Re: [Axiom-developer] Re: Axiom + High Energy Physics

[Mike Dewar]

You're correct that we originally did numerics via a separate server.
At the time this made sense but with machines being more powerful I
wouldn't try to reproduce this today, except via support for standard
protocols like SOAP and WSDL.  The Windows version of Axiom called NAG
numerical code directly by loading symbols as needed from a DLL, and had
we continued to develop Axiom we would have done the same on Unix.  This
is the approach taken by Maple and Matlab as well, by the way.

Cheers, Mike.

On Thu, Nov 10, 2005 at 09:11:19PM -0500, Bill Page wrote:
> On November 10, 2005 8:16 PM Tim Daly wrote:
> > 
> > well, i don't plan to rewrite EVERYTHING in lisp, despite what I
> > told Bill,
> 
> :o)
> 
> > although in principle you can generate machine-instruction
> > equivalent routines from lisp. Before we gave the system to NAG
> > we worked with them to generate an Axiom-Fortran interface which
> > still exists in the code. I'd like to keep it compatible with the
> > NAG library but i don't have access to that anymore.
> > 
> 
> >From a quick look at the NAG routines that are not compiled in the
> current open source release it looked to me like these assumed
> that these numeric calculations were performed by a separate server
> program that communicates with Axiom through sockets. Am I right
> about this, Tim?
> 
> 
> If that is the case, then what one would need to do in order to
> retain the NAG Axiom coding in it's current form would be to write
> a new numeric server program which would call open source math
> library routines to perform equivalent functions. Of course such
> a program could be written in any convenient language, e.g. C or
> even Fortran (post 2000 version please!).
> 
> In principle such a design seems to make sense because it allows
> some parallelism and also provides the numeric routines with their
> own address space. This also allows the conversion to and from
> Axiom data structures to native (Fortran?) data structures to be
> centralized and done as efficiently as possible.
> 
> Of course it is also possible to interface with external programs
> from Axiom in more high level ways such as Martin Rubey's interface
> to Polymake which is written entirely in Spad and communicates
> simply through files.
> 
> http://wiki.axiom-developer.org/Polymake
> 
> For example one might create such an interface for Octave or even
> Cactus http://www.cactuscode.org/aboutCactus if you are into that
> kind of thing. The main problem has to do with converting to and
> from Axiom's internal data structures and the simplest place to do
> that is in Spad (or Aldor).
> 
> Regards,
> Bill Page.
> 
> 
> 
> 
> _______________________________________________
> Axiom-developer mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/axiom-developer
> 
> ________________________________________________________________________
> This e-mail has been scanned for all viruses by Star. The
> service is powered by MessageLabs. For more information on a proactive
> anti-virus service working around the clock, around the globe, visit:
> http://www.star.net.uk
> ________________________________________________________________________

________________________________________________________________________
This e-mail has been scanned for all viruses by Star. The
service is powered by MessageLabs. For more information on a proactive
anti-virus service working around the clock, around the globe, visit:
http://www.star.net.uk
________________________________________________________________________