itaps-parallel Mesh Instances

[Mark Miller]

This is consistent with my understanding of what Jason proposed in the way
of a multi-plexor also based on my review of the code example he sent.

And, doesn't this mean that 'co-existance' cannot be solved by a multiplexor?
Or, does it only mean that it cannot be solved by the currently proposed
multiplexor?

Mark

PS: There are other issues with the currently proposed multiplexor such
as the requirement for shared libraries in order for it to work that are some
big iron may not be feasible. I want to make sure we don't forget about
that.

Carl Ollivier-Gooch wrote:

> Actually (and I think I can trust Jason to correct me if I'm wrong),
> what the multiplexer does is make it possible for two -different-
> implementations but (necessarily, in the C API) the -same- function
> names to coexist.  The multiplexer basically wraps the two different
> implementations, and passes calls through to the correct implementation
> based on the mesh instance handle.
>
> I'm sure I butchered the details there, but hopefully I got the big
> picture about right....
>
> In any event, yes, I agree that a single instance can't be both an A
> implementation and a B implementation simultaneously, although a single
> -function- from the interface can (as I sort of described above) branch
> to calls to that same function in an A or a B implementation, depending
> on the mesh instance.  (Yes, at least an implicit registration of
> instances is required for this to work smoothly, methinks...)

--
Mark C. Miller, Lawrence Livermore National Laboratory
email: mailto:miller86 at llnl.gov
(M/T/W) (925)-423-5901 (!!LLNL BUSINESS ONLY!!)
(Th/F)  (530)-753-8511 (!!LLNL BUSINESS ONLY!!)