Contributing to GSL

Contributing to GSL

[Varad Deshmukh]

Hi,
I am Varad Deshmukh, a Master's student at UCSB majoring in Computer 
Science. My research interests are in Scientific Computing and HPC, and 
I am interested in contributing to GSL. I recently took up a course in 
numerical simulation and I figured a good way to start would be looking 
into ODE solvers.

The TODO list mentions work to be done for adding other stepper
methods - especially the BDF methods for DAE's implemented in DASSL and 
DASPK. Could you please elaborate more on which new methods you want to 
focus on? I have been reading the code for the Runge-Kutta methods - the 
driver, evolve and control formats - and have a rough idea about the design.

It would be great if I could get some help regarding GSL development in 
general and ode-initval2 in particular. If you feel that I should work 
with smaller bugs and fix them before going into a new implementation, I 
would like to hear suggestions.

Thank you,
Varad.

Re: Contributing to GSL

[Tuomo Keskitalo]

Hello,

the TODO in ode-initval2 directory is slightly outdated. A BDF method 
has already been implemented in msbdf stepper in gsl-1.15. The selection 
of steppers in ode-initval2 is pretty nice, I think, as far as dense 
general-purpose ode-iv methods go. I sometimes thought about trying to 
implement Radau 5 method described in the second book (Hairer, E., 
Wanner, G., Solving Ordinary Differential Equations II: Stiff and 
Differential-Algebraic Problems) but never did. If you want to fight 
with ODEs, you could try that. However, I cannot promise I'd have much 
time to contribute.

DAEs (differential-algebraic equations) are a different story, even 
though they're close to ODEs. There's currently no DAE solver framework 
in GSL. It is something that would be very nice to have, though. If 
anyone is willing to try to design and implement one, feel free to do it!

BR,
Tuomo

On 03/28/2012 07:11 AM, Varad Deshmukh wrote:

> Hi,
> I am Varad Deshmukh, a Master's student at UCSB majoring in Computer
> Science. My research interests are in Scientific Computing and HPC, and
> I am interested in contributing to GSL. I recently took up a course in
> numerical simulation and I figured a good way to start would be looking
> into ODE solvers.
>
> The TODO list mentions work to be done for adding other stepper
> methods - especially the BDF methods for DAE's implemented in DASSL and
> DASPK. Could you please elaborate more on which new methods you want to
> focus on? I have been reading the code for the Runge-Kutta methods - the
> driver, evolve and control formats - and have a rough idea about the
> design.
>
> It would be great if I could get some help regarding GSL development in
> general and ode-initval2 in particular. If you feel that I should work
> with smaller bugs and fix them before going into a new implementation, I
> would like to hear suggestions.
>
> Thank you,
> Varad.
>


-- 
Tuomo.Keskitalo@iki.fi
http://iki.fi/tuomo.keskitalo