Version française
Home     About     Download     Resources     Contact us    
Browse thread
linear algebra libraries?
[ Home ] [ Index: by date | by threads ]
[ Search: ]

[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
Date: -- (:)
From: Markus Mottl <mottl@m...>
Subject: Re: linear algebra libraries?
On Fri, 12 Jan 2001, Brian Rogoff wrote:
> I remember using Meschach many years ago in a different life, and it was
> pretty good for "non-huge" matrices. You can find it at 
> 
> ftp://ftpmaths.anu.edu.au/pub/meschach/meschach.html

Thanks for the hint - unfortunately, it seems that Meschach is not
actively maintained anymore. The last update was more than 5 years ago. I
wouldn't want to interface to a library which gets out of use...

> It's C, not Fortran. Should be easy enough to hook to OCaml.

I haven't tried it yet (will do now), but it is probably easy enough
to interface to Fortran, too (thanks to the Bigarray module). I'll give
LAPACK and BLAS a go...

> I think it would be more interesting to write a native OCaml library. Christophe 
> Rafalli has already started something in numerical linear algebra but its
> just a start, at 
> 
> http://www.lama.univ-savoie.fr/sitelama/Membres/pages_web//RAFFALLI/formel.html

I have already seen it before - it is really nice, especially, because
it specifies algebraic structures on a very high level. I could imagine
that one can easily integrate efficient libraries into this framework.

> I don't work in this field any more but if you want to specify the interfaces
> and need someone to implement some of the algorithms I can find some time
> to hack.  

Thanks for the offer! If you (and potential other contributors) think it
is a good idea (maybe there are still better options), we could register
a project on SourceForge to interface the BLAS and LAPACK libraries to
OCaml. Since I am usually not too much into this topic, I'll probably
only implement the functionality I need (these libraries are *huge*),
but this should be enough to give the project a start.

If nobody has any counter-arguments, I'll register the project at
SourceForge on Monday...

Best regards,
Markus Mottl

-- 
Markus Mottl, mottl@miss.wu-wien.ac.at, http://miss.wu-wien.ac.at/~mottl