Do we really want Smalltalk code, or a wrapper around a C library? I've been tackling GSL, but callbacks+ffi have gotten strange. Still, it seems that for 500k element FFTs and other tricks, C _has_ to be faster than what we can create in Smalltalk.
I am not at all thrilled about GSL's license. Maybe there is a better choice? To its credit, it is fairly full featured, including wavelet transforms, which are very useful. Bill ________________________________________ From: pharo-project-boun...@lists.gforge.inria.fr [pharo-project-boun...@lists.gforge.inria.fr] on behalf of Serge Stinckwich [serge.stinckw...@gmail.com] Sent: Wednesday, March 28, 2012 1:24 AM To: Pharo Development; Moose-related development Subject: Re: [Pharo-project] SciSmalltalk BTW, i'm looking for some Smalltalk code implementing Runge-Kutta methods for solving a set of ordinary differential equations (ODEs). Regards, On Wed, Mar 28, 2012 at 8:38 AM, Serge Stinckwich <serge.stinckw...@gmail.com> wrote: > Dear all, > > we already discuss about that in the moose and pharo mailing-list. > Maybe this is too late, but please find a small proposal for gsoc 2012 below. > > ================================================================ > > Name: SciSmalltalk > Level: Intermediate > Possible mentor: Serge Stinckwich > Possible second mentor: ? > > Description > Smalltalk has at that time no equivalent to mathematical libraries > like NumPy, SciPy (Python) or SciRuby (Ruby). > The goal of the SciSmalltalk project is to develop an open-source > library of mathematical for the Smalltalk programming language (MIT > Licence). > > Technical Details > The development of this project is to be done in Pharo Smalltalk, but > the code should be portable to other Smalltalk flavors. > Numerous Smalltalk projects provide already some basic functionalities > (complex and quaternions extensions, random number generator, fuzzy > algorithms, LAPACK linear algebra package, Didier Besset's numerical > methods, ...). A first task will be to do an audit of all the existing > projects that provide some mathematical stuff and build a Pharo > Configuration to load them in a fresh Pharo Smalltalk image. After > that, the student help by his/her mentors will decide what are the > numeric algorithms to develop in priority. > > The student will need to know some basic numeric algorithms usually > found in such libraries. > Units tests should also be provided. > > Benefits to the Student > The student will help the Smalltalk community in a very concrete way. > The student will learn to design well-designed code with tests. > > Benefits to the Community > Having a package providing more elaborate numeric libraries is really > important to develop the use Smalltalk in new domains (robotics, high > performance computing, computer vision, bio-computing, ...). The lack > of numeric librairies hamper the use of the Smalltalk in a scientific > context at the moment. An another goal of this project is to develop a > community of people interested by these topic. > > Regards, > -- > Serge Stinckwich > UMI UMMISCO 209 (IRD/UPMC), Hanoi, Vietnam > Every DSL ends up being Smalltalk > http://doesnotunderstand.org/ -- Serge Stinckwich UMI UMMISCO 209 (IRD/UPMC), Hanoi, Vietnam Every DSL ends up being Smalltalk http://doesnotunderstand.org/
