Newsgroups: comp.parallel.mpi
From: brunner@mandrake.think.com (Eric Brunner)
Subject: libmpi.a vs libmpi.so
Organization: Thinking Machines Corporation, Cambridge MA, USA
Date: 16 Aug 1996 15:38:18 GMT
Message-ID: <4v24la$mor@bone.think.com>

We will be shipping a dual distribution of libmpi, dynamic and static.

I'm curious about whether or not any other mpi vendors ship a dynamic
version (major savings on disk real estate, hvectest went from 3.5MB
to .25MB), and if so, what issues other mpi vendors see in distribution
of mpilib in dynamic form.

Secondly, I'm curious as to just how "complete" an mpich 1.0.x offering
other mpi vendors ship. Toplevel construction results in the following
libraries being constructed (a TMC-centric view to be sure):

	libampi.a
	libfmpi.a
	liblmpi.a
	libmpe.a
	libmpe_nompi.a
	libmpi.a
	libpmpi.a
	libtmpi.a

I would appreciate it if some of you (comp.parallel.mpi subscribers) would
take the time to look at your mpi implementation (providors or users) and
let me know in which form the mpi implementation available to you is in,
and which libs you do use, how, and why.

I'd be happy to summarize what I learn from any responses and post it here
for general information. Thanks in advance!

--
Kitakitamatsinohpowaw,
Eric Brunner

