if it doesn't... add that path to the beginning
of your PATH).
Now just run configure...
./configure --disable-shared
Don't worry about any other options until you get the default set of
configure flags to work.
As for getting "backend signal"... this is an internal error in the
compiler. This is essentially a "segfault"... and there really isn't
much of a way to figure out why it's happening. Usually I've seen
these whenever libraries have been inconsistently compiled. We just
had a "backend error" show up earlier this week on a new intern's
machine when they were trying to build our software stack... never
really tracked down what the exact offender was... but after wiping it
all away and starting over at the bottom (mpich2 first...) it all
worked fine...
I know that's not much help. Just make sure that everything that
you're linking together has been compiled using the same compiler
against the same MPI (it's usually best if EVERYTHING uses the mpi
wrappers).
Derek
------------------------------------------------------------------------------
OpenSolaris 2009.06 is a cutting edge operating system for enterprises
looking to deploy the next generation of Solaris that includes the latest
innovations from Sun and the OpenSource community. Download a copy and
enjoy capabilities such as Networking, Storage and Virtualization.
Go to: http://p.sf.net/sfu/opensolaris-get
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users
