Re: MPI_TYPE_DUP (Actually, what I really wanted)

Eric Salo (salo@mrjones.engr.sgi.com)
Wed, 11 Dec 1996 13:05:58 -0800

Aren't we all just fooling ourselves if we try to pretend that reference counts
are optional for implementations? A correct implementation of the standard all
but requires them already. Consider:

MPI_Isend(.., my_type, ...)

MPI_Type_free(&my_type)

If the isend completes immediately, then type_free must free the resources
associated with the type. If it is not yet complete, then type_free must *not*
free the resources. Ergo, a reference count is needed.

So maybe we should just give people what they really need and define:

MPI_Type_increment_reference_count(&my_type)

Would this really be a big deal?

-- 
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?"