Fwd: Re: GSoC: Fast Linear Algebra over Extension Fields

87 views
Skip to first unread message

Martin Albrecht

unread,
Feb 25, 2014, 12:23:58 PM2/25/14
to linbox...@googlegroups.com, Dávid Fonyó
Hi Dávid,

this year we're aiming for integrating this project into the LinBox project.
You should probably subscribe to that mailing list (I'm forwarding your
introduction there).

You could - for example - repeat last year's exercise with LinBox's data types
(it should be straight forward to adapt).

Cheers,
Martin

---------- Forwarded Message ----------

Subject: Re: GSoC: Fast Linear Algebra over Extension Fields
Date: Monday 24 Feb 2014, 23:43:12
From: Dávid Fonyó <fony...@gmail.com>
To: lmnd-...@googlegroups.com

Hello,

My name is Dávid Fonyó. I'm learning theoretical mathematics and computer
science. I would like to participate in the Google Summer of Code 2014
program. Last year I submitted a proposal for "Fast Linear Algebra over
Extension Fields" project. But none of us were selected for the program
with this project. I published some sample code too. This year I would like
to try it again. I'm interested in your mathematical algorithms and data
structures project. Is there any priority among them?

Best regards,
Dávid Fonyó



2013. április 15., hétfő 14:38:45 UTC+2 időpontban Dávid Fonyó a következőt
írta:
>
> Hello,
>
> My name is Dávid Fonyó. I'm a third year Mathematics BSc and second year
> Computer Science BSc student in Eötvös Loránd University, Budapest,
> Hungary. I would like to join the Google Summer of Code 2013 program, and
> I'm really interested in the "Fast Linear Algebra over Extension Fields"
> project.
>
> As a prospective mathematician I've learned a lot of things that help me
> understand the mathematical background of algorithms:
> Algebra (4 semesters), Operation research (2 semester), Theory of
> Computation, Numerical methods (2 semester), etc.
> I have a good experience using C++, and I'm familiar with Java, Pascal,
> Ada and Python.
>
> I've already checked the listed algorithms and now I'm reading the
> references. I would like to ask what the next step will be? What kind of
> patch should I have to write?
> Thank you!
>
> Best regards,
> Dávid Fonyó
>

--
--
You received this message because you are subscribed to the Google
Groups "lmnd-devel" group. Visit this group at
http://groups.google.com/group/lmnd-devel?hl=en
lmonade: http://www.lmona.de

---
You received this message because you are subscribed to the Google Groups
"lmnd-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to lmnd-devel+...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
-----------------------------------------
signature.asc

Dávid Fonyó

unread,
Mar 3, 2014, 2:41:23 AM3/3/14
to linbox...@googlegroups.com, Dávid Fonyó, martinr...@googlemail.com
Hi,

I uploaded the exercise to my github repository: https://github.com/fonyodav/GSoC-2014/
Could you help me what I should do now to get my proposal accepted?

Greetings,
Dávid
Message has been deleted

Martin Albrecht

unread,
Mar 3, 2014, 5:04:24 AM3/3/14
to linbox...@googlegroups.com
Hi all,

I'll leave it to the LinBox developers to comment on the exercise. Just some
general comments: to make a strong proposal means to (a) demonstrate that you
can accomplish the task [this is what the exercise is for] and (b) write a
strong proposal. It is fine to discuss your proposal on this list. Your
proposal should include very clear identifiable tasks and milestones that you
intend to complete. To me these include:

- functionality: what do you plan to provide? Don't underestimate that it also
takes time to implement boring/simple stuff like: comparison, addition,
submatrices etc.
- API: what functions will you make available realising this functionality you
can take a look at the LinBox library to lean about API conventions and what
is needed
- test coverage: do you plan to provide full coverage of tests. If so, how
does it fit into LinBox's testing framework
- performance evaluation: how do you plan to test how well your code performs?

It would also make sense for you to look at Brice's code where he implemented
multiplication over GF(p^d) as GF(p)^d previously (cf. the archives of this
list).

That said, if the proposal is accepted in the end is not certain. We (lmonade)
don't know how many slots we'll get, we (linear algebra people) don't know if
other lmonade mentors will support linear algebra getting a slot.

Cheers,
Martin

On Sunday 02 Mar 2014 23:44:31 Dávid Fonyó wrote:
> Hi,
>
> I uploaded the exercise to my github repository:
> https://github.com/fonyodav/GSoC-2014/
> Could you help me what I should do now to get my proposal accepted?
>
> Greetings,
> Dávid
>
>
>
> 2014. február 25., kedd 18:23:58 UTC+1 időpontban Martin Albrecht a
>
> következőt írta:
> > Hi Dávid,
> >
> > this year we're aiming for integrating this project into the LinBox
> > project.
> > You should probably subscribe to that mailing list (I'm forwarding your
> > introduction there).
> >
> > You could - for example - repeat last year's exercise with LinBox's data
> > types
> > (it should be straight forward to adapt).
> >
> > Cheers,
> > Martin
> >
> > ---------- Forwarded Message ----------
> >
> > Subject: Re: GSoC: Fast Linear Algebra over Extension Fields
> > Date: Monday 24 Feb 2014, 23:43:12
> > From: Dávid Fonyó <fony...@gmail.com <javascript:>>
> > To: lmnd-...@googlegroups.com <javascript:>
> > ---
> > You received this message because you are subscribed to the Google Groups
> > "lmnd-devel" group.
> > To unsubscribe from this group and stop receiving emails from it, send an
> > email to lmnd-devel+...@googlegroups.com <javascript:>.

Pascal Giorgi

unread,
Mar 6, 2014, 10:54:03 AM3/6/14
to linbox...@googlegroups.com, Dávid Fonyó, martinr...@googlemail.com
Hi Dávid,

thanks for your exercise, it sounds good to me. 
In order to step further to what we expect in our project it would be nice if you can update your code with the following points:
1) use BLAS-based matrix multiplication available through FFLAS library and BlasMatrixDomain
2) incorportate a subquadratic polynomial multiplication such as Karatsuba's scheme. It could be first done on degree two extension and then extended to extension of degree 2^k.
3) compare the performances against your previous code.

Cheers,
Pascal.

Brice Boyer

unread,
Mar 6, 2014, 11:02:57 AM3/6/14
to linbox...@googlegroups.com

And 4) you may want to compare/improve on the (Toom Cook) method implemented :-)

You received this message because you are subscribed to the Google Groups "linbox-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to linbox-devel...@googlegroups.com.
To post to this group, send email to linbox...@googlegroups.com.
Visit this group at http://groups.google.com/group/linbox-devel.
Reply all
Reply to author
Forward
0 new messages