Re: [mpi-21] ABI - call for working group

[Greg Bronevetsky <bronevetsky1@xxxxxxxx>]

> - When using a morph layer, there will likely be an extra step of
> having the customer compile the morph layer against their desired
> MPI.  However, this may have two unexpected benefits:
>   1. Greatly reduce the possibility of accidentally using an
> unqualified MPI
>   2. It makes the process "just difficult enough" to discourage all
> but those who really do want to use their own MPI (i.e., they have to
> compile the morph layer)

Is it actually necessary for the user to compile the morph layer for 
their own MPI? It should be possible to write a morph layer that gets 
automatically compiled for any MPI you throw at (it should all be 
based on the given MPI's header file). It can then use dynamic 
linking to hook up the application to the appropriate morph 
conversion layer for the native MPI. I don't think that the user 
would have to be involved in the process except to identify the 
desired MPI implementation. Also, if morph layers are used, what 
would it mean to have an unqualified MPI?

Greg Bronevetsky
Post-Doctoral Researcher
1028 Building 451
Lawrence Livermore National Lab
(925) 424-5756
bronevetsky1@xxxxxxxx