Re: Progress rule vs. MPI_DELIVER

Eric Salo (salo@mrjones.engr.sgi.com)
Tue, 16 Apr 1996 09:37:19 -0700

> 1) A signal handler must be invoked at the target
> task when MPI_PUT is called.

I agree that this is a totally disgusting solution for working within a single
SMP.

> 2) Another alternative is to add an MPI_DELIVER function:

Adding such a function to the receiving end of a 1-sided operation makes it a
2-sided operation, so at that point we might as well just use MPI_Recv.

Perhaps a better solution would be to split the sending of data from the
notification of the remote agent? That way, you could do somthing like this:

MPI_Put()
MPI_Put()
MPI_Put()
.
.
.
MPI_Flush_remote_put_data()

You still have the ugliness of invoking a remote agent, but at least it's now
possible to spread the pain around a bit among all of the put calls. Just a
thought...

-- 
Eric Salo         Silicon Graphics Inc.             "Do you know what the
(415)933-2998     2011 N. Shoreline Blvd, 7L-802     last Xon said, just
salo@sgi.com      Mountain View, CA   94043-1389     before he died?"