Re: [Gcl-devel] Re: MPI

[Camm Maguire]

Greetings!  OK, I've verified that gcl-mpi is working on my cluster,
at least to the extent of running the example programs.  But I think
there is an error somewhere in the master-slave.lsp file -- namely
when curr-tag wraps back to zero once it exceeds broadcast-tag.  Here
is what I get on a large (par-eval '(myfactor-par ...)):

TRACE: 12 ->: NIL
TRACE: -> 12: 9998
TRACE: 13 ->: NIL
TRACE: -> 13: 9999
TRACE: 15 ->: NIL
TRACE: -> 15: 10000
TRACE: 1 ->: NIL
TRACE: -> 1: 10001
TRACE: 2 ->: NIL

Error: Previous task appears to not yet have completed:
       2

broad-cast-tag is 10000, and for some reason the pending array is not
getting cleared properly.  Commenting out the error check here allows
the code to proceed normally.

Advice appreciated.

Take care,

Camm Maguire <address@hidden> writes:

> Greetings, and thanks for the query.
> 
> I found this at http://www-unix.mcs.anl.gov/mpi/libraries.html:
> 
>   A beta version of GCL/MPI is now available by anonymous ftp. This is
>   a free software package built on top of MPI and GCL LISP, to easily
>   write parallel programs on networks of workstations. GCL/MPI is
>   intended as an easy-to-use master-slave distributed architecture. It
>   combines the feedback of an interactive language (the GCL or AKCL
>   dialect of LISP) with the the use of MPI (Message Passing Interface)
>   to take advantage of networks of workstations. As such, it is hoped
>   that it will make available an SPMD architecture that helps people
>   overcome the initial learning barrier in writing parallel
>   programs. Ease-of-use is emphasized while hoping to maintain
>   reasonable efficiency and a reasonable feature set. This
>   distribution, along with a paper describing it is available by
>   anonymous ftp in the directory /pub/people/gene/starmpi at
>   ftp.ccs.neu.edu . The current implementation is based on MPI and GCL
>   or AKCL, but it should be easily portable to other message passing
>   libraries (such as PVM) and other dialects of Common LISP with a
>   foreign function interface capable of loading object (.o) files and
>   library archive (.a) files. This version is still
>   experimental. Comments are gratefully accepted. Thanks to Gene
>   Cooperman .
> 
> I've downloaded the source, and have gotten the thing to compile with
> a few changes.  I have a cluster here, and will be trying to test it
> out.  If the licensing works out, I'd like to add this as an
> autoloadable module to the distribution, given the popularity of
> clustering and the relative advantage it could give to gcl.  And I
> have some reasonable mpi experience myself, so its something I can do.
> 
> Thoughts from the list welcome.
> 
> This also brings up the general question of 3rd party library
> interfaces.  If gcl could in theory do something uniquely well for the
> free lisp world, this surely should be it, given its close
> relationship to the C compiler.  
> 
> As faslink has not been implemented for Linux yet, right now the way
> one does this is to write a lisp wrapper for the library, compile,
> link using the new LINK function I added for maxima, exit and restart.
> One no longer needs the source tree.
> 
> Improvements could be the elimination of the need for the wrapper, and
> using dlopen to access the library in the live image, with this
> dynamic link being preserved on save-system.  I don't know if either
> are even possible, but they would preserve the incremental nature of
> lisp, it seems.
> 
> Coming from the C world, the idea of having to write a wrapper for
> every 3rd party library one wants to use will permanently retard the
> use of lisp, I'd think.  Reimplementing everything in lisp, while in
> theory possible and even appealing, is not practical given the volume
> of high quality free software now available.
> 
> Take care,
> 
> 
> root <address@hidden> writes:
> 
> > Camm,
> > 
> > We have just installed a beowulf cluster and I want to convince
> > the university to use GCL. Do you know of a working MPI extension?
> > 
> > Tim
> > 
> > 
> 
> -- 
> Camm Maguire                                          address@hidden
> ==========================================================================
> "The earth is but one country, and mankind its citizens."  --  Baha'u'llah
> 
> 
> _______________________________________________
> Gcl-devel mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/gcl-devel
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah