[Mpi3-abi] Open MPI and ABI

[Jeff Brown]

All,

At the last meeting I attended (Dublin) we 
discussed developing a prototype "morph" ABI 
layer with initial implementation targeting the 
December meeting.  Let's take stock of this 
effort when we meet next week.  As I recall, Ezra 
offered to lead the windows implementation, and 
Alexander linux.  This exercise is meant to 
expose some of the implementation details and 
help drive us toward a detailed standard proposal 
we can take to the larger group.

Ezra and Alexander, please brief us on this effort when we meet next Wednesday.

 From the Dublin meeting:
•MPI ABI standard, separate from MPI API 
standard, but under the umbrella of the MPI Forum
•Dynamic link is the main objective (binary compatibility)
•Startup (mpirun/mpiexec) is a secondary objective
•Include all major language bindings in the ABI 
standard (C, C++, Fortran) as we go
•Tied to MPI 2.1 and forward
•Implementers may choose to comply (or not)
•May be implemented as a morph layer or native
•Initial reference implementation will be a morph layer (tbd who provides)
•May be different per platform (Linux, Windows, …)
•Different voting rules separate from MPI API forum (?)
•Need participation from major implementors in order to make progress

mpi.h: MPICH 2 (1.0.7) as base with mods as 
needed (e.g. max values for compile time values used in declarations)
Proceed with a reference morph layer 
implementation targeting (at least) MPICH 2.0, OpenMPI
Load a different module and it just works
Demonstrate with a ring program
Defer startup consideration

Proof of concept reference morph layer implementations:
•Flesh out details by writing some code
•Linux on X86-64 cluster Red Hat Linux/SUSE
•Windows on X86-64 cluster
•Test against ring program
•Bring back to the December forum meeting

ee ya next week

Jeff 





* 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpi-forum.org/pipermail/mpi3-abi/attachments/20081211/f6023603/attachment.html>