[Beowulf] The Case for an MPI ABI
mark.westwood at ohmsurveys.com
Tue Feb 22 03:23:42 EST 2005
You make a very persuasive case for an ABI. As an end-user of MPI, and
with no ambitions to be anything else, many of the benefits of an ABI
you suggest would be very useful. My recent experience of porting our
MPI / Fortran codes to other platforms has been that getting the code to
compile has been almost trivial (replace 'call flush(6)' by 'call
flush_(6)' a few times, that sort of thing) but that getting to grips
with the foreign environment (memory management, job submission, job
start-up) is a real pain.
So, to all you salesman in the group, come back and try to sell me the
ABI when it's ready.
Greg Lindahl wrote:
> Those of you who were at the Open IB conference last week saw me give
> a talk entitled "The Case for an MPI ABI". It seems that Patrick and I
> have been channeling each other AGAIN; see what happens when I move to
> The first question is: Does an ABI provide enough benefit for people
> to care? To care enough to sit on a committee?
> If the answer is "yes", then I think we'll have one. The minimum
> technical issues revolve around the contents of <mpi.h> and the names
> of shared libraries. The amount of work for MPICH or OpenMPI to
> support that part of an ABI is modest.
> If we wanted to go farther, I have a strawman proposal which addresses
> a generic startup procedure which would allow user applications, MPI
> implementations, and queue systems to all live in peace and harmony.
> This talk:
> mostly talks about why we need an ABI, who wins and loses as a result
> of having one, and the pieces that could be in it. Please give it a
> -- greg
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
The Technology Centre
Offshore Technology Park
+44 (0)870 429 6586
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
More information about the Beowulf