> As for the MPI.OBJECT issue: we completely revamped the Java bindings to
> (a) improve performance (they are now much, much closer to the C
> performance) and to make them better conform to the Java language standard
> terminology. So we are no longer compatible with the old mpiJava library -
> not only because of the APIs, but also because the new implementation is
> MPI-3 compliant.
> > After this I tried some other tests using the MPI.OBJECT datatype, just to
> > discover that it wasn't supported. I saw that it was included in mpiJava
> > 1.2 and the current mpiJava in OMPI is 1.1(?). I guess the java bindings
> > aren't that prioritized?
I don't know how you use MPI.OBJECT, but I have ported some small
programs, which used MPI.OBJECT to the new interface with structures.
You can compile and run a program with the following commands:
mpijavac -d $HOME/mpi_classfiles BcastStructArrayMain.java
mpiexec -np 3 java -cp $HOME/mpi_classfiles BcastStructArrayMain
Perhaps you get an idea how you can port your MPI.OBJECT programs,
when you look at my small example programs.