[O-MPI users] Re: [Beowulf] Alternative to MPI ABI

[Jeff Squyres]

On Mar 24, 2005, at 8:34 PM, Greg Lindahl wrote:

> A similar idea was actually written up by Bill Gropp and was mentioned
> by him 5 weeks ago on the beowulf list. Quoth he:
>
> | I wrote a paper that appeared in the EuroPVMMPI'02 meeting that 
> discusses
> | the issues of a common ABI.  The paper is "Building Library 
> Components That
> | Can Use Any MPI Implementation" and is available as
> | http://www-unix.mcs.anl.gov/~gropp/bib/papers/2002/gmpishort.pdf .
>
> I think this overall idea falls short of the benefit of an ABI for a
> couple of reasons. The first is that it's unlikely to get wide
> distribution if it doesn't come with MPI implementations. The second
> is that it's harder to maintain "out of tree"; the minute that an MPI
> implementation changes something, MorphMPI is going to be broken.

Yes, I read this paper.  I freely admit that my post was inspired by 
this paper (and several other factors).  Mea culpa for not citing it 
(sorry Bill! :-( ).   I'm just widening the scope of the ideas a 
little, and suggesting that a bright MS student can actually go 
implement it with *FAR* less work than trying to do an MPI ABI and in a 
*MUCH* shorter timescale.

Software dependencies are a fact of life.  But also consider that MPI's 
don't change their innards frequently (or at all).  If an 
implementation chooses integers for MPI handles, for example, they'll 
stay with integers -- they won't suddenly change to pointers between 
version a.b.c and a.b.(c+1).

So the perturbation rate is actually quite low; a MorphMPI that relies 
on MPI handles being integers for a specific MPI implementation (for 
example) would be relatively stable.

>> 2. Cancel the MPI Implementor's Ultimate Prize Fighting Cage Match on
>> pay-per-view (read: no need for time-consuming, potentially fruitless
>> attempts to get MPI implementors to agree on anything)
>
> Was there a big fight over the Fortran interface? It nails down the
> types because it has to.

I was not involved in MPI-1, so I cannot say.

> All the ABI does is make C look like Fortran;
> no internals need change in any implementation.

That is an extremely inaccurate statement.  :-)

-- 
{+} Jeff Squyres
{+} The Open MPI Project
{+} http://www.open-mpi.org/