Re: CS v2.1.0 and subroutines
Posted: Fri Feb 24, 2012 5:30 pm
Hello,
Interesting to note that problems seem to be related to MPI wrappers. Things are better than they used to be a few years ago regarding alternative mpi wrappers, but when several MPI libraries are installed, it seems that the symbolic links can point to different versions (I had a problem with this when I tried to reproduce your bug).
You can tell the setup file to use MPICH2 with the following installation options:
-compc = usr/bin/gcc
-compF = usr/bin/gfortran
-mpiCompC = none
-mpi = /usr/lib/mpich2 yes no
but things might fail due to symbolic links mixing libraries, even though we try to be robust, so using the MPI wrappers should in theory help here (though they seem to be the cause of the user subroutine problems).
Did you check if you also have the user subroutine problems using the MPICH2 compiler wrapper (mpiCompC = mpicc.mpich2) ?
When using compiler wrappers, you may also want to tru using the wrapper for compc in addition to mpiCompC.
Best regards,
Yvan
Interesting to note that problems seem to be related to MPI wrappers. Things are better than they used to be a few years ago regarding alternative mpi wrappers, but when several MPI libraries are installed, it seems that the symbolic links can point to different versions (I had a problem with this when I tried to reproduce your bug).
You can tell the setup file to use MPICH2 with the following installation options:
-compc = usr/bin/gcc
-compF = usr/bin/gfortran
-mpiCompC = none
-mpi = /usr/lib/mpich2 yes no
but things might fail due to symbolic links mixing libraries, even though we try to be robust, so using the MPI wrappers should in theory help here (though they seem to be the cause of the user subroutine problems).
Did you check if you also have the user subroutine problems using the MPICH2 compiler wrapper (mpiCompC = mpicc.mpich2) ?
When using compiler wrappers, you may also want to tru using the wrapper for compc in addition to mpiCompC.
Best regards,
Yvan