[Beowulf] The Case for an MPI ABI

[Joachim Worringen]

Greg Lindahl wrote:
> The first question is: Does an ABI provide enough benefit for people
> to care? To care enough to sit on a committee?

Unfortunately, the value of an ABI is much reduced by the fact that the 
most important target platform Linux itself has no stable ABI (think of 
libc and other version nightmares). On a OS like Solaris or Windows, 
this is much more of a benefit.

Another problem are i.e. vendor-specific assertions that could conflict. 
A solution for this could be "numerical namespaces" for such extensions, 
but how should they be managed?

And what about the different calling-conventions in Fortran? Different 
library names for each variant? The different symbol names are also a 
problem, but a solvable one if a limited, but sufficient set of 
uppercase-lowercase-underscore permutations is defined.

-- 
Joachim Worringen - NEC C&C research lab St.Augustin
fon +49-2241-9252.20 - fax .99 - http://www.ccrl-nece.de