Re: Octave parallel computing toolbox

[Ganesh Bikshandi]

We plan to release the Octave/Matlab port as a toolbox. There will be
a pure Octave/Matlab layer, that users are exposed to. The layer will
internally make calls to our existing library (in C++ -- release under
EPL), thru mex interface. That library itself makes calls to IBM
Copyrighted lbraries, internally. So there is a whole lot of licensing
issues, that I am not completely aware of. I have to talk to legal
advisers of my company to resolve those issues.

Form the GNU GPL FAQ, I have inferred that there is no licensing issue
if we use the toolbox "privately" without releasing. So, we can use
Octave for development and testing of our toolbox.  I hope this should
not pose any problem, when we later distribute the toolbox under EPL
for Matlab users. (Octave users can use it too, but not redistribute
it , right?)

--Ganesh


On 1/25/08, Ben Abbott <address@hidden> wrote:
> How is the communication between your library and Octave/Matlab done?
>
> Ben
>
> On Thursday, January 24, 2008, at 02:08PM, "Ganesh Bikshandi" 
> <address@hidden> wrote:
> >The current library (which is under active development) is released
> >under Eclipse public license. But, I am not sure about the Octave or
> >Matlab port. That is part of a incubation program in the company I am
> >working for (IBM).  When the project materializes, I can give a clear
> >picture about this. Now, the project is only in the pre-selection
> >phase. A good justification is required for the project to be
> >selected. I thought that since Octave does not seem to have good
> >parallel programming interface (like Matlab), that is a sound
> >justification to push my proposal.  (Correct me, if Octave already has
> >a deluge of Parallel programming tools like Matlab).
> >
> >We want this stuff to run in IBM hardware (like Bluegene) and we don't
> >want to be bogged down by any source code or licensing issues. Plus we
> >want to run in SPMD mode (atleast in the initial phases of
> >development). That means we want N Octave instances running, where N
> >is the number of compute nodes. Using Octave we overcome all the
> >licensing, source code and portability limitations. (IMHO).
> >
> >On Jan 24, 2008 11:30 AM, John W. Eaton <address@hidden> wrote:
> >>
> >> On 24-Jan-2008, Ganesh Bikshandi wrote:
> >>
> >> | Dear Octave Developers/Users,
> >> |
> >> | We have developed a  parallel programming library intended for (C or
> >> | F77) + MPI programmers. The library is more productive and offers
> >> | better performance than MPI.
> >> |
> >> |  We are planning to have a MATLAB port of the library. But, since
> >> | MATLAB has several  parallel programming tools (Mathworks themselves
> >> | ship a Distributed Computing Toolbox) which provide adequate support
> >> | for parallel programming, we are now pondering over porting to OCTAVE
> >> | instead. We want to know the state of the art in OCTAVE parallel
> >> | programming. We need a strong business case for our project. If OCTAVE
> >> | does not have capabilities similar to MATLAB, then poring our library
> >> | to OCTAVE will result in a new product. In this regard, if you can let
> >> | us know the current support for OCTAVE parallel programming and the
> >> | necessity of parallel programming tools in OCTAVE, that will give us a
> >> | clear picture.
> >>
> >> It is Octave, not OCTAVE.
> >>
> >> What are the license terms for your library  How will it be linked
> >> with Octave?
> >>
> >> jwe
> >>
> >_______________________________________________
> >Help-octave mailing list
> >address@hidden
> >https://www.cae.wisc.edu/mailman/listinfo/help-octave
> >
> >
>