[Ocaml-biz] BLAS/LAPACK

[Brian Hurt]

On Sat, 11 Sep 2004, Brandon J. Van Every wrote:

> Brian Hurt wrote:
> >
> > The problem with BLAS/LAPACK is everyone has them.  Or at
> > least C, C++,
> > and Fortran have them.  So there is no signifigant advantage to using
> > Ocaml over those languages.
> 
> Well, the migration argument would have to be made on other grounds.
> But "yes you can still do BLAS/LAPACK under OCaml" is a good sales
> pitch.  I don't think anyone's looking to have their BLAS/LAPACK
> replaced.

And they'll reply "so why I am rewritting working code?"  There are two 
real "markets" I'm looking at here.  The first is code which needs to be 
rewritten anyways because it's hit the limitations of the original 
language- the Matlab people, in other words.  The second is new code.

One thing BLAS/LAPACK don't do is handle sparse matricies.  They were 
originally designed for vector machines, and work on dense matricies 
(including banded, triangular, etc.).  But sparse matricies are becomming 
more and more important.  And when you scratch the surface of sparse 
matrix handling, you discover that the best implementations are complex 
data structures- one of the places Ocaml really shines.

-- 
"Usenet is like a herd of performing elephants with diarrhea -- massive,
difficult to redirect, awe-inspiring, entertaining, and a source of
mind-boggling amounts of excrement when you least expect it."
                                - Gene Spafford 
Brian