In theory, yes, but we want to encourage as many implementations as possible,
right?
While I completely agree with Greg about the evilness of interrupts, I do not
believe that it is at all acceptable for 1-sided communication to depend upon
the receiver making MPI calls to guarantee progress; the term "1-sided" becomes
rather meaningless at that point. I suppose we could come up with some goofy
call like MPI_Make_some_progress() but let's not, please!
I sort of like the idea of not using communicators for shared communication.
Going back to my own counter-proposal (which I guess won't be part of the SC95
distribution), we could easily replace MPI_SHMALLOC() with something like
MPI_CREATE_SHARED_OBJECT() that takes a buffer size and returns a pointer plus
a handle for a shared window. As LLoyd points out, this has implications for
MPI_Barrier() but these are minor.
-- Eric Salo Silicon Graphics Inc. "Do you know what the (415)390-2998 2011 N. Shoreline Blvd, 7L-802 last Xon said, just salo@sgi.com Mountain View, CA 94043-1389 before he died?"