[R-sig-hpc] Rmpi on NetBSD with OpenMPI

[Kevin.Buckley at ecs.vuw.ac.nz]

> > > I believe that for us, the 'winning' change was
> > >
> > > #ifdef OPENMPI
> > > 	dlopen("libmpi.so.0", RTLD_GLOBAL | RTLD_LAZY);
> > > #endif
> > >
> > > after which we were golden.
> >
> > Exactly what I was rambling on about, that makes it explicit.
> >
> > However, I suppose the question that needs to be asked of you now
> > is, if  that worked so well for you then, how come it is no longer
> > visible (explicit) in the current version of Rmpi.
>
> I quoted from the most current sources of Rmpi.
>
> > Had it still been there, then I believe that this issue would not
> > have arisen?
>
> It is still there.

Wow, so it is!

I was assuming that must have been part of the R code and didn't look
in the C code bits of Rmpi.

Given our Rmpi is compiled with

  --configure-args=--with-Rmpi-type=OPENMPI

you would think that the ifdef would be true.

Suffice to say that, even if it is, things don't work for us.

We had to place an explict load of libmpi.so, by R, into zzz.R to
get symbols from that library to be scoped correctly.

-- 
Kevin M. Buckley                                  Room:  CO327
School of Engineering and                         Phone: +64 4 463 5971
 Computer Science
Victoria University of Wellington
New Zealand