Hi,
What I usually do is having in the MPI ensemble one GUI node
which consists of (at least) 2 threads, one doing the MPI stuff,
one doing the GUI stuff, and use some mutexes and stuff to
synchronize these two threads, where necessary.
Regards, Dominik.
On Thu, Aug 05, 2004 at 11:08:25AM -0600, John Wohlbier wrote:
> Hi,
>
> I am trying to develop an MPI application with a GUI
> using wxWidgets (http://wxWidgets.org). The approach
> that I guessed would work was to have the master process
> handle the GUI, and menu events would broadcast to
> to the slave processes.
>
> The attached code (main.cpp) is what I came up with for starters.
> What I intend the code to do is to initialize the MPI
> environment, and then exit each process, where
> the MPI::Finalize() for the master process is handled
> in the File->Exit function of the GUI window (see frame.cpp).
> What actually happens is that the window is opened and
> none of the other processes DO ANYTHING until the window
> is closed, i.e., MPI::Finalize() is executed for the master
> process. After MPI::Finalize() is called from the master,
> all of the slaves exit as expected. If I include the
> MPI::Finalize() outside of the window (i.e., uncomment the
> call that is commented out in main.cpp),
> all of the processes run. However, this isn't an acceptable
> solution because eventually the window's menu events need
> to be sent to the slave processes.
>
> Does anyone have a general recommendation for what I am
> doing wrong, or what I should be doing?
>
> Thanks!
> jgw
>
>
> p.s. I am actually using MPICH, but I couldn't find an
> MPICH mailing list! I hope this is OK since the question
> seems to be a general MPI question, and not implementation
> dependent.
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
--
|
