Mega-apologies for that. But at least we did send it out before the beginning
of the meeting! :-)
I'm not sure that I see the differences w.r.t. the extra_state. In Raja's
proposal, there can only be one active instance of a GR; after you start it,
you cannot start a subsequent one until after the first one has been both
completed and tested/waited upon. While it is true that the extra_state pointer
itself is only set once, the user is certainly free to change whatever might be
at the other end of that pointer between invocations. Depending on the
application, this might even be something that could be done by the finish_fn.
In the worst case, we could still get the functionality of the current mainline
proposal by adding one level of indirection and having the extra_state point to
another pointer. The down side of that is a very small bit of extra overhead
for the extra hop. The up side is that the new GR proposal gets to use the same
MPI_Start() that we all know and love from MPI-1.
-- Eric Salo Silicon Graphics Inc. "Do you know what the (415)933-2998 2011 N. Shoreline Blvd, 8U-802 last Xon said, just salo@sgi.com Mountain View, CA 94043-1389 before he died?"