Re: Dynamic Process comment: using MPI_Init() without mpirun

Gary Oberbrunner (garyo@avs.com)
Mon, 29 Apr 1996 18:44:25 -0400

> Date: Mon, 29 Apr 1996 15:30:35 -0700
> From: "William C. Saphir" <wcs@nas.nasa.gov>
>
> This isn't to say we shouldn't make it possible to do what you want
> to do, just that we shouldn't make it too easy, and we shouldn't
> encourage it for applications that don't need it.

I'll buy that.

> Sounds like you have a good reason to spawn processes. Given the
> current MPI-2 draft, I see no barrier to this. There is nothing
> preventing your original process from "becoming" an MPI process by
> calling MPI_Init().

Except it's not at all portable currently. I don't even know if it
would work in any current MPI implementation (since this behavior is
currently undefined). It does happen to work in LAM 6.0, I just tried
it. :-)

> It may be that current implementations require it to be started by
> mpirun, but this is just an implementation detail, [n]ot a
> fundamental problem. Then you can spawn the rest of the processes
> with MPI_Spawn. So I understand correctly?

Yup, exactly. All I really want to do is get a sentence in the
standard that says what happens if you call MPI_Init() with no mpirun
or spawn above it (it becomes a singleton, assuming the implementation
supports dynamic processes).

-- 
Gary Oberbrunner			garyo@avs.com
Advanced Visual Systems, Inc.		http://www.avs.com/~garyo
300 Fifth Avenue			(617)890-8192 x2133  TEL
Waltham, MA 02154			(617)890-8287	     FAX