[Mpi-forum] Discussion points from the MPI-<next> discussion today
nmm1 at cam.ac.uk
Sat Sep 22 05:24:47 CDT 2012
On Sep 22 2012, Supalov, Alexander wrote:
>Indeed. The road goes on and on. :)
Yes, but the real questions are:
a) Is this a sensible road to follow, or does it lead to perdition?
It isn't always a good way to provide functionality - vide Fortran versus
C++, where they are comparably flexible, but the latter is 3-5 times as
large as the former.
b) Would it be better to establish defined halting places, for the
majority of users and a very large number of implementors who do not want
to follow the path to the end? For the majority of uses, one can achieve
99% of the functionality for 10% of the implementation effort.
I haven't investigated implementations in any depth recently, but recent
quick looks confirms that quite a few MPI features are NOT available in
all of the major ones. I know that is true for heterogeneous systems,
and the more extreme threading options, but there are quite a few other
areas where I should be flabberghasted if it were not. And, of course,
at least some of the almost-unused features don't actually work quite
right, but don't get fixed because nobody gives a damn ....
More information about the mpi-forum