Re: [boost] [GSOC] Are we going to participate?

Subject: Re: [boost] [GSOC] Are we going to participate?
From: David Bellot (david.bellot_at_[hidden])
Date: 2013-02-18 10:58:25

Hi,

I would like too to propose another GSOC project, regarding the improvement
of Boost::uBLAS.
We have an old page with the list of future new features here:
http://sourceforge.net/apps/mediawiki/ublas/index.php?title=Main_Page

Basically, anything from this page is itself a GSOC project but in fact, at
least 2 are more important that the rest and have to be done in that order:

- unify representation of vector and matrices into a unique matrix class.
After all, vectors are 1-row or 1-column matrices. This has 2 advantages:
better representation of the the underlying math concept and laying the
ground for future extensions.
For code compatibility purpose, the vector<> will still exist. It will be
extended so that one can choose if it's a row or column vector. One
consequence of that, will be the implementation of the * operator for
matrices and vectors.
Fixed size vectors and matrices could be another consequence of this new ar
chitecture.

Ideally, I want to write code like this in C++11
vector<double, row, 4>  v1=randn(1,4);
vector<double, column, 4>  v2=randn(4,1);
auto v3 = v1*v2;

- basic architecture for parallel code, vectorized code, GPU code,
multi-core, you name it.

Inspiration from other libraries like Eigen, Armadillo, GotoBLAS, etc... is
highly recommended (after all, that's one of the raison d'etre of Free S
oftware)

Then if time permits, the GSOC student could start doing some SSE or
multi-core code to prove the concept of the architecture.

So basically, the GSOC project is to shuffle Boost::uBLAS architecture a
little bit to make it more modern and way faster.

Best,
David

On Mon, Feb 18, 2013 at 10:42 AM, Paul A. Bristow
<pbristow_at_[hidden]>wrote:

>
>
> > -----Original Message-----
> > From: Boost [mailto:boost-bounces_at_[hidden]] On Behalf Of Karsten
> Ahnert
> > Sent: Monday, February 18, 2013 9:59 AM
> > To: boost_at_[hidden]
> > Subject: Re: [boost] [GSOC] Are we going to participate?
> >
> > Hi,
> >
> > On 02/13/2013 11:06 PM, Boris Schaeling wrote:
> > > Andrew was the GSoC administrator for Boost in the previous years but
> > > unfortunately has no time this year. So I jump in. I just created the
> > > Wiki page <https://svn.boost.org/trac/boost/wiki/SoC2013> and will
> > > send around another email in the coming days. For now I encourage
> > > everyone to brainstorm just like Antony did about what you would like
> > > to see in Boost. Google didn't invite us last year to participate in
> > > the program, and the reason was that we didn't have enough
> > > ideas/projects. So feel free to send ideas to this mailing list or add
> them to the Wiki page.
> > > For now no idea is too stupid - and even if it is, it might help
> > > someone else to get a good idea. :)
> >
> > I would like to propose a GSOC project ideas aiming to implement a new
> feature for odeint. It
> would be
> > really nice to have more implicit ODE solvers in odeint and an extended
> algebra/operations
> mechanism for
> > matrices and solution of linear systems. This would widen the scope of
> the library enormously,
> since it
> > directly addresses partial differential equations (PDEs) and enters the
> field of HPC.
>
> This sounds a good project in that it is conceivably doable in 3 months,
> and can build on the
> existing infrastructure of testing and docs, and does not necessarily
> require a review  (or only a
> mini one).
>
> My experience mentoring the Checks project is that it was possible to get
> to a reviewable state -
> but only because I had set up the infrastructure with a template for
> testing and docs.  I think we
> would have made it to review if the student hadn't decided that there was
> a better way of doing it
> and so re-wrote it substantially!
>
>  - and then his studies intervened ;-(
>
> This suggests to me that revision or rejuvenation or extension of existing
> libraries to produce
> 'Version 2' might provide good projects?  Or picking up orphaned projects
> or libraries  (endian? -
> Beman has more important jobs to do ;-) )
>
> We seem to have a lot of nearly finished projects - and that means a lot
> of wasted work?
>
> Paul
>
> PS This does not exclude 'Blue Skies' projects suggested by Andrew Sutton
> where we can't expect a
> finished project.
> ---
> Paul A. Bristow,
> Prizet Farmhouse, Kendal LA8 8AB  UK
> [phone]  07714330204
> pbristow_at_[hidden]
>
>
>
>
>
>
>
> _______________________________________________
> Unsubscribe & other changes:
> http://lists.boost.org/mailman/listinfo.cgi/boost
>